Last active
June 27, 2016 21:06
-
-
Save stas-sl/e7aa923871d2b2ffc7f64b92718990dd to your computer and use it in GitHub Desktop.
This file has been truncated, but you can view the full file.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| I0623 14:13:34.760556 10365 solver.cpp:280] Solving mixed_lstm | |
| I0623 14:13:34.760573 10365 solver.cpp:281] Learning Rate Policy: fixed | |
| I0623 14:13:34.771200 10365 solver.cpp:338] Iteration 0, Testing net (#0) | |
| I0623 14:14:33.060055 10365 solver.cpp:393] Test loss: 4.63185 | |
| I0623 14:14:33.060313 10365 solver.cpp:406] Test net output #0: loss1/accuracy = 0.483626 | |
| I0623 14:14:33.060333 10365 solver.cpp:406] Test net output #1: loss1/accuracy01 = 0.92 | |
| I0623 14:14:33.060346 10365 solver.cpp:406] Test net output #2: loss1/accuracy02 = 0.726 | |
| I0623 14:14:33.060359 10365 solver.cpp:406] Test net output #3: loss1/accuracy03 = 0.492 | |
| I0623 14:14:33.060370 10365 solver.cpp:406] Test net output #4: loss1/accuracy04 = 0.391 | |
| I0623 14:14:33.060381 10365 solver.cpp:406] Test net output #5: loss1/accuracy05 = 0.353 | |
| I0623 14:14:33.060394 10365 solver.cpp:406] Test net output #6: loss1/accuracy06 = 0.366 | |
| I0623 14:14:33.060405 10365 solver.cpp:406] Test net output #7: loss1/accuracy07 = 0.344 | |
| I0623 14:14:33.060415 10365 solver.cpp:406] Test net output #8: loss1/accuracy08 = 0.46 | |
| I0623 14:14:33.060427 10365 solver.cpp:406] Test net output #9: loss1/accuracy09 = 0.387 | |
| I0623 14:14:33.060438 10365 solver.cpp:406] Test net output #10: loss1/accuracy10 = 0.377 | |
| I0623 14:14:33.060449 10365 solver.cpp:406] Test net output #11: loss1/accuracy11 = 0.347 | |
| I0623 14:14:33.060461 10365 solver.cpp:406] Test net output #12: loss1/accuracy12 = 0.436 | |
| I0623 14:14:33.060472 10365 solver.cpp:406] Test net output #13: loss1/accuracy13 = 0.574 | |
| I0623 14:14:33.060484 10365 solver.cpp:406] Test net output #14: loss1/accuracy14 = 0.667 | |
| I0623 14:14:33.060495 10365 solver.cpp:406] Test net output #15: loss1/accuracy15 = 0.763 | |
| I0623 14:14:33.060506 10365 solver.cpp:406] Test net output #16: loss1/accuracy16 = 0.826 | |
| I0623 14:14:33.060518 10365 solver.cpp:406] Test net output #17: loss1/accuracy17 = 0.901 | |
| I0623 14:14:33.060528 10365 solver.cpp:406] Test net output #18: loss1/accuracy18 = 0.949 | |
| I0623 14:14:33.060554 10365 solver.cpp:406] Test net output #19: loss1/accuracy19 = 0.972 | |
| I0623 14:14:33.060566 10365 solver.cpp:406] Test net output #20: loss1/accuracy20 = 0.987 | |
| I0623 14:14:33.060578 10365 solver.cpp:406] Test net output #21: loss1/accuracy21 = 0.999 | |
| I0623 14:14:33.060590 10365 solver.cpp:406] Test net output #22: loss1/accuracy22 = 1 | |
| I0623 14:14:33.060600 10365 solver.cpp:406] Test net output #23: loss1/accuracy_incl_empty = 0.685363 | |
| I0623 14:14:33.060621 10365 solver.cpp:406] Test net output #24: loss1/accuracy_top3 = 0.80516 | |
| I0623 14:14:33.060638 10365 solver.cpp:406] Test net output #25: loss1/cross_entropy_loss = 1.54183 (* 0.3 = 0.46255 loss) | |
| I0623 14:14:33.060652 10365 solver.cpp:406] Test net output #26: loss1/cross_entropy_loss_incl_empty = 0.947007 (* 0.3 = 0.284102 loss) | |
| I0623 14:14:33.060667 10365 solver.cpp:406] Test net output #27: loss1/loss01 = 0.425258 (* 0.0272727 = 0.0115979 loss) | |
| I0623 14:14:33.060681 10365 solver.cpp:406] Test net output #28: loss1/loss02 = 0.988915 (* 0.0272727 = 0.0269704 loss) | |
| I0623 14:14:33.060695 10365 solver.cpp:406] Test net output #29: loss1/loss03 = 1.60917 (* 0.0272727 = 0.0438865 loss) | |
| I0623 14:14:33.060708 10365 solver.cpp:406] Test net output #30: loss1/loss04 = 1.81765 (* 0.0272727 = 0.0495723 loss) | |
| I0623 14:14:33.060722 10365 solver.cpp:406] Test net output #31: loss1/loss05 = 1.97238 (* 0.0272727 = 0.053792 loss) | |
| I0623 14:14:33.060736 10365 solver.cpp:406] Test net output #32: loss1/loss06 = 2.02873 (* 0.0272727 = 0.055329 loss) | |
| I0623 14:14:33.060750 10365 solver.cpp:406] Test net output #33: loss1/loss07 = 1.9944 (* 0.0272727 = 0.0543926 loss) | |
| I0623 14:14:33.060763 10365 solver.cpp:406] Test net output #34: loss1/loss08 = 1.77259 (* 0.0272727 = 0.0483434 loss) | |
| I0623 14:14:33.060776 10365 solver.cpp:406] Test net output #35: loss1/loss09 = 1.9091 (* 0.0272727 = 0.0520663 loss) | |
| I0623 14:14:33.060791 10365 solver.cpp:406] Test net output #36: loss1/loss10 = 1.92095 (* 0.0272727 = 0.0523895 loss) | |
| I0623 14:14:33.060804 10365 solver.cpp:406] Test net output #37: loss1/loss11 = 1.97178 (* 0.0272727 = 0.0537758 loss) | |
| I0623 14:14:33.060818 10365 solver.cpp:406] Test net output #38: loss1/loss12 = 1.6758 (* 0.0272727 = 0.0457036 loss) | |
| I0623 14:14:33.060832 10365 solver.cpp:406] Test net output #39: loss1/loss13 = 1.32602 (* 0.0272727 = 0.0361641 loss) | |
| I0623 14:14:33.060880 10365 solver.cpp:406] Test net output #40: loss1/loss14 = 1.03335 (* 0.0272727 = 0.0281822 loss) | |
| I0623 14:14:33.060896 10365 solver.cpp:406] Test net output #41: loss1/loss15 = 0.744528 (* 0.0272727 = 0.0203053 loss) | |
| I0623 14:14:33.060910 10365 solver.cpp:406] Test net output #42: loss1/loss16 = 0.555813 (* 0.0272727 = 0.0151585 loss) | |
| I0623 14:14:33.060923 10365 solver.cpp:406] Test net output #43: loss1/loss17 = 0.347406 (* 0.0272727 = 0.00947471 loss) | |
| I0623 14:14:33.060937 10365 solver.cpp:406] Test net output #44: loss1/loss18 = 0.19763 (* 0.0272727 = 0.00538992 loss) | |
| I0623 14:14:33.060950 10365 solver.cpp:406] Test net output #45: loss1/loss19 = 0.123181 (* 0.0272727 = 0.00335947 loss) | |
| I0623 14:14:33.060964 10365 solver.cpp:406] Test net output #46: loss1/loss20 = 0.0715468 (* 0.0272727 = 0.00195128 loss) | |
| I0623 14:14:33.060977 10365 solver.cpp:406] Test net output #47: loss1/loss21 = 0.0064354 (* 0.0272727 = 0.000175511 loss) | |
| I0623 14:14:33.060992 10365 solver.cpp:406] Test net output #48: loss1/loss22 = 0.000157251 (* 0.0272727 = 4.28865e-06 loss) | |
| I0623 14:14:33.061003 10365 solver.cpp:406] Test net output #49: loss2/accuracy = 0.568327 | |
| I0623 14:14:33.061015 10365 solver.cpp:406] Test net output #50: loss2/accuracy01 = 0.964 | |
| I0623 14:14:33.061027 10365 solver.cpp:406] Test net output #51: loss2/accuracy02 = 0.921 | |
| I0623 14:14:33.061038 10365 solver.cpp:406] Test net output #52: loss2/accuracy03 = 0.815 | |
| I0623 14:14:33.061048 10365 solver.cpp:406] Test net output #53: loss2/accuracy04 = 0.629 | |
| I0623 14:14:33.061059 10365 solver.cpp:406] Test net output #54: loss2/accuracy05 = 0.483 | |
| I0623 14:14:33.061070 10365 solver.cpp:406] Test net output #55: loss2/accuracy06 = 0.455 | |
| I0623 14:14:33.061082 10365 solver.cpp:406] Test net output #56: loss2/accuracy07 = 0.409 | |
| I0623 14:14:33.061092 10365 solver.cpp:406] Test net output #57: loss2/accuracy08 = 0.479 | |
| I0623 14:14:33.061105 10365 solver.cpp:406] Test net output #58: loss2/accuracy09 = 0.447 | |
| I0623 14:14:33.061115 10365 solver.cpp:406] Test net output #59: loss2/accuracy10 = 0.392 | |
| I0623 14:14:33.061125 10365 solver.cpp:406] Test net output #60: loss2/accuracy11 = 0.374 | |
| I0623 14:14:33.061136 10365 solver.cpp:406] Test net output #61: loss2/accuracy12 = 0.479 | |
| I0623 14:14:33.061147 10365 solver.cpp:406] Test net output #62: loss2/accuracy13 = 0.583 | |
| I0623 14:14:33.061158 10365 solver.cpp:406] Test net output #63: loss2/accuracy14 = 0.671 | |
| I0623 14:14:33.061169 10365 solver.cpp:406] Test net output #64: loss2/accuracy15 = 0.771 | |
| I0623 14:14:33.061180 10365 solver.cpp:406] Test net output #65: loss2/accuracy16 = 0.829 | |
| I0623 14:14:33.061192 10365 solver.cpp:406] Test net output #66: loss2/accuracy17 = 0.901 | |
| I0623 14:14:33.061203 10365 solver.cpp:406] Test net output #67: loss2/accuracy18 = 0.95 | |
| I0623 14:14:33.061213 10365 solver.cpp:406] Test net output #68: loss2/accuracy19 = 0.972 | |
| I0623 14:14:33.061224 10365 solver.cpp:406] Test net output #69: loss2/accuracy20 = 0.987 | |
| I0623 14:14:33.061235 10365 solver.cpp:406] Test net output #70: loss2/accuracy21 = 0.999 | |
| I0623 14:14:33.061246 10365 solver.cpp:406] Test net output #71: loss2/accuracy22 = 1 | |
| I0623 14:14:33.061257 10365 solver.cpp:406] Test net output #72: loss2/accuracy_incl_empty = 0.730864 | |
| I0623 14:14:33.061271 10365 solver.cpp:406] Test net output #73: loss2/accuracy_top3 = 0.869695 | |
| I0623 14:14:33.061285 10365 solver.cpp:406] Test net output #74: loss2/cross_entropy_loss = 1.23732 (* 0.3 = 0.371196 loss) | |
| I0623 14:14:33.061298 10365 solver.cpp:406] Test net output #75: loss2/cross_entropy_loss_incl_empty = 0.770825 (* 0.3 = 0.231247 loss) | |
| I0623 14:14:33.061312 10365 solver.cpp:406] Test net output #76: loss2/loss01 = 0.221866 (* 0.0272727 = 0.00605089 loss) | |
| I0623 14:14:33.061326 10365 solver.cpp:406] Test net output #77: loss2/loss02 = 0.387539 (* 0.0272727 = 0.0105692 loss) | |
| I0623 14:14:33.061354 10365 solver.cpp:406] Test net output #78: loss2/loss03 = 0.703188 (* 0.0272727 = 0.0191779 loss) | |
| I0623 14:14:33.061369 10365 solver.cpp:406] Test net output #79: loss2/loss04 = 1.14818 (* 0.0272727 = 0.0313141 loss) | |
| I0623 14:14:33.061379 10365 solver.cpp:406] Test net output #80: loss2/loss05 = 1.44089 (* 0.0272727 = 0.039297 loss) | |
| I0623 14:14:33.061389 10365 solver.cpp:406] Test net output #81: loss2/loss06 = 1.65874 (* 0.0272727 = 0.0452384 loss) | |
| I0623 14:14:33.061404 10365 solver.cpp:406] Test net output #82: loss2/loss07 = 1.74339 (* 0.0272727 = 0.0475469 loss) | |
| I0623 14:14:33.061417 10365 solver.cpp:406] Test net output #83: loss2/loss08 = 1.61813 (* 0.0272727 = 0.0441309 loss) | |
| I0623 14:14:33.061431 10365 solver.cpp:406] Test net output #84: loss2/loss09 = 1.73008 (* 0.0272727 = 0.0471839 loss) | |
| I0623 14:14:33.061445 10365 solver.cpp:406] Test net output #85: loss2/loss10 = 1.79506 (* 0.0272727 = 0.0489561 loss) | |
| I0623 14:14:33.061458 10365 solver.cpp:406] Test net output #86: loss2/loss11 = 1.8371 (* 0.0272727 = 0.0501028 loss) | |
| I0623 14:14:33.061472 10365 solver.cpp:406] Test net output #87: loss2/loss12 = 1.52091 (* 0.0272727 = 0.0414794 loss) | |
| I0623 14:14:33.061486 10365 solver.cpp:406] Test net output #88: loss2/loss13 = 1.21411 (* 0.0272727 = 0.0331122 loss) | |
| I0623 14:14:33.061499 10365 solver.cpp:406] Test net output #89: loss2/loss14 = 0.933804 (* 0.0272727 = 0.0254674 loss) | |
| I0623 14:14:33.061512 10365 solver.cpp:406] Test net output #90: loss2/loss15 = 0.668074 (* 0.0272727 = 0.0182202 loss) | |
| I0623 14:14:33.061527 10365 solver.cpp:406] Test net output #91: loss2/loss16 = 0.507575 (* 0.0272727 = 0.0138429 loss) | |
| I0623 14:14:33.061539 10365 solver.cpp:406] Test net output #92: loss2/loss17 = 0.332998 (* 0.0272727 = 0.00908176 loss) | |
| I0623 14:14:33.061553 10365 solver.cpp:406] Test net output #93: loss2/loss18 = 0.182017 (* 0.0272727 = 0.00496409 loss) | |
| I0623 14:14:33.061566 10365 solver.cpp:406] Test net output #94: loss2/loss19 = 0.113026 (* 0.0272727 = 0.00308252 loss) | |
| I0623 14:14:33.061580 10365 solver.cpp:406] Test net output #95: loss2/loss20 = 0.0646569 (* 0.0272727 = 0.00176337 loss) | |
| I0623 14:14:33.061594 10365 solver.cpp:406] Test net output #96: loss2/loss21 = 0.00830641 (* 0.0272727 = 0.000226538 loss) | |
| I0623 14:14:33.061607 10365 solver.cpp:406] Test net output #97: loss2/loss22 = 0.000274394 (* 0.0272727 = 7.48349e-06 loss) | |
| I0623 14:14:33.061619 10365 solver.cpp:406] Test net output #98: loss3/accuracy = 0.824308 | |
| I0623 14:14:33.061630 10365 solver.cpp:406] Test net output #99: loss3/accuracy01 = 0.974 | |
| I0623 14:14:33.061642 10365 solver.cpp:406] Test net output #100: loss3/accuracy02 = 0.964 | |
| I0623 14:14:33.061652 10365 solver.cpp:406] Test net output #101: loss3/accuracy03 = 0.937 | |
| I0623 14:14:33.061663 10365 solver.cpp:406] Test net output #102: loss3/accuracy04 = 0.919 | |
| I0623 14:14:33.061676 10365 solver.cpp:406] Test net output #103: loss3/accuracy05 = 0.917 | |
| I0623 14:14:33.061686 10365 solver.cpp:406] Test net output #104: loss3/accuracy06 = 0.87 | |
| I0623 14:14:33.061697 10365 solver.cpp:406] Test net output #105: loss3/accuracy07 = 0.84 | |
| I0623 14:14:33.061708 10365 solver.cpp:406] Test net output #106: loss3/accuracy08 = 0.789 | |
| I0623 14:14:33.061719 10365 solver.cpp:406] Test net output #107: loss3/accuracy09 = 0.687 | |
| I0623 14:14:33.061730 10365 solver.cpp:406] Test net output #108: loss3/accuracy10 = 0.563 | |
| I0623 14:14:33.061741 10365 solver.cpp:406] Test net output #109: loss3/accuracy11 = 0.508 | |
| I0623 14:14:33.061760 10365 solver.cpp:406] Test net output #110: loss3/accuracy12 = 0.568 | |
| I0623 14:14:33.061771 10365 solver.cpp:406] Test net output #111: loss3/accuracy13 = 0.65 | |
| I0623 14:14:33.061782 10365 solver.cpp:406] Test net output #112: loss3/accuracy14 = 0.713 | |
| I0623 14:14:33.061794 10365 solver.cpp:406] Test net output #113: loss3/accuracy15 = 0.795 | |
| I0623 14:14:33.061805 10365 solver.cpp:406] Test net output #114: loss3/accuracy16 = 0.836 | |
| I0623 14:14:33.061835 10365 solver.cpp:406] Test net output #115: loss3/accuracy17 = 0.908 | |
| I0623 14:14:33.061846 10365 solver.cpp:406] Test net output #116: loss3/accuracy18 = 0.957 | |
| I0623 14:14:33.061857 10365 solver.cpp:406] Test net output #117: loss3/accuracy19 = 0.973 | |
| I0623 14:14:33.061868 10365 solver.cpp:406] Test net output #118: loss3/accuracy20 = 0.987 | |
| I0623 14:14:33.061879 10365 solver.cpp:406] Test net output #119: loss3/accuracy21 = 0.999 | |
| I0623 14:14:33.061892 10365 solver.cpp:406] Test net output #120: loss3/accuracy22 = 1 | |
| I0623 14:14:33.061902 10365 solver.cpp:406] Test net output #121: loss3/accuracy_incl_empty = 0.883092 | |
| I0623 14:14:33.061913 10365 solver.cpp:406] Test net output #122: loss3/accuracy_top3 = 0.96174 | |
| I0623 14:14:33.061926 10365 solver.cpp:406] Test net output #123: loss3/cross_entropy_loss = 0.630206 (* 1 = 0.630206 loss) | |
| I0623 14:14:33.061939 10365 solver.cpp:406] Test net output #124: loss3/cross_entropy_loss_incl_empty = 0.39967 (* 1 = 0.39967 loss) | |
| I0623 14:14:33.061954 10365 solver.cpp:406] Test net output #125: loss3/loss01 = 0.1817 (* 0.0909091 = 0.0165182 loss) | |
| I0623 14:14:33.061969 10365 solver.cpp:406] Test net output #126: loss3/loss02 = 0.236424 (* 0.0909091 = 0.0214931 loss) | |
| I0623 14:14:33.061982 10365 solver.cpp:406] Test net output #127: loss3/loss03 = 0.350202 (* 0.0909091 = 0.0318366 loss) | |
| I0623 14:14:33.061996 10365 solver.cpp:406] Test net output #128: loss3/loss04 = 0.416311 (* 0.0909091 = 0.0378464 loss) | |
| I0623 14:14:33.062010 10365 solver.cpp:406] Test net output #129: loss3/loss05 = 0.454873 (* 0.0909091 = 0.0413521 loss) | |
| I0623 14:14:33.062024 10365 solver.cpp:406] Test net output #130: loss3/loss06 = 0.583978 (* 0.0909091 = 0.0530889 loss) | |
| I0623 14:14:33.062038 10365 solver.cpp:406] Test net output #131: loss3/loss07 = 0.658571 (* 0.0909091 = 0.0598701 loss) | |
| I0623 14:14:33.062048 10365 solver.cpp:406] Test net output #132: loss3/loss08 = 0.741631 (* 0.0909091 = 0.067421 loss) | |
| I0623 14:14:33.062062 10365 solver.cpp:406] Test net output #133: loss3/loss09 = 0.95179 (* 0.0909091 = 0.0865264 loss) | |
| I0623 14:14:33.062077 10365 solver.cpp:406] Test net output #134: loss3/loss10 = 1.21927 (* 0.0909091 = 0.110843 loss) | |
| I0623 14:14:33.062090 10365 solver.cpp:406] Test net output #135: loss3/loss11 = 1.31837 (* 0.0909091 = 0.119852 loss) | |
| I0623 14:14:33.062111 10365 solver.cpp:406] Test net output #136: loss3/loss12 = 1.13646 (* 0.0909091 = 0.103314 loss) | |
| I0623 14:14:33.062130 10365 solver.cpp:406] Test net output #137: loss3/loss13 = 0.989969 (* 0.0909091 = 0.0899972 loss) | |
| I0623 14:14:33.062145 10365 solver.cpp:406] Test net output #138: loss3/loss14 = 0.758606 (* 0.0909091 = 0.0689642 loss) | |
| I0623 14:14:33.062166 10365 solver.cpp:406] Test net output #139: loss3/loss15 = 0.517559 (* 0.0909091 = 0.0470508 loss) | |
| I0623 14:14:33.062180 10365 solver.cpp:406] Test net output #140: loss3/loss16 = 0.424152 (* 0.0909091 = 0.0385593 loss) | |
| I0623 14:14:33.062193 10365 solver.cpp:406] Test net output #141: loss3/loss17 = 0.255278 (* 0.0909091 = 0.0232071 loss) | |
| I0623 14:14:33.062207 10365 solver.cpp:406] Test net output #142: loss3/loss18 = 0.144045 (* 0.0909091 = 0.013095 loss) | |
| I0623 14:14:33.062221 10365 solver.cpp:406] Test net output #143: loss3/loss19 = 0.0864126 (* 0.0909091 = 0.00785569 loss) | |
| I0623 14:14:33.062234 10365 solver.cpp:406] Test net output #144: loss3/loss20 = 0.0533713 (* 0.0909091 = 0.00485194 loss) | |
| I0623 14:14:33.062247 10365 solver.cpp:406] Test net output #145: loss3/loss21 = 0.00578669 (* 0.0909091 = 0.000526063 loss) | |
| I0623 14:14:33.062261 10365 solver.cpp:406] Test net output #146: loss3/loss22 = 9.36713e-05 (* 0.0909091 = 8.51557e-06 loss) | |
| I0623 14:14:33.062273 10365 solver.cpp:406] Test net output #147: total_accuracy = 0.278 | |
| I0623 14:14:33.062284 10365 solver.cpp:406] Test net output #148: total_accuracy_not_rec = 0.146 | |
| I0623 14:14:33.062295 10365 solver.cpp:406] Test net output #149: total_confidence = 0.139813 | |
| I0623 14:14:33.062317 10365 solver.cpp:406] Test net output #150: total_confidence_not_rec = 0.100067 | |
| I0623 14:14:33.552907 10365 solver.cpp:229] Iteration 0, loss = 5.12142 | |
| I0623 14:14:33.552981 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.401961 | |
| I0623 14:14:33.552999 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 14:14:33.553014 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 14:14:33.553025 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 14:14:33.553037 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 14:14:33.553050 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.5 | |
| I0623 14:14:33.553061 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 14:14:33.553074 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.625 | |
| I0623 14:14:33.553086 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 14:14:33.553098 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 14:14:33.553109 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 14:14:33.553122 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.125 | |
| I0623 14:14:33.553133 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 14:14:33.553144 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 14:14:33.553156 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 14:14:33.553169 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 14:14:33.553179 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 14:14:33.553191 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 14:14:33.553203 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 14:14:33.553215 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 14:14:33.553226 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 14:14:33.553237 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 14:14:33.553249 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 14:14:33.553261 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.647727 | |
| I0623 14:14:33.553272 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.735294 | |
| I0623 14:14:33.553289 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.79252 (* 0.3 = 0.537755 loss) | |
| I0623 14:14:33.553305 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.06922 (* 0.3 = 0.320767 loss) | |
| I0623 14:14:33.553319 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.324429 (* 0.0272727 = 0.00884805 loss) | |
| I0623 14:14:33.553333 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.0714 (* 0.0272727 = 0.02922 loss) | |
| I0623 14:14:33.553349 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.60965 (* 0.0272727 = 0.0438995 loss) | |
| I0623 14:14:33.553361 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.86818 (* 0.0272727 = 0.0509504 loss) | |
| I0623 14:14:33.553375 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.90872 (* 0.0272727 = 0.0520559 loss) | |
| I0623 14:14:33.553390 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.53158 (* 0.0272727 = 0.0690431 loss) | |
| I0623 14:14:33.553403 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.5928 (* 0.0272727 = 0.0434399 loss) | |
| I0623 14:14:33.553418 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.88923 (* 0.0272727 = 0.0515244 loss) | |
| I0623 14:14:33.553431 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.70306 (* 0.0272727 = 0.0464472 loss) | |
| I0623 14:14:33.553445 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.09238 (* 0.0272727 = 0.0570648 loss) | |
| I0623 14:14:33.553459 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.44719 (* 0.0272727 = 0.0667417 loss) | |
| I0623 14:14:33.553501 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.30285 (* 0.0272727 = 0.062805 loss) | |
| I0623 14:14:33.553524 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.78017 (* 0.0272727 = 0.0485502 loss) | |
| I0623 14:14:33.553539 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.30083 (* 0.0272727 = 0.0354772 loss) | |
| I0623 14:14:33.553552 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.24533 (* 0.0272727 = 0.0339636 loss) | |
| I0623 14:14:33.553566 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.00064 (* 0.0272727 = 0.0272902 loss) | |
| I0623 14:14:33.553588 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0466472 (* 0.0272727 = 0.0012722 loss) | |
| I0623 14:14:33.553602 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00791877 (* 0.0272727 = 0.000215966 loss) | |
| I0623 14:14:33.553617 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00312984 (* 0.0272727 = 8.53593e-05 loss) | |
| I0623 14:14:33.553633 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000648265 (* 0.0272727 = 1.768e-05 loss) | |
| I0623 14:14:33.553648 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000746471 (* 0.0272727 = 2.03583e-05 loss) | |
| I0623 14:14:33.553663 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000106871 (* 0.0272727 = 2.91465e-06 loss) | |
| I0623 14:14:33.553675 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.5 | |
| I0623 14:14:33.553688 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 14:14:33.553699 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 14:14:33.553711 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.5 | |
| I0623 14:14:33.553722 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 14:14:33.553735 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 14:14:33.553745 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.25 | |
| I0623 14:14:33.553762 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 14:14:33.553773 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.25 | |
| I0623 14:14:33.553786 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 14:14:33.553797 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 14:14:33.553807 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.125 | |
| I0623 14:14:33.553825 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.375 | |
| I0623 14:14:33.553838 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 14:14:33.553848 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 14:14:33.553860 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 14:14:33.553871 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 14:14:33.553885 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 14:14:33.553897 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 14:14:33.553908 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 14:14:33.553920 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 14:14:33.553930 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 14:14:33.553942 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 14:14:33.553953 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.704545 | |
| I0623 14:14:33.553966 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.813725 | |
| I0623 14:14:33.553978 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.49093 (* 0.3 = 0.447279 loss) | |
| I0623 14:14:33.553992 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.892697 (* 0.3 = 0.267809 loss) | |
| I0623 14:14:33.554006 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.41995 (* 0.0272727 = 0.0114532 loss) | |
| I0623 14:14:33.554033 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.667512 (* 0.0272727 = 0.0182049 loss) | |
| I0623 14:14:33.554047 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.932408 (* 0.0272727 = 0.0254293 loss) | |
| I0623 14:14:33.554061 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.885443 (* 0.0272727 = 0.0241484 loss) | |
| I0623 14:14:33.554075 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.02774 (* 0.0272727 = 0.055302 loss) | |
| I0623 14:14:33.554090 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.051 (* 0.0272727 = 0.0559365 loss) | |
| I0623 14:14:33.554105 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.80063 (* 0.0272727 = 0.0491081 loss) | |
| I0623 14:14:33.554118 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.06991 (* 0.0272727 = 0.0564521 loss) | |
| I0623 14:14:33.554132 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.99384 (* 0.0272727 = 0.0543775 loss) | |
| I0623 14:14:33.554147 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.96296 (* 0.0272727 = 0.0535352 loss) | |
| I0623 14:14:33.554160 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.4945 (* 0.0272727 = 0.0680318 loss) | |
| I0623 14:14:33.554174 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 2.14919 (* 0.0272727 = 0.0586143 loss) | |
| I0623 14:14:33.554188 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.408 (* 0.0272727 = 0.0384001 loss) | |
| I0623 14:14:33.554201 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.935762 (* 0.0272727 = 0.0255208 loss) | |
| I0623 14:14:33.554215 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.973234 (* 0.0272727 = 0.0265427 loss) | |
| I0623 14:14:33.554230 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.970905 (* 0.0272727 = 0.0264792 loss) | |
| I0623 14:14:33.554244 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0856895 (* 0.0272727 = 0.00233699 loss) | |
| I0623 14:14:33.554258 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0186307 (* 0.0272727 = 0.000508111 loss) | |
| I0623 14:14:33.554273 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00553131 (* 0.0272727 = 0.000150854 loss) | |
| I0623 14:14:33.554287 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00141805 (* 0.0272727 = 3.8674e-05 loss) | |
| I0623 14:14:33.554301 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000240579 (* 0.0272727 = 6.56125e-06 loss) | |
| I0623 14:14:33.554316 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 8.45916e-05 (* 0.0272727 = 2.30704e-06 loss) | |
| I0623 14:14:33.554328 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.803922 | |
| I0623 14:14:33.554340 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 14:14:33.554352 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 14:14:33.554363 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 14:14:33.554375 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 14:14:33.554386 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 14:14:33.554399 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 14:14:33.554409 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 14:14:33.554421 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 14:14:33.554432 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.5 | |
| I0623 14:14:33.554443 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.375 | |
| I0623 14:14:33.554455 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.25 | |
| I0623 14:14:33.554466 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 14:14:33.554477 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 14:14:33.554489 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 14:14:33.554507 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 14:14:33.554515 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 14:14:33.554530 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 14:14:33.554541 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 14:14:33.554553 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 14:14:33.554564 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 14:14:33.554574 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 14:14:33.554585 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 14:14:33.554596 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.880682 | |
| I0623 14:14:33.554607 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.95098 | |
| I0623 14:14:33.554621 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.644893 (* 1 = 0.644893 loss) | |
| I0623 14:14:33.554635 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.384166 (* 1 = 0.384166 loss) | |
| I0623 14:14:33.554648 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.079225 (* 0.0909091 = 0.00720227 loss) | |
| I0623 14:14:33.554662 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0830121 (* 0.0909091 = 0.00754656 loss) | |
| I0623 14:14:33.554675 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0467799 (* 0.0909091 = 0.00425272 loss) | |
| I0623 14:14:33.554692 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.130574 (* 0.0909091 = 0.0118704 loss) | |
| I0623 14:14:33.554707 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.832513 (* 0.0909091 = 0.075683 loss) | |
| I0623 14:14:33.554720 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.28307 (* 0.0909091 = 0.0257336 loss) | |
| I0623 14:14:33.554733 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.598593 (* 0.0909091 = 0.0544175 loss) | |
| I0623 14:14:33.554746 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.936576 (* 0.0909091 = 0.0851433 loss) | |
| I0623 14:14:33.554760 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.926023 (* 0.0909091 = 0.0841839 loss) | |
| I0623 14:14:33.554774 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.85965 (* 0.0909091 = 0.169059 loss) | |
| I0623 14:14:33.554786 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.91403 (* 0.0909091 = 0.174003 loss) | |
| I0623 14:14:33.554800 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.51265 (* 0.0909091 = 0.137514 loss) | |
| I0623 14:14:33.554813 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.11668 (* 0.0909091 = 0.101517 loss) | |
| I0623 14:14:33.554826 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.619355 (* 0.0909091 = 0.056305 loss) | |
| I0623 14:14:33.554841 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.714942 (* 0.0909091 = 0.0649947 loss) | |
| I0623 14:14:33.554853 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.852556 (* 0.0909091 = 0.0775051 loss) | |
| I0623 14:14:33.554867 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0190559 (* 0.0909091 = 0.00173235 loss) | |
| I0623 14:14:33.554880 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00355098 (* 0.0909091 = 0.000322817 loss) | |
| I0623 14:14:33.554894 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00154537 (* 0.0909091 = 0.000140488 loss) | |
| I0623 14:14:33.554908 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000877499 (* 0.0909091 = 7.97726e-05 loss) | |
| I0623 14:14:33.554922 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000159306 (* 0.0909091 = 1.44823e-05 loss) | |
| I0623 14:14:33.554939 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 0.00013564 (* 0.0909091 = 1.23309e-05 loss) | |
| I0623 14:14:33.554951 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 14:14:33.554976 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 14:14:33.554991 10365 solver.cpp:245] Train net output #149: total_confidence = 0.109761 | |
| I0623 14:14:33.555001 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.108266 | |
| I0623 14:14:33.555022 10365 sgd_solver.cpp:106] Iteration 0, lr = 0.001 | |
| I0623 14:15:19.266675 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.4511 > 30) by scale factor 0.985185 | |
| I0623 14:15:52.274754 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.5472 > 30) by scale factor 0.820855 | |
| I0623 14:17:08.293911 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.9362 > 30) by scale factor 0.834814 | |
| I0623 14:17:29.809190 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.9036 > 30) by scale factor 0.884861 | |
| I0623 14:20:24.121537 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.3355 > 30) by scale factor 0.708625 | |
| I0623 14:20:40.253177 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 56.9096 > 30) by scale factor 0.527152 | |
| I0623 14:20:57.513164 10365 solver.cpp:229] Iteration 500, loss = 5.02442 | |
| I0623 14:20:57.513280 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.459459 | |
| I0623 14:20:57.513300 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 14:20:57.513312 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.25 | |
| I0623 14:20:57.513325 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.75 | |
| I0623 14:20:57.513337 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 14:20:57.513350 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.125 | |
| I0623 14:20:57.513361 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 14:20:57.513373 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 14:20:57.513386 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.125 | |
| I0623 14:20:57.513397 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 14:20:57.513408 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 14:20:57.513420 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.125 | |
| I0623 14:20:57.513432 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 14:20:57.513443 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.25 | |
| I0623 14:20:57.513455 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 14:20:57.513468 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 14:20:57.513478 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 14:20:57.513490 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 14:20:57.513501 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 14:20:57.513514 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 14:20:57.513525 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 14:20:57.513536 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 14:20:57.513547 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 14:20:57.513559 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.653409 | |
| I0623 14:20:57.513571 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.711712 | |
| I0623 14:20:57.513588 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.78541 (* 0.3 = 0.535622 loss) | |
| I0623 14:20:57.513603 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.17627 (* 0.3 = 0.352881 loss) | |
| I0623 14:20:57.513617 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.792957 (* 0.0272727 = 0.0216261 loss) | |
| I0623 14:20:57.513631 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 2.44411 (* 0.0272727 = 0.0666575 loss) | |
| I0623 14:20:57.513644 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.18889 (* 0.0272727 = 0.0324242 loss) | |
| I0623 14:20:57.513659 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.94172 (* 0.0272727 = 0.052956 loss) | |
| I0623 14:20:57.513672 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.49155 (* 0.0272727 = 0.0679513 loss) | |
| I0623 14:20:57.513686 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.43514 (* 0.0272727 = 0.066413 loss) | |
| I0623 14:20:57.513700 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.79543 (* 0.0272727 = 0.0489663 loss) | |
| I0623 14:20:57.513713 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.72213 (* 0.0272727 = 0.07424 loss) | |
| I0623 14:20:57.513727 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.0564 (* 0.0272727 = 0.0560837 loss) | |
| I0623 14:20:57.513741 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.14042 (* 0.0272727 = 0.058375 loss) | |
| I0623 14:20:57.513754 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.76514 (* 0.0272727 = 0.0754129 loss) | |
| I0623 14:20:57.513768 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.29966 (* 0.0272727 = 0.0627181 loss) | |
| I0623 14:20:57.513799 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.70819 (* 0.0272727 = 0.0465871 loss) | |
| I0623 14:20:57.513814 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.34664 (* 0.0272727 = 0.0367266 loss) | |
| I0623 14:20:57.513828 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.00315 (* 0.0272727 = 0.0273586 loss) | |
| I0623 14:20:57.513841 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.18193 (* 0.0272727 = 0.0322345 loss) | |
| I0623 14:20:57.513855 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.729755 (* 0.0272727 = 0.0199024 loss) | |
| I0623 14:20:57.513870 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0136932 (* 0.0272727 = 0.000373452 loss) | |
| I0623 14:20:57.513883 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00239847 (* 0.0272727 = 6.54127e-05 loss) | |
| I0623 14:20:57.513897 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000722501 (* 0.0272727 = 1.97046e-05 loss) | |
| I0623 14:20:57.513911 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000203126 (* 0.0272727 = 5.53981e-06 loss) | |
| I0623 14:20:57.513926 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 4.89618e-05 (* 0.0272727 = 1.33532e-06 loss) | |
| I0623 14:20:57.513937 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.558559 | |
| I0623 14:20:57.513949 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 14:20:57.513962 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 14:20:57.513972 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 14:20:57.513983 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 14:20:57.513994 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 14:20:57.514005 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 14:20:57.514017 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.125 | |
| I0623 14:20:57.514029 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.25 | |
| I0623 14:20:57.514039 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 14:20:57.514050 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 14:20:57.514061 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0 | |
| I0623 14:20:57.514073 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0 | |
| I0623 14:20:57.514084 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 14:20:57.514096 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 14:20:57.514106 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 14:20:57.514117 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 14:20:57.514128 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 14:20:57.514139 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 14:20:57.514150 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 14:20:57.514163 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 14:20:57.514173 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 14:20:57.514185 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 14:20:57.514196 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.715909 | |
| I0623 14:20:57.514209 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.801802 | |
| I0623 14:20:57.514222 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.49167 (* 0.3 = 0.447501 loss) | |
| I0623 14:20:57.514236 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.964172 (* 0.3 = 0.289252 loss) | |
| I0623 14:20:57.514250 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.207342 (* 0.0272727 = 0.00565479 loss) | |
| I0623 14:20:57.514268 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.551789 (* 0.0272727 = 0.0150488 loss) | |
| I0623 14:20:57.514294 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.211737 (* 0.0272727 = 0.00577466 loss) | |
| I0623 14:20:57.514309 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.830071 (* 0.0272727 = 0.0226383 loss) | |
| I0623 14:20:57.514323 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.09761 (* 0.0272727 = 0.0299347 loss) | |
| I0623 14:20:57.514338 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.83724 (* 0.0272727 = 0.0501065 loss) | |
| I0623 14:20:57.514351 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.79621 (* 0.0272727 = 0.0489876 loss) | |
| I0623 14:20:57.514365 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.61365 (* 0.0272727 = 0.0712814 loss) | |
| I0623 14:20:57.514379 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.70172 (* 0.0272727 = 0.0464107 loss) | |
| I0623 14:20:57.514392 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.58875 (* 0.0272727 = 0.0433294 loss) | |
| I0623 14:20:57.514406 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.7115 (* 0.0272727 = 0.0739501 loss) | |
| I0623 14:20:57.514416 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 2.77846 (* 0.0272727 = 0.0757761 loss) | |
| I0623 14:20:57.514431 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.08562 (* 0.0272727 = 0.0296077 loss) | |
| I0623 14:20:57.514446 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.69045 (* 0.0272727 = 0.0461033 loss) | |
| I0623 14:20:57.514458 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.02518 (* 0.0272727 = 0.0279595 loss) | |
| I0623 14:20:57.514472 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.8921 (* 0.0272727 = 0.02433 loss) | |
| I0623 14:20:57.514485 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.994337 (* 0.0272727 = 0.0271183 loss) | |
| I0623 14:20:57.514499 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00238301 (* 0.0272727 = 6.49911e-05 loss) | |
| I0623 14:20:57.514513 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000596238 (* 0.0272727 = 1.6261e-05 loss) | |
| I0623 14:20:57.514528 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000135191 (* 0.0272727 = 3.68702e-06 loss) | |
| I0623 14:20:57.514541 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 2.43196e-05 (* 0.0272727 = 6.63262e-07 loss) | |
| I0623 14:20:57.514555 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 3.87432e-06 (* 0.0272727 = 1.05663e-07 loss) | |
| I0623 14:20:57.514567 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.738739 | |
| I0623 14:20:57.514578 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 14:20:57.514590 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 14:20:57.514600 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 14:20:57.514612 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 14:20:57.514623 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 14:20:57.514634 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 14:20:57.514645 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 14:20:57.514657 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.625 | |
| I0623 14:20:57.514667 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 14:20:57.514679 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.375 | |
| I0623 14:20:57.514690 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.375 | |
| I0623 14:20:57.514701 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.25 | |
| I0623 14:20:57.514713 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.375 | |
| I0623 14:20:57.514724 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.5 | |
| I0623 14:20:57.514734 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 14:20:57.514745 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 14:20:57.514766 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 14:20:57.514780 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 14:20:57.514791 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 14:20:57.514802 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 14:20:57.514813 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 14:20:57.514824 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 14:20:57.514835 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.835227 | |
| I0623 14:20:57.514847 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.936937 | |
| I0623 14:20:57.514860 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.836642 (* 1 = 0.836642 loss) | |
| I0623 14:20:57.514873 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.533756 (* 1 = 0.533756 loss) | |
| I0623 14:20:57.514888 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0850979 (* 0.0909091 = 0.00773617 loss) | |
| I0623 14:20:57.514901 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0624322 (* 0.0909091 = 0.00567565 loss) | |
| I0623 14:20:57.514915 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0335406 (* 0.0909091 = 0.00304914 loss) | |
| I0623 14:20:57.514930 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0513493 (* 0.0909091 = 0.00466812 loss) | |
| I0623 14:20:57.514943 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.493017 (* 0.0909091 = 0.0448197 loss) | |
| I0623 14:20:57.514957 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.656601 (* 0.0909091 = 0.059691 loss) | |
| I0623 14:20:57.514971 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.814834 (* 0.0909091 = 0.0740758 loss) | |
| I0623 14:20:57.514984 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.56302 (* 0.0909091 = 0.142093 loss) | |
| I0623 14:20:57.514998 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.18253 (* 0.0909091 = 0.107503 loss) | |
| I0623 14:20:57.515012 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.23405 (* 0.0909091 = 0.112186 loss) | |
| I0623 14:20:57.515025 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.48682 (* 0.0909091 = 0.135165 loss) | |
| I0623 14:20:57.515038 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.47172 (* 0.0909091 = 0.133793 loss) | |
| I0623 14:20:57.515053 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.33224 (* 0.0909091 = 0.121113 loss) | |
| I0623 14:20:57.515065 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.41434 (* 0.0909091 = 0.128576 loss) | |
| I0623 14:20:57.515079 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.05772 (* 0.0909091 = 0.0961561 loss) | |
| I0623 14:20:57.515092 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.798388 (* 0.0909091 = 0.0725808 loss) | |
| I0623 14:20:57.515106 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.981651 (* 0.0909091 = 0.089241 loss) | |
| I0623 14:20:57.515120 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000568332 (* 0.0909091 = 5.16665e-05 loss) | |
| I0623 14:20:57.515135 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000143639 (* 0.0909091 = 1.30581e-05 loss) | |
| I0623 14:20:57.515148 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 4.31788e-05 (* 0.0909091 = 3.92534e-06 loss) | |
| I0623 14:20:57.515161 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 1.25171e-05 (* 0.0909091 = 1.13792e-06 loss) | |
| I0623 14:20:57.515175 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 9.08984e-06 (* 0.0909091 = 8.26349e-07 loss) | |
| I0623 14:20:57.515187 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 14:20:57.515199 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 14:20:57.515210 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0310181 | |
| I0623 14:20:57.515231 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.00863898 | |
| I0623 14:20:57.515246 10365 sgd_solver.cpp:106] Iteration 500, lr = 0.001 | |
| I0623 14:21:35.439049 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.2764 > 30) by scale factor 0.990871 | |
| I0623 14:23:42.637328 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.8471 > 30) by scale factor 0.772259 | |
| I0623 14:24:21.707362 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 39.9716 > 30) by scale factor 0.750533 | |
| I0623 14:24:33.197569 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.2116 > 30) by scale factor 0.851992 | |
| I0623 14:25:29.132534 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.0346 > 30) by scale factor 0.908138 | |
| I0623 14:26:56.500510 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.7428 > 30) by scale factor 0.774337 | |
| I0623 14:27:20.673233 10365 solver.cpp:229] Iteration 1000, loss = 5.0137 | |
| I0623 14:27:20.673290 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.4 | |
| I0623 14:27:20.673308 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.5 | |
| I0623 14:27:20.673321 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 14:27:20.673333 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 14:27:20.673346 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 14:27:20.673357 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0 | |
| I0623 14:27:20.673369 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 14:27:20.673382 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.125 | |
| I0623 14:27:20.673393 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 14:27:20.673405 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 14:27:20.673416 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.125 | |
| I0623 14:27:20.673429 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 14:27:20.673439 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 14:27:20.673451 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.375 | |
| I0623 14:27:20.673462 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 14:27:20.673475 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 1 | |
| I0623 14:27:20.673485 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 1 | |
| I0623 14:27:20.673497 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 14:27:20.673508 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 14:27:20.673519 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 14:27:20.673532 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 14:27:20.673542 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 14:27:20.673553 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 14:27:20.673564 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.647727 | |
| I0623 14:27:20.673576 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.73 | |
| I0623 14:27:20.673593 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.95348 (* 0.3 = 0.586043 loss) | |
| I0623 14:27:20.673606 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.15325 (* 0.3 = 0.345975 loss) | |
| I0623 14:27:20.673620 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 1.00459 (* 0.0272727 = 0.027398 loss) | |
| I0623 14:27:20.673635 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.73638 (* 0.0272727 = 0.0473558 loss) | |
| I0623 14:27:20.673648 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.89167 (* 0.0272727 = 0.0515909 loss) | |
| I0623 14:27:20.673665 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.0593 (* 0.0272727 = 0.0561628 loss) | |
| I0623 14:27:20.673678 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 3.29933 (* 0.0272727 = 0.0899817 loss) | |
| I0623 14:27:20.673692 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.37125 (* 0.0272727 = 0.0646706 loss) | |
| I0623 14:27:20.673705 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.0382 (* 0.0272727 = 0.0555871 loss) | |
| I0623 14:27:20.673719 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.20754 (* 0.0272727 = 0.0602055 loss) | |
| I0623 14:27:20.673732 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.05245 (* 0.0272727 = 0.055976 loss) | |
| I0623 14:27:20.673746 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.85184 (* 0.0272727 = 0.0505047 loss) | |
| I0623 14:27:20.673760 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 3.06303 (* 0.0272727 = 0.0835371 loss) | |
| I0623 14:27:20.673774 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.68697 (* 0.0272727 = 0.0732809 loss) | |
| I0623 14:27:20.673816 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.11863 (* 0.0272727 = 0.0577807 loss) | |
| I0623 14:27:20.673831 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.830733 (* 0.0272727 = 0.0226564 loss) | |
| I0623 14:27:20.673846 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.169665 (* 0.0272727 = 0.00462723 loss) | |
| I0623 14:27:20.673859 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.0521971 (* 0.0272727 = 0.00142356 loss) | |
| I0623 14:27:20.673873 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00874121 (* 0.0272727 = 0.000238397 loss) | |
| I0623 14:27:20.673887 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00658462 (* 0.0272727 = 0.000179581 loss) | |
| I0623 14:27:20.673902 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.000559141 (* 0.0272727 = 1.52493e-05 loss) | |
| I0623 14:27:20.673915 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000220096 (* 0.0272727 = 6.00262e-06 loss) | |
| I0623 14:27:20.673929 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000131226 (* 0.0272727 = 3.5789e-06 loss) | |
| I0623 14:27:20.673943 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 8.71745e-06 (* 0.0272727 = 2.37749e-07 loss) | |
| I0623 14:27:20.673956 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.54 | |
| I0623 14:27:20.673967 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 14:27:20.673979 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 14:27:20.673990 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 14:27:20.674001 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 14:27:20.674013 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 14:27:20.674024 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 14:27:20.674036 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.25 | |
| I0623 14:27:20.674047 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 14:27:20.674059 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.25 | |
| I0623 14:27:20.674070 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.25 | |
| I0623 14:27:20.674082 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 14:27:20.674093 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.375 | |
| I0623 14:27:20.674105 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 14:27:20.674119 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 14:27:20.674131 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 1 | |
| I0623 14:27:20.674142 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 14:27:20.674154 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 14:27:20.674165 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 14:27:20.674175 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 14:27:20.674187 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 14:27:20.674199 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 14:27:20.674211 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 14:27:20.674221 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.738636 | |
| I0623 14:27:20.674233 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.8 | |
| I0623 14:27:20.674247 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.35541 (* 0.3 = 0.406622 loss) | |
| I0623 14:27:20.674262 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.79964 (* 0.3 = 0.239892 loss) | |
| I0623 14:27:20.674278 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.185283 (* 0.0272727 = 0.00505318 loss) | |
| I0623 14:27:20.674293 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.297175 (* 0.0272727 = 0.00810478 loss) | |
| I0623 14:27:20.674319 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.355719 (* 0.0272727 = 0.00970144 loss) | |
| I0623 14:27:20.674334 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.00669 (* 0.0272727 = 0.0274553 loss) | |
| I0623 14:27:20.674347 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.44598 (* 0.0272727 = 0.0667086 loss) | |
| I0623 14:27:20.674361 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.98102 (* 0.0272727 = 0.0540278 loss) | |
| I0623 14:27:20.674374 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.8353 (* 0.0272727 = 0.0500536 loss) | |
| I0623 14:27:20.674388 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.13399 (* 0.0272727 = 0.0581998 loss) | |
| I0623 14:27:20.674401 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 2.10105 (* 0.0272727 = 0.0573014 loss) | |
| I0623 14:27:20.674414 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.9969 (* 0.0272727 = 0.0544608 loss) | |
| I0623 14:27:20.674428 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.48258 (* 0.0272727 = 0.0677067 loss) | |
| I0623 14:27:20.674441 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.60997 (* 0.0272727 = 0.0439084 loss) | |
| I0623 14:27:20.674454 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.73087 (* 0.0272727 = 0.0472056 loss) | |
| I0623 14:27:20.674468 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.571708 (* 0.0272727 = 0.015592 loss) | |
| I0623 14:27:20.674482 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.0720522 (* 0.0272727 = 0.00196506 loss) | |
| I0623 14:27:20.674497 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.0269818 (* 0.0272727 = 0.000735868 loss) | |
| I0623 14:27:20.674507 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.00143195 (* 0.0272727 = 3.90533e-05 loss) | |
| I0623 14:27:20.674522 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.000274465 (* 0.0272727 = 7.48541e-06 loss) | |
| I0623 14:27:20.674536 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000105223 (* 0.0272727 = 2.86972e-06 loss) | |
| I0623 14:27:20.674551 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 3.79486e-05 (* 0.0272727 = 1.03496e-06 loss) | |
| I0623 14:27:20.674564 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 9.47731e-06 (* 0.0272727 = 2.58472e-07 loss) | |
| I0623 14:27:20.674578 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.77325e-06 (* 0.0272727 = 4.83613e-08 loss) | |
| I0623 14:27:20.674590 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.89 | |
| I0623 14:27:20.674602 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 14:27:20.674614 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 14:27:20.674625 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 14:27:20.674636 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 14:27:20.674648 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 14:27:20.674659 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 14:27:20.674670 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 14:27:20.674681 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.625 | |
| I0623 14:27:20.674693 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.25 | |
| I0623 14:27:20.674705 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 14:27:20.674715 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 14:27:20.674726 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.75 | |
| I0623 14:27:20.674737 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 14:27:20.674749 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 14:27:20.674760 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 14:27:20.674782 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 14:27:20.674793 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 14:27:20.674804 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 14:27:20.674816 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 14:27:20.674828 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 14:27:20.674839 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 14:27:20.674849 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 14:27:20.674860 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.931818 | |
| I0623 14:27:20.674872 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.98 | |
| I0623 14:27:20.674885 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.437546 (* 1 = 0.437546 loss) | |
| I0623 14:27:20.674899 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.271503 (* 1 = 0.271503 loss) | |
| I0623 14:27:20.674912 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.200327 (* 0.0909091 = 0.0182115 loss) | |
| I0623 14:27:20.674926 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0666142 (* 0.0909091 = 0.00605583 loss) | |
| I0623 14:27:20.674940 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0982584 (* 0.0909091 = 0.00893258 loss) | |
| I0623 14:27:20.674953 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.258239 (* 0.0909091 = 0.0234763 loss) | |
| I0623 14:27:20.674968 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.699413 (* 0.0909091 = 0.063583 loss) | |
| I0623 14:27:20.674981 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.268446 (* 0.0909091 = 0.0244042 loss) | |
| I0623 14:27:20.674994 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.289023 (* 0.0909091 = 0.0262749 loss) | |
| I0623 14:27:20.675009 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.93049 (* 0.0909091 = 0.08459 loss) | |
| I0623 14:27:20.675022 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.44596 (* 0.0909091 = 0.131451 loss) | |
| I0623 14:27:20.675036 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.840798 (* 0.0909091 = 0.0764362 loss) | |
| I0623 14:27:20.675050 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.876739 (* 0.0909091 = 0.0797035 loss) | |
| I0623 14:27:20.675063 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.877419 (* 0.0909091 = 0.0797654 loss) | |
| I0623 14:27:20.675077 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.940941 (* 0.0909091 = 0.0855401 loss) | |
| I0623 14:27:20.675091 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.589183 (* 0.0909091 = 0.0535621 loss) | |
| I0623 14:27:20.675104 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.14662 (* 0.0909091 = 0.0133291 loss) | |
| I0623 14:27:20.675117 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.0129137 (* 0.0909091 = 0.00117397 loss) | |
| I0623 14:27:20.675132 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.00143373 (* 0.0909091 = 0.000130339 loss) | |
| I0623 14:27:20.675145 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 8.6886e-05 (* 0.0909091 = 7.89873e-06 loss) | |
| I0623 14:27:20.675159 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 4.02736e-05 (* 0.0909091 = 3.66124e-06 loss) | |
| I0623 14:27:20.675176 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 2.55423e-05 (* 0.0909091 = 2.32203e-06 loss) | |
| I0623 14:27:20.675190 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 1.35156e-05 (* 0.0909091 = 1.2287e-06 loss) | |
| I0623 14:27:20.675204 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 1.53482e-06 (* 0.0909091 = 1.39529e-07 loss) | |
| I0623 14:27:20.675216 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 14:27:20.675228 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 14:27:20.675249 10365 solver.cpp:245] Train net output #149: total_confidence = 0.046828 | |
| I0623 14:27:20.675263 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0284861 | |
| I0623 14:27:20.675276 10365 sgd_solver.cpp:106] Iteration 1000, lr = 0.001 | |
| I0623 14:29:23.664837 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.874 > 30) by scale factor 0.771725 | |
| I0623 14:30:48.724439 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.3493 > 30) by scale factor 0.825326 | |
| I0623 14:31:12.488049 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 49.8042 > 30) by scale factor 0.602359 | |
| I0623 14:33:43.876998 10365 solver.cpp:229] Iteration 1500, loss = 4.99068 | |
| I0623 14:33:43.877152 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.436893 | |
| I0623 14:33:43.877172 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 14:33:43.877185 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 14:33:43.877198 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 14:33:43.877209 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.25 | |
| I0623 14:33:43.877221 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.125 | |
| I0623 14:33:43.877233 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 14:33:43.877245 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 14:33:43.877257 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 14:33:43.877280 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 14:33:43.877298 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 14:33:43.877311 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 14:33:43.877322 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 14:33:43.877334 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 14:33:43.877346 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 14:33:43.877358 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.5 | |
| I0623 14:33:43.877370 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.5 | |
| I0623 14:33:43.877382 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 14:33:43.877394 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 14:33:43.877405 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 14:33:43.877418 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 14:33:43.877429 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 14:33:43.877441 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 14:33:43.877452 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.659091 | |
| I0623 14:33:43.877465 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.76699 | |
| I0623 14:33:43.877482 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.63592 (* 0.3 = 0.490776 loss) | |
| I0623 14:33:43.877497 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.99772 (* 0.3 = 0.299316 loss) | |
| I0623 14:33:43.877512 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.35398 (* 0.0272727 = 0.00965401 loss) | |
| I0623 14:33:43.877526 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.51821 (* 0.0272727 = 0.0414058 loss) | |
| I0623 14:33:43.877540 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.72349 (* 0.0272727 = 0.0470044 loss) | |
| I0623 14:33:43.877554 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.80135 (* 0.0272727 = 0.0491277 loss) | |
| I0623 14:33:43.877569 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.1189 (* 0.0272727 = 0.0577882 loss) | |
| I0623 14:33:43.877583 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.21734 (* 0.0272727 = 0.0604728 loss) | |
| I0623 14:33:43.877596 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.60432 (* 0.0272727 = 0.0437542 loss) | |
| I0623 14:33:43.877610 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.35414 (* 0.0272727 = 0.0369312 loss) | |
| I0623 14:33:43.877624 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.20818 (* 0.0272727 = 0.0602232 loss) | |
| I0623 14:33:43.877638 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.20143 (* 0.0272727 = 0.060039 loss) | |
| I0623 14:33:43.877651 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.63611 (* 0.0272727 = 0.0446211 loss) | |
| I0623 14:33:43.877665 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.94855 (* 0.0272727 = 0.0531422 loss) | |
| I0623 14:33:43.877701 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.67807 (* 0.0272727 = 0.0457655 loss) | |
| I0623 14:33:43.877715 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.71274 (* 0.0272727 = 0.0467112 loss) | |
| I0623 14:33:43.877728 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.37181 (* 0.0272727 = 0.0374129 loss) | |
| I0623 14:33:43.877743 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.20252 (* 0.0272727 = 0.032796 loss) | |
| I0623 14:33:43.877756 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.48925 (* 0.0272727 = 0.0133432 loss) | |
| I0623 14:33:43.877770 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.448091 (* 0.0272727 = 0.0122207 loss) | |
| I0623 14:33:43.877784 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00811442 (* 0.0272727 = 0.000221302 loss) | |
| I0623 14:33:43.877799 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00415387 (* 0.0272727 = 0.000113287 loss) | |
| I0623 14:33:43.877813 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000825719 (* 0.0272727 = 2.25196e-05 loss) | |
| I0623 14:33:43.877828 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 5.03877e-05 (* 0.0272727 = 1.37421e-06 loss) | |
| I0623 14:33:43.877840 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.524272 | |
| I0623 14:33:43.877853 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 14:33:43.877866 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 14:33:43.877876 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 14:33:43.877888 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.875 | |
| I0623 14:33:43.877900 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 14:33:43.877912 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 14:33:43.877921 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.625 | |
| I0623 14:33:43.877929 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.5 | |
| I0623 14:33:43.877943 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.75 | |
| I0623 14:33:43.877954 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 14:33:43.877966 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 14:33:43.877977 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.25 | |
| I0623 14:33:43.877990 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 14:33:43.878001 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.375 | |
| I0623 14:33:43.878012 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.5 | |
| I0623 14:33:43.878024 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 14:33:43.878036 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 14:33:43.878046 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 14:33:43.878057 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 14:33:43.878069 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 14:33:43.878080 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 14:33:43.878093 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 14:33:43.878103 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.704545 | |
| I0623 14:33:43.878115 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.825243 | |
| I0623 14:33:43.878129 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.24201 (* 0.3 = 0.372604 loss) | |
| I0623 14:33:43.878144 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.75484 (* 0.3 = 0.226452 loss) | |
| I0623 14:33:43.878157 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.326613 (* 0.0272727 = 0.00890763 loss) | |
| I0623 14:33:43.878188 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.11451 (* 0.0272727 = 0.003123 loss) | |
| I0623 14:33:43.878224 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.858004 (* 0.0272727 = 0.0234001 loss) | |
| I0623 14:33:43.878240 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.14072 (* 0.0272727 = 0.0311106 loss) | |
| I0623 14:33:43.878254 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.07329 (* 0.0272727 = 0.0292716 loss) | |
| I0623 14:33:43.878268 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.50703 (* 0.0272727 = 0.0411007 loss) | |
| I0623 14:33:43.878283 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.20357 (* 0.0272727 = 0.0328247 loss) | |
| I0623 14:33:43.878295 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.34352 (* 0.0272727 = 0.0366415 loss) | |
| I0623 14:33:43.878309 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.2565 (* 0.0272727 = 0.0342683 loss) | |
| I0623 14:33:43.878326 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.54333 (* 0.0272727 = 0.0420908 loss) | |
| I0623 14:33:43.878340 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.30628 (* 0.0272727 = 0.0356259 loss) | |
| I0623 14:33:43.878355 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.66233 (* 0.0272727 = 0.0453363 loss) | |
| I0623 14:33:43.878367 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.859647 (* 0.0272727 = 0.0234449 loss) | |
| I0623 14:33:43.878381 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.6104 (* 0.0272727 = 0.0439201 loss) | |
| I0623 14:33:43.878396 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.54237 (* 0.0272727 = 0.0420646 loss) | |
| I0623 14:33:43.878409 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.867605 (* 0.0272727 = 0.023662 loss) | |
| I0623 14:33:43.878422 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.428251 (* 0.0272727 = 0.0116796 loss) | |
| I0623 14:33:43.878437 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.360381 (* 0.0272727 = 0.00982856 loss) | |
| I0623 14:33:43.878450 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0439134 (* 0.0272727 = 0.00119764 loss) | |
| I0623 14:33:43.878464 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00346557 (* 0.0272727 = 9.45157e-05 loss) | |
| I0623 14:33:43.878479 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00228619 (* 0.0272727 = 6.23506e-05 loss) | |
| I0623 14:33:43.878494 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.00054124 (* 0.0272727 = 1.47611e-05 loss) | |
| I0623 14:33:43.878505 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.708738 | |
| I0623 14:33:43.878517 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 14:33:43.878530 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 14:33:43.878541 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 14:33:43.878553 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 14:33:43.878566 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 14:33:43.878576 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.75 | |
| I0623 14:33:43.878588 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 14:33:43.878600 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 14:33:43.878612 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 14:33:43.878624 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 14:33:43.878635 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.5 | |
| I0623 14:33:43.878648 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 14:33:43.878659 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 14:33:43.878670 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.375 | |
| I0623 14:33:43.878682 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.5 | |
| I0623 14:33:43.878695 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 14:33:43.878716 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.75 | |
| I0623 14:33:43.878729 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.875 | |
| I0623 14:33:43.878741 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 14:33:43.878753 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 14:33:43.878765 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 14:33:43.878777 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 14:33:43.878788 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.8125 | |
| I0623 14:33:43.878800 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.92233 | |
| I0623 14:33:43.878814 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.919121 (* 1 = 0.919121 loss) | |
| I0623 14:33:43.878829 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.580445 (* 1 = 0.580445 loss) | |
| I0623 14:33:43.878842 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0695891 (* 0.0909091 = 0.00632628 loss) | |
| I0623 14:33:43.878857 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0775979 (* 0.0909091 = 0.00705436 loss) | |
| I0623 14:33:43.878871 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.994121 (* 0.0909091 = 0.0903746 loss) | |
| I0623 14:33:43.878885 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.743891 (* 0.0909091 = 0.0676265 loss) | |
| I0623 14:33:43.878900 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.836761 (* 0.0909091 = 0.0760692 loss) | |
| I0623 14:33:43.878913 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.574879 (* 0.0909091 = 0.0522618 loss) | |
| I0623 14:33:43.878927 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.614564 (* 0.0909091 = 0.0558695 loss) | |
| I0623 14:33:43.878942 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.520687 (* 0.0909091 = 0.0473352 loss) | |
| I0623 14:33:43.878955 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.28832 (* 0.0909091 = 0.11712 loss) | |
| I0623 14:33:43.878969 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.1674 (* 0.0909091 = 0.106127 loss) | |
| I0623 14:33:43.878983 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.11603 (* 0.0909091 = 0.101457 loss) | |
| I0623 14:33:43.878996 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.10812 (* 0.0909091 = 0.100738 loss) | |
| I0623 14:33:43.879010 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.05021 (* 0.0909091 = 0.0954736 loss) | |
| I0623 14:33:43.879024 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.49 (* 0.0909091 = 0.135455 loss) | |
| I0623 14:33:43.879039 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.917307 (* 0.0909091 = 0.0833915 loss) | |
| I0623 14:33:43.879052 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.69531 (* 0.0909091 = 0.06321 loss) | |
| I0623 14:33:43.879066 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.621631 (* 0.0909091 = 0.0565119 loss) | |
| I0623 14:33:43.879081 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.485634 (* 0.0909091 = 0.0441486 loss) | |
| I0623 14:33:43.879094 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0202815 (* 0.0909091 = 0.00184377 loss) | |
| I0623 14:33:43.879108 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00393216 (* 0.0909091 = 0.000357469 loss) | |
| I0623 14:33:43.879123 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.00140446 (* 0.0909091 = 0.000127678 loss) | |
| I0623 14:33:43.879137 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 0.000310339 (* 0.0909091 = 2.82127e-05 loss) | |
| I0623 14:33:43.879150 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 14:33:43.879161 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.25 | |
| I0623 14:33:43.879173 10365 solver.cpp:245] Train net output #149: total_confidence = 0.224659 | |
| I0623 14:33:43.879195 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.22658 | |
| I0623 14:33:43.879212 10365 sgd_solver.cpp:106] Iteration 1500, lr = 0.001 | |
| I0623 14:34:59.331606 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.986 > 30) by scale factor 0.769506 | |
| I0623 14:40:07.028699 10365 solver.cpp:229] Iteration 2000, loss = 4.96318 | |
| I0623 14:40:07.028813 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.505155 | |
| I0623 14:40:07.028832 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 14:40:07.028846 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 14:40:07.028857 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.625 | |
| I0623 14:40:07.028869 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 14:40:07.028882 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.625 | |
| I0623 14:40:07.028893 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 14:40:07.028905 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 14:40:07.028918 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 14:40:07.028928 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.625 | |
| I0623 14:40:07.028940 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 14:40:07.028952 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 14:40:07.028964 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 14:40:07.028975 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 14:40:07.028987 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 14:40:07.028998 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 14:40:07.029011 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 14:40:07.029022 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 14:40:07.029033 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 14:40:07.029045 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 14:40:07.029058 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 14:40:07.029072 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 14:40:07.029083 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 14:40:07.029094 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.704545 | |
| I0623 14:40:07.029106 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.742268 | |
| I0623 14:40:07.029122 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.52434 (* 0.3 = 0.457301 loss) | |
| I0623 14:40:07.029137 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.901596 (* 0.3 = 0.270479 loss) | |
| I0623 14:40:07.029152 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.154487 (* 0.0272727 = 0.00421327 loss) | |
| I0623 14:40:07.029166 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.39093 (* 0.0272727 = 0.0379345 loss) | |
| I0623 14:40:07.029181 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.29701 (* 0.0272727 = 0.035373 loss) | |
| I0623 14:40:07.029194 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.01864 (* 0.0272727 = 0.0550539 loss) | |
| I0623 14:40:07.029208 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.55337 (* 0.0272727 = 0.0423645 loss) | |
| I0623 14:40:07.029222 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.84741 (* 0.0272727 = 0.0503838 loss) | |
| I0623 14:40:07.029237 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.24955 (* 0.0272727 = 0.0613515 loss) | |
| I0623 14:40:07.029250 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.01946 (* 0.0272727 = 0.0550762 loss) | |
| I0623 14:40:07.029263 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.25868 (* 0.0272727 = 0.0343276 loss) | |
| I0623 14:40:07.029278 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.35633 (* 0.0272727 = 0.0369908 loss) | |
| I0623 14:40:07.029291 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.84703 (* 0.0272727 = 0.0503736 loss) | |
| I0623 14:40:07.029304 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.38104 (* 0.0272727 = 0.0376647 loss) | |
| I0623 14:40:07.029335 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.21026 (* 0.0272727 = 0.0330072 loss) | |
| I0623 14:40:07.029350 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.564738 (* 0.0272727 = 0.0154019 loss) | |
| I0623 14:40:07.029364 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.0343 (* 0.0272727 = 0.0282082 loss) | |
| I0623 14:40:07.029377 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.05757 (* 0.0272727 = 0.0288428 loss) | |
| I0623 14:40:07.029392 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00716972 (* 0.0272727 = 0.000195538 loss) | |
| I0623 14:40:07.029407 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.000416087 (* 0.0272727 = 1.13478e-05 loss) | |
| I0623 14:40:07.029420 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 6.06702e-05 (* 0.0272727 = 1.65464e-06 loss) | |
| I0623 14:40:07.029434 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 1.7122e-05 (* 0.0272727 = 4.66964e-07 loss) | |
| I0623 14:40:07.029448 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 2.65242e-06 (* 0.0272727 = 7.23387e-08 loss) | |
| I0623 14:40:07.029463 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 6.36291e-06 (* 0.0272727 = 1.73534e-07 loss) | |
| I0623 14:40:07.029474 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.536082 | |
| I0623 14:40:07.029486 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 14:40:07.029498 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.5 | |
| I0623 14:40:07.029510 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.625 | |
| I0623 14:40:07.029520 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 14:40:07.029532 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 14:40:07.029543 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 14:40:07.029556 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 14:40:07.029567 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 14:40:07.029577 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 14:40:07.029589 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.625 | |
| I0623 14:40:07.029600 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 14:40:07.029611 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.75 | |
| I0623 14:40:07.029623 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 14:40:07.029634 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 14:40:07.029645 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 14:40:07.029657 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 14:40:07.029669 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 14:40:07.029680 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 14:40:07.029690 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 14:40:07.029705 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 14:40:07.029716 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 14:40:07.029727 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 14:40:07.029739 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.727273 | |
| I0623 14:40:07.029752 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.845361 | |
| I0623 14:40:07.029765 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.27407 (* 0.3 = 0.382221 loss) | |
| I0623 14:40:07.029778 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.749142 (* 0.3 = 0.224743 loss) | |
| I0623 14:40:07.029793 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.289582 (* 0.0272727 = 0.00789769 loss) | |
| I0623 14:40:07.029808 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 1.12146 (* 0.0272727 = 0.0305852 loss) | |
| I0623 14:40:07.029832 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.23205 (* 0.0272727 = 0.0336015 loss) | |
| I0623 14:40:07.029847 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.98002 (* 0.0272727 = 0.0267278 loss) | |
| I0623 14:40:07.029860 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.08595 (* 0.0272727 = 0.0296167 loss) | |
| I0623 14:40:07.029875 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.41348 (* 0.0272727 = 0.0658221 loss) | |
| I0623 14:40:07.029888 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.34985 (* 0.0272727 = 0.036814 loss) | |
| I0623 14:40:07.029901 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.67264 (* 0.0272727 = 0.0456173 loss) | |
| I0623 14:40:07.029916 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.36723 (* 0.0272727 = 0.0372882 loss) | |
| I0623 14:40:07.029928 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.29962 (* 0.0272727 = 0.0354442 loss) | |
| I0623 14:40:07.029942 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.4849 (* 0.0272727 = 0.0404972 loss) | |
| I0623 14:40:07.029956 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.00619 (* 0.0272727 = 0.0274414 loss) | |
| I0623 14:40:07.029970 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.18223 (* 0.0272727 = 0.0322426 loss) | |
| I0623 14:40:07.029980 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.75579 (* 0.0272727 = 0.0206124 loss) | |
| I0623 14:40:07.029995 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.12627 (* 0.0272727 = 0.0307163 loss) | |
| I0623 14:40:07.030009 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.731459 (* 0.0272727 = 0.0199489 loss) | |
| I0623 14:40:07.030024 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.00600807 (* 0.0272727 = 0.000163856 loss) | |
| I0623 14:40:07.030038 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.000911532 (* 0.0272727 = 2.486e-05 loss) | |
| I0623 14:40:07.030052 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000379858 (* 0.0272727 = 1.03598e-05 loss) | |
| I0623 14:40:07.030066 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000172318 (* 0.0272727 = 4.69958e-06 loss) | |
| I0623 14:40:07.030081 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 8.93462e-05 (* 0.0272727 = 2.43672e-06 loss) | |
| I0623 14:40:07.030094 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 3.42399e-05 (* 0.0272727 = 9.33816e-07 loss) | |
| I0623 14:40:07.030105 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.783505 | |
| I0623 14:40:07.030120 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 14:40:07.030133 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 14:40:07.030144 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 14:40:07.030155 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 14:40:07.030166 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 14:40:07.030177 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.75 | |
| I0623 14:40:07.030189 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 14:40:07.030200 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 14:40:07.030211 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 14:40:07.030223 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 14:40:07.030235 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 14:40:07.030246 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.75 | |
| I0623 14:40:07.030256 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 14:40:07.030268 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 14:40:07.030279 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 14:40:07.030290 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 14:40:07.030311 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 14:40:07.030324 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 14:40:07.030335 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 14:40:07.030347 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 14:40:07.030359 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 14:40:07.030369 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 14:40:07.030380 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.875 | |
| I0623 14:40:07.030392 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.969072 | |
| I0623 14:40:07.030405 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.617477 (* 1 = 0.617477 loss) | |
| I0623 14:40:07.030419 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.355193 (* 1 = 0.355193 loss) | |
| I0623 14:40:07.030433 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.048981 (* 0.0909091 = 0.00445282 loss) | |
| I0623 14:40:07.030447 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.160332 (* 0.0909091 = 0.0145757 loss) | |
| I0623 14:40:07.030462 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.648969 (* 0.0909091 = 0.0589972 loss) | |
| I0623 14:40:07.030475 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.326771 (* 0.0909091 = 0.0297065 loss) | |
| I0623 14:40:07.030488 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.106711 (* 0.0909091 = 0.00970104 loss) | |
| I0623 14:40:07.030503 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.559422 (* 0.0909091 = 0.0508566 loss) | |
| I0623 14:40:07.030516 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.594721 (* 0.0909091 = 0.0540655 loss) | |
| I0623 14:40:07.030529 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.08508 (* 0.0909091 = 0.098644 loss) | |
| I0623 14:40:07.030544 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.16768 (* 0.0909091 = 0.106153 loss) | |
| I0623 14:40:07.030557 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.815407 (* 0.0909091 = 0.0741279 loss) | |
| I0623 14:40:07.030570 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.818788 (* 0.0909091 = 0.0744353 loss) | |
| I0623 14:40:07.030585 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.80094 (* 0.0909091 = 0.0728128 loss) | |
| I0623 14:40:07.030597 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.702165 (* 0.0909091 = 0.0638332 loss) | |
| I0623 14:40:07.030611 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.563488 (* 0.0909091 = 0.0512262 loss) | |
| I0623 14:40:07.030625 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.59477 (* 0.0909091 = 0.05407 loss) | |
| I0623 14:40:07.030638 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.294131 (* 0.0909091 = 0.0267392 loss) | |
| I0623 14:40:07.030652 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.020137 (* 0.0909091 = 0.00183063 loss) | |
| I0623 14:40:07.030666 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00129659 (* 0.0909091 = 0.000117872 loss) | |
| I0623 14:40:07.030680 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000211465 (* 0.0909091 = 1.92241e-05 loss) | |
| I0623 14:40:07.030694 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 8.47347e-05 (* 0.0909091 = 7.70316e-06 loss) | |
| I0623 14:40:07.030709 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 9.73994e-05 (* 0.0909091 = 8.85449e-06 loss) | |
| I0623 14:40:07.030722 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 1.54678e-05 (* 0.0909091 = 1.40617e-06 loss) | |
| I0623 14:40:07.030735 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 14:40:07.030746 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 14:40:07.030771 10365 solver.cpp:245] Train net output #149: total_confidence = 0.142096 | |
| I0623 14:40:07.030786 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.122268 | |
| I0623 14:40:07.030798 10365 sgd_solver.cpp:106] Iteration 2000, lr = 0.001 | |
| I0623 14:41:37.035840 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 45.1753 > 30) by scale factor 0.66408 | |
| I0623 14:41:43.169929 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.2328 > 30) by scale factor 0.930729 | |
| I0623 14:46:30.048812 10365 solver.cpp:229] Iteration 2500, loss = 4.94346 | |
| I0623 14:46:30.048944 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.458716 | |
| I0623 14:46:30.048976 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 14:46:30.049000 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 14:46:30.049021 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 14:46:30.049043 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 14:46:30.049067 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 14:46:30.049090 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.125 | |
| I0623 14:46:30.049113 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 14:46:30.049134 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.125 | |
| I0623 14:46:30.049154 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 14:46:30.049175 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 14:46:30.049196 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.125 | |
| I0623 14:46:30.049218 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 14:46:30.049239 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.125 | |
| I0623 14:46:30.049263 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.375 | |
| I0623 14:46:30.049286 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 1 | |
| I0623 14:46:30.049307 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 14:46:30.049329 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 14:46:30.049350 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 14:46:30.049371 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 14:46:30.049393 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 14:46:30.049417 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 14:46:30.049438 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 14:46:30.049459 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.659091 | |
| I0623 14:46:30.049481 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.752294 | |
| I0623 14:46:30.049510 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.65286 (* 0.3 = 0.495859 loss) | |
| I0623 14:46:30.049535 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.0534 (* 0.3 = 0.316021 loss) | |
| I0623 14:46:30.049563 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.332442 (* 0.0272727 = 0.00906661 loss) | |
| I0623 14:46:30.049590 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.74377 (* 0.0272727 = 0.0202846 loss) | |
| I0623 14:46:30.049617 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.78767 (* 0.0272727 = 0.0487547 loss) | |
| I0623 14:46:30.049641 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.83659 (* 0.0272727 = 0.0500888 loss) | |
| I0623 14:46:30.049667 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.10959 (* 0.0272727 = 0.0575343 loss) | |
| I0623 14:46:30.049693 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.24122 (* 0.0272727 = 0.0611243 loss) | |
| I0623 14:46:30.049720 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.02527 (* 0.0272727 = 0.0552346 loss) | |
| I0623 14:46:30.049746 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.68539 (* 0.0272727 = 0.0732379 loss) | |
| I0623 14:46:30.049770 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.55994 (* 0.0272727 = 0.0698164 loss) | |
| I0623 14:46:30.049796 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.7932 (* 0.0272727 = 0.0761783 loss) | |
| I0623 14:46:30.049821 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.7438 (* 0.0272727 = 0.0748308 loss) | |
| I0623 14:46:30.049846 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.09966 (* 0.0272727 = 0.0572634 loss) | |
| I0623 14:46:30.049896 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.25319 (* 0.0272727 = 0.0614506 loss) | |
| I0623 14:46:30.049923 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.79734 (* 0.0272727 = 0.0490183 loss) | |
| I0623 14:46:30.049949 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.307537 (* 0.0272727 = 0.00838738 loss) | |
| I0623 14:46:30.049981 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.238412 (* 0.0272727 = 0.00650213 loss) | |
| I0623 14:46:30.050009 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00769272 (* 0.0272727 = 0.000209801 loss) | |
| I0623 14:46:30.050036 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00262171 (* 0.0272727 = 7.15013e-05 loss) | |
| I0623 14:46:30.050062 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00128659 (* 0.0272727 = 3.50889e-05 loss) | |
| I0623 14:46:30.050089 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00174246 (* 0.0272727 = 4.75217e-05 loss) | |
| I0623 14:46:30.050114 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000290219 (* 0.0272727 = 7.91506e-06 loss) | |
| I0623 14:46:30.050140 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 1.63474e-05 (* 0.0272727 = 4.45837e-07 loss) | |
| I0623 14:46:30.050163 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.541284 | |
| I0623 14:46:30.050184 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 14:46:30.050206 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 14:46:30.050227 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 14:46:30.050247 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.375 | |
| I0623 14:46:30.050269 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 14:46:30.050289 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 14:46:30.050313 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 14:46:30.050336 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.125 | |
| I0623 14:46:30.050356 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 14:46:30.050379 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.125 | |
| I0623 14:46:30.050400 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 14:46:30.050422 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.125 | |
| I0623 14:46:30.050447 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 14:46:30.050472 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.5 | |
| I0623 14:46:30.050493 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 14:46:30.050514 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 14:46:30.050535 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 14:46:30.050555 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 14:46:30.050576 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 14:46:30.050597 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 14:46:30.050618 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 14:46:30.050639 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 14:46:30.050660 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.704545 | |
| I0623 14:46:30.050681 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.853211 | |
| I0623 14:46:30.050706 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.36306 (* 0.3 = 0.408919 loss) | |
| I0623 14:46:30.050731 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.882276 (* 0.3 = 0.264683 loss) | |
| I0623 14:46:30.050757 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.129856 (* 0.0272727 = 0.00354153 loss) | |
| I0623 14:46:30.050782 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.314755 (* 0.0272727 = 0.00858423 loss) | |
| I0623 14:46:30.050824 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.61505 (* 0.0272727 = 0.0440467 loss) | |
| I0623 14:46:30.050851 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.34749 (* 0.0272727 = 0.0367496 loss) | |
| I0623 14:46:30.050876 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.35053 (* 0.0272727 = 0.0368326 loss) | |
| I0623 14:46:30.050901 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.63173 (* 0.0272727 = 0.0445017 loss) | |
| I0623 14:46:30.050925 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.469 (* 0.0272727 = 0.0400635 loss) | |
| I0623 14:46:30.050950 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.39347 (* 0.0272727 = 0.0652764 loss) | |
| I0623 14:46:30.050976 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.80129 (* 0.0272727 = 0.049126 loss) | |
| I0623 14:46:30.050999 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.54239 (* 0.0272727 = 0.0693378 loss) | |
| I0623 14:46:30.051030 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.06496 (* 0.0272727 = 0.056317 loss) | |
| I0623 14:46:30.051057 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 2.47508 (* 0.0272727 = 0.0675022 loss) | |
| I0623 14:46:30.051082 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.89046 (* 0.0272727 = 0.0515581 loss) | |
| I0623 14:46:30.051107 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.69731 (* 0.0272727 = 0.0462902 loss) | |
| I0623 14:46:30.051133 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.300582 (* 0.0272727 = 0.00819768 loss) | |
| I0623 14:46:30.051157 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.204932 (* 0.0272727 = 0.00558905 loss) | |
| I0623 14:46:30.051183 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.00578984 (* 0.0272727 = 0.000157905 loss) | |
| I0623 14:46:30.051208 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00145277 (* 0.0272727 = 3.96211e-05 loss) | |
| I0623 14:46:30.051234 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000764299 (* 0.0272727 = 2.08445e-05 loss) | |
| I0623 14:46:30.051257 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000429066 (* 0.0272727 = 1.17018e-05 loss) | |
| I0623 14:46:30.051283 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000126682 (* 0.0272727 = 3.45495e-06 loss) | |
| I0623 14:46:30.051308 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 7.36114e-05 (* 0.0272727 = 2.00758e-06 loss) | |
| I0623 14:46:30.051331 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.834862 | |
| I0623 14:46:30.051352 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 14:46:30.051376 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 14:46:30.051396 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 14:46:30.051417 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 14:46:30.051439 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 14:46:30.051458 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 14:46:30.051478 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 14:46:30.051499 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 14:46:30.051520 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 14:46:30.051542 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 14:46:30.051561 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.25 | |
| I0623 14:46:30.051583 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 14:46:30.051625 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.25 | |
| I0623 14:46:30.051650 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 14:46:30.051671 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 1 | |
| I0623 14:46:30.051709 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 14:46:30.051731 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 14:46:30.051753 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 14:46:30.051774 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 14:46:30.051794 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 14:46:30.051813 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 14:46:30.051833 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 14:46:30.051854 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.897727 | |
| I0623 14:46:30.051875 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.990826 | |
| I0623 14:46:30.051899 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.494439 (* 1 = 0.494439 loss) | |
| I0623 14:46:30.051925 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.310383 (* 1 = 0.310383 loss) | |
| I0623 14:46:30.051950 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0250574 (* 0.0909091 = 0.00227794 loss) | |
| I0623 14:46:30.051976 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0381695 (* 0.0909091 = 0.00346995 loss) | |
| I0623 14:46:30.052002 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.731704 (* 0.0909091 = 0.0665186 loss) | |
| I0623 14:46:30.052027 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.136593 (* 0.0909091 = 0.0124175 loss) | |
| I0623 14:46:30.052052 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.208606 (* 0.0909091 = 0.0189641 loss) | |
| I0623 14:46:30.052083 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.436378 (* 0.0909091 = 0.0396707 loss) | |
| I0623 14:46:30.052109 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.271636 (* 0.0909091 = 0.0246942 loss) | |
| I0623 14:46:30.052135 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.610426 (* 0.0909091 = 0.0554933 loss) | |
| I0623 14:46:30.052160 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.0013 (* 0.0909091 = 0.0910276 loss) | |
| I0623 14:46:30.052186 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.632721 (* 0.0909091 = 0.0575201 loss) | |
| I0623 14:46:30.052211 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.62459 (* 0.0909091 = 0.14769 loss) | |
| I0623 14:46:30.052237 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.37371 (* 0.0909091 = 0.124883 loss) | |
| I0623 14:46:30.052261 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 2.02801 (* 0.0909091 = 0.184365 loss) | |
| I0623 14:46:30.052286 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.04196 (* 0.0909091 = 0.094724 loss) | |
| I0623 14:46:30.052311 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.143899 (* 0.0909091 = 0.0130817 loss) | |
| I0623 14:46:30.052336 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.250614 (* 0.0909091 = 0.0227831 loss) | |
| I0623 14:46:30.052361 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0108264 (* 0.0909091 = 0.000984214 loss) | |
| I0623 14:46:30.052387 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000755582 (* 0.0909091 = 6.86893e-05 loss) | |
| I0623 14:46:30.052412 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000127671 (* 0.0909091 = 1.16065e-05 loss) | |
| I0623 14:46:30.052443 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 3.37465e-05 (* 0.0909091 = 3.06786e-06 loss) | |
| I0623 14:46:30.052469 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 1.93276e-05 (* 0.0909091 = 1.75706e-06 loss) | |
| I0623 14:46:30.052495 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.38098e-06 (* 0.0909091 = 3.98271e-07 loss) | |
| I0623 14:46:30.052516 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 14:46:30.052537 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 14:46:30.052575 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0256491 | |
| I0623 14:46:30.052599 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.00997235 | |
| I0623 14:46:30.052621 10365 sgd_solver.cpp:106] Iteration 2500, lr = 0.001 | |
| I0623 14:46:48.023708 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 53.4176 > 30) by scale factor 0.561612 | |
| I0623 14:51:23.551838 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.856 > 30) by scale factor 0.83668 | |
| I0623 14:52:03.353504 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.6474 > 30) by scale factor 0.841576 | |
| I0623 14:52:15.609285 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.3726 > 30) by scale factor 0.987732 | |
| I0623 14:52:52.738499 10365 solver.cpp:229] Iteration 3000, loss = 4.81239 | |
| I0623 14:52:52.738693 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.40566 | |
| I0623 14:52:52.738714 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 14:52:52.738728 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 14:52:52.738740 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 14:52:52.738752 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.625 | |
| I0623 14:52:52.738765 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.125 | |
| I0623 14:52:52.738776 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 14:52:52.738788 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0 | |
| I0623 14:52:52.738801 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 14:52:52.738812 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 14:52:52.738823 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.125 | |
| I0623 14:52:52.738836 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.625 | |
| I0623 14:52:52.738847 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 14:52:52.738859 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 14:52:52.738872 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 14:52:52.738883 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 1 | |
| I0623 14:52:52.738894 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 14:52:52.738906 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 14:52:52.738917 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 14:52:52.738929 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 14:52:52.738940 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 14:52:52.738951 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 14:52:52.738963 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 14:52:52.738975 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.602273 | |
| I0623 14:52:52.738986 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.764151 | |
| I0623 14:52:52.739004 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.7083 (* 0.3 = 0.51249 loss) | |
| I0623 14:52:52.739019 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.13982 (* 0.3 = 0.341945 loss) | |
| I0623 14:52:52.739033 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.440445 (* 0.0272727 = 0.0120121 loss) | |
| I0623 14:52:52.739048 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 2.26076 (* 0.0272727 = 0.061657 loss) | |
| I0623 14:52:52.739061 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.89772 (* 0.0272727 = 0.0517559 loss) | |
| I0623 14:52:52.739075 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.83502 (* 0.0272727 = 0.050046 loss) | |
| I0623 14:52:52.739089 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.11274 (* 0.0272727 = 0.0576202 loss) | |
| I0623 14:52:52.739104 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.98587 (* 0.0272727 = 0.05416 loss) | |
| I0623 14:52:52.739117 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 3.20005 (* 0.0272727 = 0.0872741 loss) | |
| I0623 14:52:52.739140 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.92236 (* 0.0272727 = 0.0524279 loss) | |
| I0623 14:52:52.739156 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.00636 (* 0.0272727 = 0.054719 loss) | |
| I0623 14:52:52.739169 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.13686 (* 0.0272727 = 0.058278 loss) | |
| I0623 14:52:52.739183 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.64007 (* 0.0272727 = 0.0447291 loss) | |
| I0623 14:52:52.739197 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.90232 (* 0.0272727 = 0.0518815 loss) | |
| I0623 14:52:52.739212 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.16639 (* 0.0272727 = 0.0590832 loss) | |
| I0623 14:52:52.739248 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.05631 (* 0.0272727 = 0.0288084 loss) | |
| I0623 14:52:52.739264 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.511445 (* 0.0272727 = 0.0139485 loss) | |
| I0623 14:52:52.739279 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.667652 (* 0.0272727 = 0.0182087 loss) | |
| I0623 14:52:52.739295 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.152876 (* 0.0272727 = 0.00416936 loss) | |
| I0623 14:52:52.739308 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0573011 (* 0.0272727 = 0.00156276 loss) | |
| I0623 14:52:52.739322 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0224864 (* 0.0272727 = 0.000613264 loss) | |
| I0623 14:52:52.739337 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00142196 (* 0.0272727 = 3.87807e-05 loss) | |
| I0623 14:52:52.739351 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000250147 (* 0.0272727 = 6.82219e-06 loss) | |
| I0623 14:52:52.739367 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 5.07369e-05 (* 0.0272727 = 1.38373e-06 loss) | |
| I0623 14:52:52.739378 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.528302 | |
| I0623 14:52:52.739392 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 14:52:52.739403 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 14:52:52.739414 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 14:52:52.739426 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 14:52:52.739439 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 14:52:52.739450 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.25 | |
| I0623 14:52:52.739462 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 14:52:52.739473 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.625 | |
| I0623 14:52:52.739485 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.25 | |
| I0623 14:52:52.739497 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 14:52:52.739508 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 14:52:52.739521 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 14:52:52.739531 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 14:52:52.739543 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.5 | |
| I0623 14:52:52.739555 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 14:52:52.739568 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 14:52:52.739579 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 14:52:52.739591 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 14:52:52.739616 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 14:52:52.739629 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 14:52:52.739641 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 14:52:52.739653 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 14:52:52.739665 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.704545 | |
| I0623 14:52:52.739677 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.858491 | |
| I0623 14:52:52.739691 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.29741 (* 0.3 = 0.389222 loss) | |
| I0623 14:52:52.739708 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.818423 (* 0.3 = 0.245527 loss) | |
| I0623 14:52:52.739723 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.267801 (* 0.0272727 = 0.00730367 loss) | |
| I0623 14:52:52.739738 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.382662 (* 0.0272727 = 0.0104362 loss) | |
| I0623 14:52:52.739764 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.861127 (* 0.0272727 = 0.0234853 loss) | |
| I0623 14:52:52.739780 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.853439 (* 0.0272727 = 0.0232756 loss) | |
| I0623 14:52:52.739794 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.91908 (* 0.0272727 = 0.0523384 loss) | |
| I0623 14:52:52.739809 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.79485 (* 0.0272727 = 0.0489506 loss) | |
| I0623 14:52:52.739821 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 2.67437 (* 0.0272727 = 0.0729374 loss) | |
| I0623 14:52:52.739835 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.5347 (* 0.0272727 = 0.0418554 loss) | |
| I0623 14:52:52.739850 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.65995 (* 0.0272727 = 0.0452714 loss) | |
| I0623 14:52:52.739863 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.89547 (* 0.0272727 = 0.0516946 loss) | |
| I0623 14:52:52.739874 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.11307 (* 0.0272727 = 0.0576292 loss) | |
| I0623 14:52:52.739883 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.36235 (* 0.0272727 = 0.0371549 loss) | |
| I0623 14:52:52.739893 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.4795 (* 0.0272727 = 0.04035 loss) | |
| I0623 14:52:52.739907 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.33441 (* 0.0272727 = 0.036393 loss) | |
| I0623 14:52:52.739922 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.548604 (* 0.0272727 = 0.0149619 loss) | |
| I0623 14:52:52.739936 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.363595 (* 0.0272727 = 0.00991624 loss) | |
| I0623 14:52:52.739956 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0089043 (* 0.0272727 = 0.000242844 loss) | |
| I0623 14:52:52.739971 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00556622 (* 0.0272727 = 0.000151806 loss) | |
| I0623 14:52:52.739985 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00191248 (* 0.0272727 = 5.21587e-05 loss) | |
| I0623 14:52:52.740000 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000505539 (* 0.0272727 = 1.37874e-05 loss) | |
| I0623 14:52:52.740013 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 6.95917e-05 (* 0.0272727 = 1.89795e-06 loss) | |
| I0623 14:52:52.740027 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000112214 (* 0.0272727 = 3.06037e-06 loss) | |
| I0623 14:52:52.740041 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.877358 | |
| I0623 14:52:52.740052 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 14:52:52.740063 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 14:52:52.740074 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 14:52:52.740087 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 14:52:52.740097 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 14:52:52.740108 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 14:52:52.740120 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 14:52:52.740131 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 14:52:52.740142 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 14:52:52.740154 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 14:52:52.740166 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.5 | |
| I0623 14:52:52.740178 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.375 | |
| I0623 14:52:52.740190 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.375 | |
| I0623 14:52:52.740201 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 14:52:52.740212 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 14:52:52.740224 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 14:52:52.740245 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 14:52:52.740258 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 14:52:52.740270 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 14:52:52.740283 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 14:52:52.740293 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 14:52:52.740305 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 14:52:52.740319 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.909091 | |
| I0623 14:52:52.740331 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.990566 | |
| I0623 14:52:52.740345 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.408787 (* 1 = 0.408787 loss) | |
| I0623 14:52:52.740360 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.28637 (* 1 = 0.28637 loss) | |
| I0623 14:52:52.740375 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.102309 (* 0.0909091 = 0.00930086 loss) | |
| I0623 14:52:52.740389 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0741905 (* 0.0909091 = 0.00674459 loss) | |
| I0623 14:52:52.740404 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0925174 (* 0.0909091 = 0.00841068 loss) | |
| I0623 14:52:52.740418 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.139711 (* 0.0909091 = 0.012701 loss) | |
| I0623 14:52:52.740432 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.060201 (* 0.0909091 = 0.00547281 loss) | |
| I0623 14:52:52.740447 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.133415 (* 0.0909091 = 0.0121286 loss) | |
| I0623 14:52:52.740461 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.0823182 (* 0.0909091 = 0.00748348 loss) | |
| I0623 14:52:52.740475 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.30457 (* 0.0909091 = 0.0276882 loss) | |
| I0623 14:52:52.740489 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.51384 (* 0.0909091 = 0.0467127 loss) | |
| I0623 14:52:52.740504 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.780804 (* 0.0909091 = 0.0709821 loss) | |
| I0623 14:52:52.740516 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.19866 (* 0.0909091 = 0.108969 loss) | |
| I0623 14:52:52.740530 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.94527 (* 0.0909091 = 0.176843 loss) | |
| I0623 14:52:52.740550 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.26344 (* 0.0909091 = 0.114859 loss) | |
| I0623 14:52:52.740576 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.726534 (* 0.0909091 = 0.0660486 loss) | |
| I0623 14:52:52.740592 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.413277 (* 0.0909091 = 0.0375707 loss) | |
| I0623 14:52:52.740607 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.101109 (* 0.0909091 = 0.00919169 loss) | |
| I0623 14:52:52.740622 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0285528 (* 0.0909091 = 0.00259571 loss) | |
| I0623 14:52:52.740636 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00150093 (* 0.0909091 = 0.000136448 loss) | |
| I0623 14:52:52.740650 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000403559 (* 0.0909091 = 3.66872e-05 loss) | |
| I0623 14:52:52.740665 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 6.15639e-05 (* 0.0909091 = 5.59672e-06 loss) | |
| I0623 14:52:52.740679 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 2.57208e-05 (* 0.0909091 = 2.33825e-06 loss) | |
| I0623 14:52:52.740694 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 1.25471e-05 (* 0.0909091 = 1.14065e-06 loss) | |
| I0623 14:52:52.740706 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 14:52:52.740718 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 14:52:52.740731 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0608495 | |
| I0623 14:52:52.740757 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0372617 | |
| I0623 14:52:52.740773 10365 sgd_solver.cpp:106] Iteration 3000, lr = 0.001 | |
| I0623 14:59:15.378171 10365 solver.cpp:229] Iteration 3500, loss = 4.89668 | |
| I0623 14:59:15.378301 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.4375 | |
| I0623 14:59:15.378321 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 14:59:15.378334 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 14:59:15.378346 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 14:59:15.378358 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 14:59:15.378371 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 14:59:15.378383 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 14:59:15.378396 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.25 | |
| I0623 14:59:15.378407 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.125 | |
| I0623 14:59:15.378419 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 14:59:15.378430 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 14:59:15.378443 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.625 | |
| I0623 14:59:15.378454 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 14:59:15.378465 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 14:59:15.378478 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 14:59:15.378489 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.875 | |
| I0623 14:59:15.378501 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 14:59:15.378512 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 14:59:15.378525 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 14:59:15.378535 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 14:59:15.378547 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 14:59:15.378558 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 14:59:15.378569 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 14:59:15.378582 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.681818 | |
| I0623 14:59:15.378592 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.802083 | |
| I0623 14:59:15.378608 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.69347 (* 0.3 = 0.508041 loss) | |
| I0623 14:59:15.378623 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.972951 (* 0.3 = 0.291885 loss) | |
| I0623 14:59:15.378638 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.656143 (* 0.0272727 = 0.0178948 loss) | |
| I0623 14:59:15.378651 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.38512 (* 0.0272727 = 0.037776 loss) | |
| I0623 14:59:15.378665 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.96425 (* 0.0272727 = 0.0535703 loss) | |
| I0623 14:59:15.378679 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.62156 (* 0.0272727 = 0.0442244 loss) | |
| I0623 14:59:15.378692 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.28502 (* 0.0272727 = 0.0623186 loss) | |
| I0623 14:59:15.378706 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.71377 (* 0.0272727 = 0.0467392 loss) | |
| I0623 14:59:15.378720 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.09839 (* 0.0272727 = 0.0572287 loss) | |
| I0623 14:59:15.378733 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.65771 (* 0.0272727 = 0.072483 loss) | |
| I0623 14:59:15.378746 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.65907 (* 0.0272727 = 0.0452473 loss) | |
| I0623 14:59:15.378760 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.79907 (* 0.0272727 = 0.0490656 loss) | |
| I0623 14:59:15.378774 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.43016 (* 0.0272727 = 0.0390043 loss) | |
| I0623 14:59:15.378787 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.51645 (* 0.0272727 = 0.0413577 loss) | |
| I0623 14:59:15.378818 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.17933 (* 0.0272727 = 0.0321636 loss) | |
| I0623 14:59:15.378834 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.16761 (* 0.0272727 = 0.0318438 loss) | |
| I0623 14:59:15.378847 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.665982 (* 0.0272727 = 0.0181631 loss) | |
| I0623 14:59:15.378861 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.41184 (* 0.0272727 = 0.011232 loss) | |
| I0623 14:59:15.378875 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00993008 (* 0.0272727 = 0.00027082 loss) | |
| I0623 14:59:15.378890 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00128075 (* 0.0272727 = 3.49296e-05 loss) | |
| I0623 14:59:15.378904 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.000160468 (* 0.0272727 = 4.37641e-06 loss) | |
| I0623 14:59:15.378918 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 2.51985e-05 (* 0.0272727 = 6.87231e-07 loss) | |
| I0623 14:59:15.378932 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 2.66443e-05 (* 0.0272727 = 7.26663e-07 loss) | |
| I0623 14:59:15.378947 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.50179e-06 (* 0.0272727 = 9.55033e-08 loss) | |
| I0623 14:59:15.378958 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.59375 | |
| I0623 14:59:15.378970 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 14:59:15.378981 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 14:59:15.378993 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 14:59:15.379004 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 14:59:15.379015 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 14:59:15.379026 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 14:59:15.379039 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 14:59:15.379050 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.125 | |
| I0623 14:59:15.379060 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.625 | |
| I0623 14:59:15.379072 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.25 | |
| I0623 14:59:15.379083 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.25 | |
| I0623 14:59:15.379094 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 14:59:15.379106 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 14:59:15.379117 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 14:59:15.379128 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 14:59:15.379139 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 14:59:15.379150 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 14:59:15.379163 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 14:59:15.379173 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 14:59:15.379184 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 14:59:15.379195 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 14:59:15.379207 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 14:59:15.379218 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.761364 | |
| I0623 14:59:15.379230 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.895833 | |
| I0623 14:59:15.379243 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.18819 (* 0.3 = 0.356456 loss) | |
| I0623 14:59:15.379257 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.707515 (* 0.3 = 0.212254 loss) | |
| I0623 14:59:15.379276 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.397564 (* 0.0272727 = 0.0108427 loss) | |
| I0623 14:59:15.379289 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.267576 (* 0.0272727 = 0.00729752 loss) | |
| I0623 14:59:15.379317 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.457735 (* 0.0272727 = 0.0124837 loss) | |
| I0623 14:59:15.379333 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.7306 (* 0.0272727 = 0.0199255 loss) | |
| I0623 14:59:15.379346 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.38243 (* 0.0272727 = 0.0377025 loss) | |
| I0623 14:59:15.379360 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.72218 (* 0.0272727 = 0.0469686 loss) | |
| I0623 14:59:15.379374 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.32987 (* 0.0272727 = 0.0362693 loss) | |
| I0623 14:59:15.379389 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.49933 (* 0.0272727 = 0.0681636 loss) | |
| I0623 14:59:15.379401 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.08446 (* 0.0272727 = 0.0295762 loss) | |
| I0623 14:59:15.379415 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.66457 (* 0.0272727 = 0.0453972 loss) | |
| I0623 14:59:15.379429 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.93683 (* 0.0272727 = 0.0528225 loss) | |
| I0623 14:59:15.379439 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.26982 (* 0.0272727 = 0.0346315 loss) | |
| I0623 14:59:15.379448 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.93142 (* 0.0272727 = 0.0254024 loss) | |
| I0623 14:59:15.379462 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.16356 (* 0.0272727 = 0.0317336 loss) | |
| I0623 14:59:15.379477 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.557932 (* 0.0272727 = 0.0152163 loss) | |
| I0623 14:59:15.379490 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.419746 (* 0.0272727 = 0.0114476 loss) | |
| I0623 14:59:15.379504 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0407682 (* 0.0272727 = 0.00111186 loss) | |
| I0623 14:59:15.379518 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00311974 (* 0.0272727 = 8.50837e-05 loss) | |
| I0623 14:59:15.379533 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0013185 (* 0.0272727 = 3.5959e-05 loss) | |
| I0623 14:59:15.379545 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000438077 (* 0.0272727 = 1.19476e-05 loss) | |
| I0623 14:59:15.379560 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00010142 (* 0.0272727 = 2.766e-06 loss) | |
| I0623 14:59:15.379573 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 8.50598e-05 (* 0.0272727 = 2.31981e-06 loss) | |
| I0623 14:59:15.379585 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.885417 | |
| I0623 14:59:15.379617 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 14:59:15.379632 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 14:59:15.379644 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 14:59:15.379655 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 14:59:15.379667 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 14:59:15.379678 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 14:59:15.379689 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 14:59:15.379700 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 14:59:15.379711 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.375 | |
| I0623 14:59:15.379722 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 14:59:15.379734 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 14:59:15.379745 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 14:59:15.379756 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 1 | |
| I0623 14:59:15.379767 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 14:59:15.379778 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 1 | |
| I0623 14:59:15.379789 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 14:59:15.379812 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 14:59:15.379825 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 14:59:15.379837 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 14:59:15.379848 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 14:59:15.379858 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 14:59:15.379870 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 14:59:15.379881 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.9375 | |
| I0623 14:59:15.379892 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.979167 | |
| I0623 14:59:15.379906 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.415226 (* 1 = 0.415226 loss) | |
| I0623 14:59:15.379920 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.233742 (* 1 = 0.233742 loss) | |
| I0623 14:59:15.379935 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.103762 (* 0.0909091 = 0.00943293 loss) | |
| I0623 14:59:15.379948 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0679485 (* 0.0909091 = 0.00617714 loss) | |
| I0623 14:59:15.379962 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.137116 (* 0.0909091 = 0.0124651 loss) | |
| I0623 14:59:15.379976 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.158172 (* 0.0909091 = 0.0143793 loss) | |
| I0623 14:59:15.379989 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.108909 (* 0.0909091 = 0.00990086 loss) | |
| I0623 14:59:15.380003 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.119962 (* 0.0909091 = 0.0109057 loss) | |
| I0623 14:59:15.380017 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.389655 (* 0.0909091 = 0.0354232 loss) | |
| I0623 14:59:15.380030 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.447159 (* 0.0909091 = 0.0406508 loss) | |
| I0623 14:59:15.380043 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.00276 (* 0.0909091 = 0.0911601 loss) | |
| I0623 14:59:15.380058 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.839829 (* 0.0909091 = 0.0763481 loss) | |
| I0623 14:59:15.380071 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.50242 (* 0.0909091 = 0.136583 loss) | |
| I0623 14:59:15.380084 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.10702 (* 0.0909091 = 0.100638 loss) | |
| I0623 14:59:15.380098 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.392737 (* 0.0909091 = 0.0357034 loss) | |
| I0623 14:59:15.380111 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.830633 (* 0.0909091 = 0.0755121 loss) | |
| I0623 14:59:15.380125 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.15395 (* 0.0909091 = 0.0139954 loss) | |
| I0623 14:59:15.380138 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.343277 (* 0.0909091 = 0.031207 loss) | |
| I0623 14:59:15.380152 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.00760761 (* 0.0909091 = 0.000691601 loss) | |
| I0623 14:59:15.380167 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00118204 (* 0.0909091 = 0.000107458 loss) | |
| I0623 14:59:15.380182 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000276157 (* 0.0909091 = 2.51052e-05 loss) | |
| I0623 14:59:15.380194 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000101284 (* 0.0909091 = 9.2076e-06 loss) | |
| I0623 14:59:15.380208 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 9.17041e-05 (* 0.0909091 = 8.33673e-06 loss) | |
| I0623 14:59:15.380223 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 5.69719e-05 (* 0.0909091 = 5.17927e-06 loss) | |
| I0623 14:59:15.380234 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 14:59:15.380246 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 14:59:15.380257 10365 solver.cpp:245] Train net output #149: total_confidence = 0.134066 | |
| I0623 14:59:15.380280 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0841617 | |
| I0623 14:59:15.380295 10365 sgd_solver.cpp:106] Iteration 3500, lr = 0.001 | |
| I0623 15:00:57.489078 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.7825 > 30) by scale factor 0.943917 | |
| I0623 15:01:06.678267 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.607 > 30) by scale factor 0.949158 | |
| I0623 15:01:18.913363 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.4911 > 30) by scale factor 0.779401 | |
| I0623 15:05:37.941593 10365 solver.cpp:229] Iteration 4000, loss = 4.7189 | |
| I0623 15:05:37.941735 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.343434 | |
| I0623 15:05:37.941756 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 15:05:37.941768 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.25 | |
| I0623 15:05:37.941781 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 15:05:37.941792 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.125 | |
| I0623 15:05:37.941803 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.5 | |
| I0623 15:05:37.941815 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 15:05:37.941828 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.125 | |
| I0623 15:05:37.941839 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 15:05:37.941851 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 15:05:37.941864 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 15:05:37.941875 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 15:05:37.941886 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 15:05:37.941898 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.375 | |
| I0623 15:05:37.941910 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 15:05:37.941921 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 15:05:37.941933 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 15:05:37.941944 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.75 | |
| I0623 15:05:37.941956 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 15:05:37.941967 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 15:05:37.941979 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:05:37.941990 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:05:37.942003 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:05:37.942013 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.556818 | |
| I0623 15:05:37.942025 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.636364 | |
| I0623 15:05:37.942042 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.97195 (* 0.3 = 0.591586 loss) | |
| I0623 15:05:37.942056 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.56886 (* 0.3 = 0.470657 loss) | |
| I0623 15:05:37.942073 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 1.37548 (* 0.0272727 = 0.0375132 loss) | |
| I0623 15:05:37.942087 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 2.6714 (* 0.0272727 = 0.0728564 loss) | |
| I0623 15:05:37.942101 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.27999 (* 0.0272727 = 0.0621816 loss) | |
| I0623 15:05:37.942114 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 3.90989 (* 0.0272727 = 0.106633 loss) | |
| I0623 15:05:37.942128 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.25643 (* 0.0272727 = 0.061539 loss) | |
| I0623 15:05:37.942142 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.31426 (* 0.0272727 = 0.0631162 loss) | |
| I0623 15:05:37.942157 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.75914 (* 0.0272727 = 0.0752493 loss) | |
| I0623 15:05:37.942170 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.96325 (* 0.0272727 = 0.0535432 loss) | |
| I0623 15:05:37.942184 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.08682 (* 0.0272727 = 0.0569133 loss) | |
| I0623 15:05:37.942198 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.80577 (* 0.0272727 = 0.0492482 loss) | |
| I0623 15:05:37.942211 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.2623 (* 0.0272727 = 0.0616992 loss) | |
| I0623 15:05:37.942224 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.81972 (* 0.0272727 = 0.0496286 loss) | |
| I0623 15:05:37.942251 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.6073 (* 0.0272727 = 0.0438355 loss) | |
| I0623 15:05:37.942270 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.35362 (* 0.0272727 = 0.036917 loss) | |
| I0623 15:05:37.942284 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.05002 (* 0.0272727 = 0.0286369 loss) | |
| I0623 15:05:37.942298 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.814023 (* 0.0272727 = 0.0222006 loss) | |
| I0623 15:05:37.942312 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.791333 (* 0.0272727 = 0.0215818 loss) | |
| I0623 15:05:37.942325 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.336407 (* 0.0272727 = 0.00917474 loss) | |
| I0623 15:05:37.942339 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.136067 (* 0.0272727 = 0.00371092 loss) | |
| I0623 15:05:37.942353 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.0621995 (* 0.0272727 = 0.00169635 loss) | |
| I0623 15:05:37.942366 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.0048601 (* 0.0272727 = 0.000132548 loss) | |
| I0623 15:05:37.942380 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 8.88681e-05 (* 0.0272727 = 2.42367e-06 loss) | |
| I0623 15:05:37.942394 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.464646 | |
| I0623 15:05:37.942405 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 15:05:37.942416 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 15:05:37.942427 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.5 | |
| I0623 15:05:37.942440 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 15:05:37.942451 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 15:05:37.942461 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 15:05:37.942472 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.125 | |
| I0623 15:05:37.942484 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.5 | |
| I0623 15:05:37.942495 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.25 | |
| I0623 15:05:37.942507 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.625 | |
| I0623 15:05:37.942517 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 15:05:37.942529 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.25 | |
| I0623 15:05:37.942540 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 15:05:37.942551 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.5 | |
| I0623 15:05:37.942562 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 15:05:37.942574 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.5 | |
| I0623 15:05:37.942585 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.75 | |
| I0623 15:05:37.942596 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 15:05:37.942607 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 15:05:37.942620 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:05:37.942631 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:05:37.942641 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:05:37.942652 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.625 | |
| I0623 15:05:37.942664 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.79798 | |
| I0623 15:05:37.942677 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.64319 (* 0.3 = 0.492958 loss) | |
| I0623 15:05:37.942692 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 1.35236 (* 0.3 = 0.405709 loss) | |
| I0623 15:05:37.942705 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 1.74263 (* 0.0272727 = 0.0475262 loss) | |
| I0623 15:05:37.942719 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.891262 (* 0.0272727 = 0.0243071 loss) | |
| I0623 15:05:37.942744 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 2.23443 (* 0.0272727 = 0.060939 loss) | |
| I0623 15:05:37.942762 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 3.55954 (* 0.0272727 = 0.0970784 loss) | |
| I0623 15:05:37.942776 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.89693 (* 0.0272727 = 0.0517344 loss) | |
| I0623 15:05:37.942790 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.95617 (* 0.0272727 = 0.05335 loss) | |
| I0623 15:05:37.942803 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 3.26143 (* 0.0272727 = 0.0889481 loss) | |
| I0623 15:05:37.942816 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.26939 (* 0.0272727 = 0.0618923 loss) | |
| I0623 15:05:37.942829 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 2.17293 (* 0.0272727 = 0.0592617 loss) | |
| I0623 15:05:37.942843 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.78501 (* 0.0272727 = 0.0486822 loss) | |
| I0623 15:05:37.942857 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.9837 (* 0.0272727 = 0.054101 loss) | |
| I0623 15:05:37.942870 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.34215 (* 0.0272727 = 0.0366041 loss) | |
| I0623 15:05:37.942883 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.49029 (* 0.0272727 = 0.0406443 loss) | |
| I0623 15:05:37.942896 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.85396 (* 0.0272727 = 0.0505626 loss) | |
| I0623 15:05:37.942910 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.2933 (* 0.0272727 = 0.0352719 loss) | |
| I0623 15:05:37.942924 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 1.20277 (* 0.0272727 = 0.0328029 loss) | |
| I0623 15:05:37.942937 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.750674 (* 0.0272727 = 0.0204729 loss) | |
| I0623 15:05:37.942950 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.396116 (* 0.0272727 = 0.0108032 loss) | |
| I0623 15:05:37.942965 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.101589 (* 0.0272727 = 0.0027706 loss) | |
| I0623 15:05:37.942978 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.0467076 (* 0.0272727 = 0.00127384 loss) | |
| I0623 15:05:37.942992 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00156432 (* 0.0272727 = 4.26634e-05 loss) | |
| I0623 15:05:37.943006 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.00111674 (* 0.0272727 = 3.04566e-05 loss) | |
| I0623 15:05:37.943017 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.636364 | |
| I0623 15:05:37.943030 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.875 | |
| I0623 15:05:37.943042 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 15:05:37.943053 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.75 | |
| I0623 15:05:37.943064 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.75 | |
| I0623 15:05:37.943075 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.75 | |
| I0623 15:05:37.943087 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.625 | |
| I0623 15:05:37.943099 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.5 | |
| I0623 15:05:37.943110 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.625 | |
| I0623 15:05:37.943123 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.25 | |
| I0623 15:05:37.943136 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 15:05:37.943147 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.125 | |
| I0623 15:05:37.943159 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.25 | |
| I0623 15:05:37.943171 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.375 | |
| I0623 15:05:37.943183 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.375 | |
| I0623 15:05:37.943192 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 15:05:37.943199 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.625 | |
| I0623 15:05:37.943222 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.75 | |
| I0623 15:05:37.943234 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 15:05:37.943245 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 15:05:37.943256 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:05:37.943267 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:05:37.943279 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:05:37.943289 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.721591 | |
| I0623 15:05:37.943301 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.959596 | |
| I0623 15:05:37.943315 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 1.16799 (* 1 = 1.16799 loss) | |
| I0623 15:05:37.943328 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 1.11914 (* 1 = 1.11914 loss) | |
| I0623 15:05:37.943341 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 1.26992 (* 0.0909091 = 0.115448 loss) | |
| I0623 15:05:37.943356 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 1.13651 (* 0.0909091 = 0.103319 loss) | |
| I0623 15:05:37.943368 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 1.5697 (* 0.0909091 = 0.1427 loss) | |
| I0623 15:05:37.943382 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 1.70312 (* 0.0909091 = 0.15483 loss) | |
| I0623 15:05:37.943395 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 1.22748 (* 0.0909091 = 0.111589 loss) | |
| I0623 15:05:37.943409 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 2.64739 (* 0.0909091 = 0.240672 loss) | |
| I0623 15:05:37.943423 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 2.34839 (* 0.0909091 = 0.21349 loss) | |
| I0623 15:05:37.943435 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.75045 (* 0.0909091 = 0.159131 loss) | |
| I0623 15:05:37.943449 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 2.01441 (* 0.0909091 = 0.183128 loss) | |
| I0623 15:05:37.943461 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.22626 (* 0.0909091 = 0.111478 loss) | |
| I0623 15:05:37.943475 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.95582 (* 0.0909091 = 0.177802 loss) | |
| I0623 15:05:37.943488 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.81106 (* 0.0909091 = 0.164642 loss) | |
| I0623 15:05:37.943501 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.31545 (* 0.0909091 = 0.119587 loss) | |
| I0623 15:05:37.943514 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.26388 (* 0.0909091 = 0.114899 loss) | |
| I0623 15:05:37.943527 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.12719 (* 0.0909091 = 0.102471 loss) | |
| I0623 15:05:37.943541 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.783099 (* 0.0909091 = 0.0711908 loss) | |
| I0623 15:05:37.943553 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.550135 (* 0.0909091 = 0.0500123 loss) | |
| I0623 15:05:37.943567 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0983122 (* 0.0909091 = 0.00893748 loss) | |
| I0623 15:05:37.943580 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0177637 (* 0.0909091 = 0.00161488 loss) | |
| I0623 15:05:37.943593 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00703782 (* 0.0909091 = 0.000639802 loss) | |
| I0623 15:05:37.943624 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.001339 (* 0.0909091 = 0.000121727 loss) | |
| I0623 15:05:37.943637 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.79349e-05 (* 0.0909091 = 4.35772e-06 loss) | |
| I0623 15:05:37.943650 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 15:05:37.943660 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 15:05:37.943672 10365 solver.cpp:245] Train net output #149: total_confidence = 0.124627 | |
| I0623 15:05:37.943696 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.127342 | |
| I0623 15:05:37.943709 10365 sgd_solver.cpp:106] Iteration 4000, lr = 0.001 | |
| I0623 15:06:53.232090 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 62.8073 > 30) by scale factor 0.477651 | |
| I0623 15:07:12.369156 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 37.0164 > 30) by scale factor 0.810451 | |
| I0623 15:11:01.956571 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.2599 > 30) by scale factor 0.850824 | |
| I0623 15:11:44.034663 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.9961 > 30) by scale factor 0.882453 | |
| I0623 15:12:00.529906 10365 solver.cpp:229] Iteration 4500, loss = 4.75575 | |
| I0623 15:12:00.529971 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.521277 | |
| I0623 15:12:00.529989 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 15:12:00.530002 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 15:12:00.530014 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.625 | |
| I0623 15:12:00.530026 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 15:12:00.530038 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 15:12:00.530050 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 15:12:00.530062 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 15:12:00.530076 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 15:12:00.530089 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.625 | |
| I0623 15:12:00.530100 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 15:12:00.530112 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 15:12:00.530123 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 15:12:00.530135 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 15:12:00.530146 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 15:12:00.530158 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.875 | |
| I0623 15:12:00.530170 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 15:12:00.530181 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 15:12:00.530194 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 15:12:00.530205 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 15:12:00.530215 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:12:00.530227 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:12:00.530238 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:12:00.530251 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.727273 | |
| I0623 15:12:00.530262 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.776596 | |
| I0623 15:12:00.530278 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.46941 (* 0.3 = 0.440823 loss) | |
| I0623 15:12:00.530292 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.837564 (* 0.3 = 0.251269 loss) | |
| I0623 15:12:00.530306 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.537703 (* 0.0272727 = 0.0146646 loss) | |
| I0623 15:12:00.530320 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.2214 (* 0.0272727 = 0.0333109 loss) | |
| I0623 15:12:00.530334 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 0.9079 (* 0.0272727 = 0.0247609 loss) | |
| I0623 15:12:00.530349 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.45341 (* 0.0272727 = 0.0396386 loss) | |
| I0623 15:12:00.530361 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.19488 (* 0.0272727 = 0.0598604 loss) | |
| I0623 15:12:00.530375 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.53633 (* 0.0272727 = 0.0418998 loss) | |
| I0623 15:12:00.530390 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.28028 (* 0.0272727 = 0.0621895 loss) | |
| I0623 15:12:00.530403 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.25325 (* 0.0272727 = 0.0341796 loss) | |
| I0623 15:12:00.530416 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.37647 (* 0.0272727 = 0.0375402 loss) | |
| I0623 15:12:00.530429 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.12058 (* 0.0272727 = 0.0578339 loss) | |
| I0623 15:12:00.530443 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.75101 (* 0.0272727 = 0.0477547 loss) | |
| I0623 15:12:00.530457 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.43266 (* 0.0272727 = 0.0390725 loss) | |
| I0623 15:12:00.530500 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.40447 (* 0.0272727 = 0.0383037 loss) | |
| I0623 15:12:00.530525 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.10244 (* 0.0272727 = 0.0300666 loss) | |
| I0623 15:12:00.530539 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.597345 (* 0.0272727 = 0.0162912 loss) | |
| I0623 15:12:00.530553 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.797813 (* 0.0272727 = 0.0217585 loss) | |
| I0623 15:12:00.530567 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00693995 (* 0.0272727 = 0.000189271 loss) | |
| I0623 15:12:00.530581 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00114927 (* 0.0272727 = 3.13438e-05 loss) | |
| I0623 15:12:00.530596 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0002945 (* 0.0272727 = 8.03183e-06 loss) | |
| I0623 15:12:00.530609 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 6.41628e-05 (* 0.0272727 = 1.74989e-06 loss) | |
| I0623 15:12:00.530623 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 1.62427e-05 (* 0.0272727 = 4.42983e-07 loss) | |
| I0623 15:12:00.530637 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 4.27667e-06 (* 0.0272727 = 1.16636e-07 loss) | |
| I0623 15:12:00.530650 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.606383 | |
| I0623 15:12:00.530663 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 15:12:00.530673 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 15:12:00.530685 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 15:12:00.530696 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 15:12:00.530707 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.125 | |
| I0623 15:12:00.530719 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 15:12:00.530730 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 15:12:00.530742 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.625 | |
| I0623 15:12:00.530756 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.25 | |
| I0623 15:12:00.530768 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.75 | |
| I0623 15:12:00.530779 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 15:12:00.530791 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 15:12:00.530802 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 15:12:00.530813 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 15:12:00.530824 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 15:12:00.530836 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 15:12:00.530848 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 15:12:00.530858 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 15:12:00.530869 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 15:12:00.530880 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:12:00.530892 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:12:00.530902 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:12:00.530915 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.789773 | |
| I0623 15:12:00.530925 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.882979 | |
| I0623 15:12:00.530939 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.16585 (* 0.3 = 0.349755 loss) | |
| I0623 15:12:00.530952 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.638165 (* 0.3 = 0.19145 loss) | |
| I0623 15:12:00.530966 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.363134 (* 0.0272727 = 0.00990364 loss) | |
| I0623 15:12:00.530992 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.462497 (* 0.0272727 = 0.0126136 loss) | |
| I0623 15:12:00.531008 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.394001 (* 0.0272727 = 0.0107455 loss) | |
| I0623 15:12:00.531021 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.971272 (* 0.0272727 = 0.0264892 loss) | |
| I0623 15:12:00.531034 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.14806 (* 0.0272727 = 0.0585835 loss) | |
| I0623 15:12:00.531049 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.2416 (* 0.0272727 = 0.0338618 loss) | |
| I0623 15:12:00.531061 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.97814 (* 0.0272727 = 0.0539492 loss) | |
| I0623 15:12:00.531075 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.19017 (* 0.0272727 = 0.0324592 loss) | |
| I0623 15:12:00.531088 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.75951 (* 0.0272727 = 0.0479867 loss) | |
| I0623 15:12:00.531101 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.30373 (* 0.0272727 = 0.0355562 loss) | |
| I0623 15:12:00.531114 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.43326 (* 0.0272727 = 0.0390888 loss) | |
| I0623 15:12:00.531131 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.45376 (* 0.0272727 = 0.0396481 loss) | |
| I0623 15:12:00.531146 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.10756 (* 0.0272727 = 0.0302061 loss) | |
| I0623 15:12:00.531158 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.798834 (* 0.0272727 = 0.0217864 loss) | |
| I0623 15:12:00.531172 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.568323 (* 0.0272727 = 0.0154997 loss) | |
| I0623 15:12:00.531185 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.513104 (* 0.0272727 = 0.0139937 loss) | |
| I0623 15:12:00.531199 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.00900011 (* 0.0272727 = 0.000245458 loss) | |
| I0623 15:12:00.531213 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00131582 (* 0.0272727 = 3.58861e-05 loss) | |
| I0623 15:12:00.531226 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 5.15837e-05 (* 0.0272727 = 1.40683e-06 loss) | |
| I0623 15:12:00.531240 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 1.23981e-05 (* 0.0272727 = 3.38129e-07 loss) | |
| I0623 15:12:00.531255 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 4.90252e-06 (* 0.0272727 = 1.33705e-07 loss) | |
| I0623 15:12:00.531268 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 3.69553e-06 (* 0.0272727 = 1.00787e-07 loss) | |
| I0623 15:12:00.531280 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.904255 | |
| I0623 15:12:00.531292 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 15:12:00.531304 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 15:12:00.531316 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 15:12:00.531327 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 15:12:00.531338 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 15:12:00.531349 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 15:12:00.531360 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 15:12:00.531371 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 15:12:00.531383 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 15:12:00.531394 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 15:12:00.531406 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.875 | |
| I0623 15:12:00.531417 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 15:12:00.531429 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 15:12:00.531443 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 15:12:00.531455 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 15:12:00.531476 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 15:12:00.531489 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 15:12:00.531500 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 15:12:00.531510 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 15:12:00.531522 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:12:00.531533 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:12:00.531543 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:12:00.531554 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.926136 | |
| I0623 15:12:00.531566 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 1 | |
| I0623 15:12:00.531579 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.291806 (* 1 = 0.291806 loss) | |
| I0623 15:12:00.531592 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.188262 (* 1 = 0.188262 loss) | |
| I0623 15:12:00.531621 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0282195 (* 0.0909091 = 0.00256541 loss) | |
| I0623 15:12:00.531635 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0808289 (* 0.0909091 = 0.00734808 loss) | |
| I0623 15:12:00.531649 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0501915 (* 0.0909091 = 0.00456286 loss) | |
| I0623 15:12:00.531663 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.07595 (* 0.0909091 = 0.00690454 loss) | |
| I0623 15:12:00.531677 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.114051 (* 0.0909091 = 0.0103682 loss) | |
| I0623 15:12:00.531690 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.103858 (* 0.0909091 = 0.0094416 loss) | |
| I0623 15:12:00.531704 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.0960977 (* 0.0909091 = 0.00873615 loss) | |
| I0623 15:12:00.531718 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.743839 (* 0.0909091 = 0.0676217 loss) | |
| I0623 15:12:00.531731 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.749018 (* 0.0909091 = 0.0680925 loss) | |
| I0623 15:12:00.531744 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.742976 (* 0.0909091 = 0.0675432 loss) | |
| I0623 15:12:00.531757 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.556033 (* 0.0909091 = 0.0505484 loss) | |
| I0623 15:12:00.531771 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.868181 (* 0.0909091 = 0.0789256 loss) | |
| I0623 15:12:00.531785 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.661758 (* 0.0909091 = 0.0601598 loss) | |
| I0623 15:12:00.531797 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.789923 (* 0.0909091 = 0.0718112 loss) | |
| I0623 15:12:00.531815 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.641951 (* 0.0909091 = 0.0583592 loss) | |
| I0623 15:12:00.531828 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.198909 (* 0.0909091 = 0.0180826 loss) | |
| I0623 15:12:00.531841 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.00485608 (* 0.0909091 = 0.000441462 loss) | |
| I0623 15:12:00.531855 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000565632 (* 0.0909091 = 5.14211e-05 loss) | |
| I0623 15:12:00.531868 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 7.45254e-05 (* 0.0909091 = 6.77504e-06 loss) | |
| I0623 15:12:00.531883 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 3.76723e-05 (* 0.0909091 = 3.42475e-06 loss) | |
| I0623 15:12:00.531895 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 2.35898e-05 (* 0.0909091 = 2.14453e-06 loss) | |
| I0623 15:12:00.531909 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 3.6359e-06 (* 0.0909091 = 3.30536e-07 loss) | |
| I0623 15:12:00.531921 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 15:12:00.531934 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 15:12:00.531957 10365 solver.cpp:245] Train net output #149: total_confidence = 0.147452 | |
| I0623 15:12:00.531970 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.121225 | |
| I0623 15:12:00.531983 10365 sgd_solver.cpp:106] Iteration 4500, lr = 0.001 | |
| I0623 15:12:18.493130 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.3915 > 30) by scale factor 0.824368 | |
| I0623 15:13:15.864212 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.4634 > 30) by scale factor 0.896503 | |
| I0623 15:14:10.181311 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 37.0829 > 30) by scale factor 0.808998 | |
| I0623 15:15:03.746193 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.2962 > 30) by scale factor 0.849951 | |
| I0623 15:15:41.224390 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.909 > 30) by scale factor 0.97059 | |
| I0623 15:18:22.618973 10365 solver.cpp:338] Iteration 5000, Testing net (#0) | |
| I0623 15:19:19.799612 10365 solver.cpp:393] Test loss: 4.13349 | |
| I0623 15:19:19.799731 10365 solver.cpp:406] Test net output #0: loss1/accuracy = 0.50051 | |
| I0623 15:19:19.799751 10365 solver.cpp:406] Test net output #1: loss1/accuracy01 = 0.923 | |
| I0623 15:19:19.799764 10365 solver.cpp:406] Test net output #2: loss1/accuracy02 = 0.761 | |
| I0623 15:19:19.799777 10365 solver.cpp:406] Test net output #3: loss1/accuracy03 = 0.533 | |
| I0623 15:19:19.799788 10365 solver.cpp:406] Test net output #4: loss1/accuracy04 = 0.428 | |
| I0623 15:19:19.799800 10365 solver.cpp:406] Test net output #5: loss1/accuracy05 = 0.4 | |
| I0623 15:19:19.799811 10365 solver.cpp:406] Test net output #6: loss1/accuracy06 = 0.402 | |
| I0623 15:19:19.799823 10365 solver.cpp:406] Test net output #7: loss1/accuracy07 = 0.359 | |
| I0623 15:19:19.799834 10365 solver.cpp:406] Test net output #8: loss1/accuracy08 = 0.478 | |
| I0623 15:19:19.799846 10365 solver.cpp:406] Test net output #9: loss1/accuracy09 = 0.416 | |
| I0623 15:19:19.799857 10365 solver.cpp:406] Test net output #10: loss1/accuracy10 = 0.41 | |
| I0623 15:19:19.799868 10365 solver.cpp:406] Test net output #11: loss1/accuracy11 = 0.39 | |
| I0623 15:19:19.799880 10365 solver.cpp:406] Test net output #12: loss1/accuracy12 = 0.484 | |
| I0623 15:19:19.799891 10365 solver.cpp:406] Test net output #13: loss1/accuracy13 = 0.591 | |
| I0623 15:19:19.799902 10365 solver.cpp:406] Test net output #14: loss1/accuracy14 = 0.671 | |
| I0623 15:19:19.799913 10365 solver.cpp:406] Test net output #15: loss1/accuracy15 = 0.769 | |
| I0623 15:19:19.799924 10365 solver.cpp:406] Test net output #16: loss1/accuracy16 = 0.827 | |
| I0623 15:19:19.799935 10365 solver.cpp:406] Test net output #17: loss1/accuracy17 = 0.902 | |
| I0623 15:19:19.799947 10365 solver.cpp:406] Test net output #18: loss1/accuracy18 = 0.949 | |
| I0623 15:19:19.799957 10365 solver.cpp:406] Test net output #19: loss1/accuracy19 = 0.972 | |
| I0623 15:19:19.799969 10365 solver.cpp:406] Test net output #20: loss1/accuracy20 = 0.987 | |
| I0623 15:19:19.799980 10365 solver.cpp:406] Test net output #21: loss1/accuracy21 = 0.999 | |
| I0623 15:19:19.799991 10365 solver.cpp:406] Test net output #22: loss1/accuracy22 = 1 | |
| I0623 15:19:19.800003 10365 solver.cpp:406] Test net output #23: loss1/accuracy_incl_empty = 0.694182 | |
| I0623 15:19:19.800014 10365 solver.cpp:406] Test net output #24: loss1/accuracy_top3 = 0.83487 | |
| I0623 15:19:19.800029 10365 solver.cpp:406] Test net output #25: loss1/cross_entropy_loss = 1.45476 (* 0.3 = 0.436429 loss) | |
| I0623 15:19:19.800043 10365 solver.cpp:406] Test net output #26: loss1/cross_entropy_loss_incl_empty = 0.894121 (* 0.3 = 0.268236 loss) | |
| I0623 15:19:19.800058 10365 solver.cpp:406] Test net output #27: loss1/loss01 = 0.343591 (* 0.0272727 = 0.00937065 loss) | |
| I0623 15:19:19.800072 10365 solver.cpp:406] Test net output #28: loss1/loss02 = 0.814289 (* 0.0272727 = 0.0222079 loss) | |
| I0623 15:19:19.800086 10365 solver.cpp:406] Test net output #29: loss1/loss03 = 1.47671 (* 0.0272727 = 0.0402738 loss) | |
| I0623 15:19:19.800099 10365 solver.cpp:406] Test net output #30: loss1/loss04 = 1.65237 (* 0.0272727 = 0.0450646 loss) | |
| I0623 15:19:19.800112 10365 solver.cpp:406] Test net output #31: loss1/loss05 = 1.79879 (* 0.0272727 = 0.0490578 loss) | |
| I0623 15:19:19.800127 10365 solver.cpp:406] Test net output #32: loss1/loss06 = 1.89115 (* 0.0272727 = 0.0515768 loss) | |
| I0623 15:19:19.800139 10365 solver.cpp:406] Test net output #33: loss1/loss07 = 1.90352 (* 0.0272727 = 0.0519142 loss) | |
| I0623 15:19:19.800153 10365 solver.cpp:406] Test net output #34: loss1/loss08 = 1.6837 (* 0.0272727 = 0.045919 loss) | |
| I0623 15:19:19.800165 10365 solver.cpp:406] Test net output #35: loss1/loss09 = 1.77611 (* 0.0272727 = 0.0484394 loss) | |
| I0623 15:19:19.800179 10365 solver.cpp:406] Test net output #36: loss1/loss10 = 1.80655 (* 0.0272727 = 0.0492694 loss) | |
| I0623 15:19:19.800192 10365 solver.cpp:406] Test net output #37: loss1/loss11 = 1.89256 (* 0.0272727 = 0.0516152 loss) | |
| I0623 15:19:19.800205 10365 solver.cpp:406] Test net output #38: loss1/loss12 = 1.56893 (* 0.0272727 = 0.042789 loss) | |
| I0623 15:19:19.800218 10365 solver.cpp:406] Test net output #39: loss1/loss13 = 1.25648 (* 0.0272727 = 0.0342676 loss) | |
| I0623 15:19:19.800251 10365 solver.cpp:406] Test net output #40: loss1/loss14 = 0.972732 (* 0.0272727 = 0.0265291 loss) | |
| I0623 15:19:19.800269 10365 solver.cpp:406] Test net output #41: loss1/loss15 = 0.693935 (* 0.0272727 = 0.0189255 loss) | |
| I0623 15:19:19.800283 10365 solver.cpp:406] Test net output #42: loss1/loss16 = 0.534658 (* 0.0272727 = 0.0145816 loss) | |
| I0623 15:19:19.800297 10365 solver.cpp:406] Test net output #43: loss1/loss17 = 0.341222 (* 0.0272727 = 0.00930606 loss) | |
| I0623 15:19:19.800312 10365 solver.cpp:406] Test net output #44: loss1/loss18 = 0.198091 (* 0.0272727 = 0.00540249 loss) | |
| I0623 15:19:19.800324 10365 solver.cpp:406] Test net output #45: loss1/loss19 = 0.126552 (* 0.0272727 = 0.00345142 loss) | |
| I0623 15:19:19.800338 10365 solver.cpp:406] Test net output #46: loss1/loss20 = 0.0713861 (* 0.0272727 = 0.00194689 loss) | |
| I0623 15:19:19.800351 10365 solver.cpp:406] Test net output #47: loss1/loss21 = 0.00713148 (* 0.0272727 = 0.000194495 loss) | |
| I0623 15:19:19.800365 10365 solver.cpp:406] Test net output #48: loss1/loss22 = 8.71755e-05 (* 0.0272727 = 2.37751e-06 loss) | |
| I0623 15:19:19.800377 10365 solver.cpp:406] Test net output #49: loss2/accuracy = 0.61994 | |
| I0623 15:19:19.800389 10365 solver.cpp:406] Test net output #50: loss2/accuracy01 = 0.971 | |
| I0623 15:19:19.800400 10365 solver.cpp:406] Test net output #51: loss2/accuracy02 = 0.945 | |
| I0623 15:19:19.800411 10365 solver.cpp:406] Test net output #52: loss2/accuracy03 = 0.832 | |
| I0623 15:19:19.800422 10365 solver.cpp:406] Test net output #53: loss2/accuracy04 = 0.704 | |
| I0623 15:19:19.800433 10365 solver.cpp:406] Test net output #54: loss2/accuracy05 = 0.537 | |
| I0623 15:19:19.800444 10365 solver.cpp:406] Test net output #55: loss2/accuracy06 = 0.484 | |
| I0623 15:19:19.800456 10365 solver.cpp:406] Test net output #56: loss2/accuracy07 = 0.454 | |
| I0623 15:19:19.800467 10365 solver.cpp:406] Test net output #57: loss2/accuracy08 = 0.506 | |
| I0623 15:19:19.800477 10365 solver.cpp:406] Test net output #58: loss2/accuracy09 = 0.473 | |
| I0623 15:19:19.800488 10365 solver.cpp:406] Test net output #59: loss2/accuracy10 = 0.42 | |
| I0623 15:19:19.800499 10365 solver.cpp:406] Test net output #60: loss2/accuracy11 = 0.414 | |
| I0623 15:19:19.800510 10365 solver.cpp:406] Test net output #61: loss2/accuracy12 = 0.524 | |
| I0623 15:19:19.800521 10365 solver.cpp:406] Test net output #62: loss2/accuracy13 = 0.61 | |
| I0623 15:19:19.800532 10365 solver.cpp:406] Test net output #63: loss2/accuracy14 = 0.693 | |
| I0623 15:19:19.800544 10365 solver.cpp:406] Test net output #64: loss2/accuracy15 = 0.778 | |
| I0623 15:19:19.800554 10365 solver.cpp:406] Test net output #65: loss2/accuracy16 = 0.839 | |
| I0623 15:19:19.800565 10365 solver.cpp:406] Test net output #66: loss2/accuracy17 = 0.905 | |
| I0623 15:19:19.800576 10365 solver.cpp:406] Test net output #67: loss2/accuracy18 = 0.95 | |
| I0623 15:19:19.800587 10365 solver.cpp:406] Test net output #68: loss2/accuracy19 = 0.972 | |
| I0623 15:19:19.800598 10365 solver.cpp:406] Test net output #69: loss2/accuracy20 = 0.987 | |
| I0623 15:19:19.800609 10365 solver.cpp:406] Test net output #70: loss2/accuracy21 = 0.999 | |
| I0623 15:19:19.800621 10365 solver.cpp:406] Test net output #71: loss2/accuracy22 = 1 | |
| I0623 15:19:19.800631 10365 solver.cpp:406] Test net output #72: loss2/accuracy_incl_empty = 0.759409 | |
| I0623 15:19:19.800643 10365 solver.cpp:406] Test net output #73: loss2/accuracy_top3 = 0.900178 | |
| I0623 15:19:19.800657 10365 solver.cpp:406] Test net output #74: loss2/cross_entropy_loss = 1.09847 (* 0.3 = 0.32954 loss) | |
| I0623 15:19:19.800669 10365 solver.cpp:406] Test net output #75: loss2/cross_entropy_loss_incl_empty = 0.691491 (* 0.3 = 0.207447 loss) | |
| I0623 15:19:19.800684 10365 solver.cpp:406] Test net output #76: loss2/loss01 = 0.195611 (* 0.0272727 = 0.00533484 loss) | |
| I0623 15:19:19.800698 10365 solver.cpp:406] Test net output #77: loss2/loss02 = 0.290989 (* 0.0272727 = 0.00793606 loss) | |
| I0623 15:19:19.800722 10365 solver.cpp:406] Test net output #78: loss2/loss03 = 0.624667 (* 0.0272727 = 0.0170364 loss) | |
| I0623 15:19:19.800740 10365 solver.cpp:406] Test net output #79: loss2/loss04 = 0.966282 (* 0.0272727 = 0.0263531 loss) | |
| I0623 15:19:19.800752 10365 solver.cpp:406] Test net output #80: loss2/loss05 = 1.26002 (* 0.0272727 = 0.0343641 loss) | |
| I0623 15:19:19.800766 10365 solver.cpp:406] Test net output #81: loss2/loss06 = 1.51084 (* 0.0272727 = 0.0412047 loss) | |
| I0623 15:19:19.800779 10365 solver.cpp:406] Test net output #82: loss2/loss07 = 1.60048 (* 0.0272727 = 0.0436496 loss) | |
| I0623 15:19:19.800792 10365 solver.cpp:406] Test net output #83: loss2/loss08 = 1.49974 (* 0.0272727 = 0.040902 loss) | |
| I0623 15:19:19.800806 10365 solver.cpp:406] Test net output #84: loss2/loss09 = 1.56109 (* 0.0272727 = 0.0425752 loss) | |
| I0623 15:19:19.800819 10365 solver.cpp:406] Test net output #85: loss2/loss10 = 1.64685 (* 0.0272727 = 0.044914 loss) | |
| I0623 15:19:19.800832 10365 solver.cpp:406] Test net output #86: loss2/loss11 = 1.71539 (* 0.0272727 = 0.0467835 loss) | |
| I0623 15:19:19.800845 10365 solver.cpp:406] Test net output #87: loss2/loss12 = 1.38885 (* 0.0272727 = 0.0378777 loss) | |
| I0623 15:19:19.800858 10365 solver.cpp:406] Test net output #88: loss2/loss13 = 1.14341 (* 0.0272727 = 0.031184 loss) | |
| I0623 15:19:19.800871 10365 solver.cpp:406] Test net output #89: loss2/loss14 = 0.869613 (* 0.0272727 = 0.0237167 loss) | |
| I0623 15:19:19.800884 10365 solver.cpp:406] Test net output #90: loss2/loss15 = 0.629922 (* 0.0272727 = 0.0171797 loss) | |
| I0623 15:19:19.800897 10365 solver.cpp:406] Test net output #91: loss2/loss16 = 0.470202 (* 0.0272727 = 0.0128237 loss) | |
| I0623 15:19:19.800910 10365 solver.cpp:406] Test net output #92: loss2/loss17 = 0.322601 (* 0.0272727 = 0.0087982 loss) | |
| I0623 15:19:19.800925 10365 solver.cpp:406] Test net output #93: loss2/loss18 = 0.177043 (* 0.0272727 = 0.00482843 loss) | |
| I0623 15:19:19.800937 10365 solver.cpp:406] Test net output #94: loss2/loss19 = 0.111702 (* 0.0272727 = 0.00304642 loss) | |
| I0623 15:19:19.800951 10365 solver.cpp:406] Test net output #95: loss2/loss20 = 0.0664249 (* 0.0272727 = 0.00181159 loss) | |
| I0623 15:19:19.800964 10365 solver.cpp:406] Test net output #96: loss2/loss21 = 0.00825273 (* 0.0272727 = 0.000225074 loss) | |
| I0623 15:19:19.800977 10365 solver.cpp:406] Test net output #97: loss2/loss22 = 7.74321e-05 (* 0.0272727 = 2.11178e-06 loss) | |
| I0623 15:19:19.800989 10365 solver.cpp:406] Test net output #98: loss3/accuracy = 0.869437 | |
| I0623 15:19:19.801000 10365 solver.cpp:406] Test net output #99: loss3/accuracy01 = 0.975 | |
| I0623 15:19:19.801012 10365 solver.cpp:406] Test net output #100: loss3/accuracy02 = 0.97 | |
| I0623 15:19:19.801023 10365 solver.cpp:406] Test net output #101: loss3/accuracy03 = 0.953 | |
| I0623 15:19:19.801033 10365 solver.cpp:406] Test net output #102: loss3/accuracy04 = 0.945 | |
| I0623 15:19:19.801044 10365 solver.cpp:406] Test net output #103: loss3/accuracy05 = 0.933 | |
| I0623 15:19:19.801055 10365 solver.cpp:406] Test net output #104: loss3/accuracy06 = 0.899 | |
| I0623 15:19:19.801065 10365 solver.cpp:406] Test net output #105: loss3/accuracy07 = 0.893 | |
| I0623 15:19:19.801076 10365 solver.cpp:406] Test net output #106: loss3/accuracy08 = 0.869 | |
| I0623 15:19:19.801087 10365 solver.cpp:406] Test net output #107: loss3/accuracy09 = 0.786 | |
| I0623 15:19:19.801097 10365 solver.cpp:406] Test net output #108: loss3/accuracy10 = 0.639 | |
| I0623 15:19:19.801108 10365 solver.cpp:406] Test net output #109: loss3/accuracy11 = 0.574 | |
| I0623 15:19:19.801120 10365 solver.cpp:406] Test net output #110: loss3/accuracy12 = 0.633 | |
| I0623 15:19:19.801131 10365 solver.cpp:406] Test net output #111: loss3/accuracy13 = 0.678 | |
| I0623 15:19:19.801141 10365 solver.cpp:406] Test net output #112: loss3/accuracy14 = 0.74 | |
| I0623 15:19:19.801151 10365 solver.cpp:406] Test net output #113: loss3/accuracy15 = 0.824 | |
| I0623 15:19:19.801162 10365 solver.cpp:406] Test net output #114: loss3/accuracy16 = 0.874 | |
| I0623 15:19:19.801183 10365 solver.cpp:406] Test net output #115: loss3/accuracy17 = 0.921 | |
| I0623 15:19:19.801195 10365 solver.cpp:406] Test net output #116: loss3/accuracy18 = 0.966 | |
| I0623 15:19:19.801206 10365 solver.cpp:406] Test net output #117: loss3/accuracy19 = 0.976 | |
| I0623 15:19:19.801218 10365 solver.cpp:406] Test net output #118: loss3/accuracy20 = 0.987 | |
| I0623 15:19:19.801229 10365 solver.cpp:406] Test net output #119: loss3/accuracy21 = 0.999 | |
| I0623 15:19:19.801240 10365 solver.cpp:406] Test net output #120: loss3/accuracy22 = 1 | |
| I0623 15:19:19.801251 10365 solver.cpp:406] Test net output #121: loss3/accuracy_incl_empty = 0.909274 | |
| I0623 15:19:19.801264 10365 solver.cpp:406] Test net output #122: loss3/accuracy_top3 = 0.966978 | |
| I0623 15:19:19.801276 10365 solver.cpp:406] Test net output #123: loss3/cross_entropy_loss = 0.537938 (* 1 = 0.537938 loss) | |
| I0623 15:19:19.801290 10365 solver.cpp:406] Test net output #124: loss3/cross_entropy_loss_incl_empty = 0.351334 (* 1 = 0.351334 loss) | |
| I0623 15:19:19.801303 10365 solver.cpp:406] Test net output #125: loss3/loss01 = 0.155262 (* 0.0909091 = 0.0141148 loss) | |
| I0623 15:19:19.801319 10365 solver.cpp:406] Test net output #126: loss3/loss02 = 0.18181 (* 0.0909091 = 0.0165282 loss) | |
| I0623 15:19:19.801333 10365 solver.cpp:406] Test net output #127: loss3/loss03 = 0.31223 (* 0.0909091 = 0.0283846 loss) | |
| I0623 15:19:19.801347 10365 solver.cpp:406] Test net output #128: loss3/loss04 = 0.359497 (* 0.0909091 = 0.0326816 loss) | |
| I0623 15:19:19.801360 10365 solver.cpp:406] Test net output #129: loss3/loss05 = 0.380586 (* 0.0909091 = 0.0345987 loss) | |
| I0623 15:19:19.801373 10365 solver.cpp:406] Test net output #130: loss3/loss06 = 0.498533 (* 0.0909091 = 0.0453212 loss) | |
| I0623 15:19:19.801386 10365 solver.cpp:406] Test net output #131: loss3/loss07 = 0.547753 (* 0.0909091 = 0.0497957 loss) | |
| I0623 15:19:19.801400 10365 solver.cpp:406] Test net output #132: loss3/loss08 = 0.55748 (* 0.0909091 = 0.05068 loss) | |
| I0623 15:19:19.801414 10365 solver.cpp:406] Test net output #133: loss3/loss09 = 0.74361 (* 0.0909091 = 0.0676009 loss) | |
| I0623 15:19:19.801427 10365 solver.cpp:406] Test net output #134: loss3/loss10 = 1.01823 (* 0.0909091 = 0.0925663 loss) | |
| I0623 15:19:19.801440 10365 solver.cpp:406] Test net output #135: loss3/loss11 = 1.16624 (* 0.0909091 = 0.106022 loss) | |
| I0623 15:19:19.801453 10365 solver.cpp:406] Test net output #136: loss3/loss12 = 0.969957 (* 0.0909091 = 0.0881779 loss) | |
| I0623 15:19:19.801466 10365 solver.cpp:406] Test net output #137: loss3/loss13 = 0.880675 (* 0.0909091 = 0.0800614 loss) | |
| I0623 15:19:19.801479 10365 solver.cpp:406] Test net output #138: loss3/loss14 = 0.674098 (* 0.0909091 = 0.0612817 loss) | |
| I0623 15:19:19.801493 10365 solver.cpp:406] Test net output #139: loss3/loss15 = 0.479104 (* 0.0909091 = 0.0435549 loss) | |
| I0623 15:19:19.801506 10365 solver.cpp:406] Test net output #140: loss3/loss16 = 0.366366 (* 0.0909091 = 0.033306 loss) | |
| I0623 15:19:19.801519 10365 solver.cpp:406] Test net output #141: loss3/loss17 = 0.219897 (* 0.0909091 = 0.0199906 loss) | |
| I0623 15:19:19.801532 10365 solver.cpp:406] Test net output #142: loss3/loss18 = 0.130909 (* 0.0909091 = 0.0119008 loss) | |
| I0623 15:19:19.801545 10365 solver.cpp:406] Test net output #143: loss3/loss19 = 0.0752946 (* 0.0909091 = 0.00684496 loss) | |
| I0623 15:19:19.801559 10365 solver.cpp:406] Test net output #144: loss3/loss20 = 0.0434854 (* 0.0909091 = 0.00395322 loss) | |
| I0623 15:19:19.801573 10365 solver.cpp:406] Test net output #145: loss3/loss21 = 0.00593866 (* 0.0909091 = 0.000539878 loss) | |
| I0623 15:19:19.801585 10365 solver.cpp:406] Test net output #146: loss3/loss22 = 9.62999e-05 (* 0.0909091 = 8.75453e-06 loss) | |
| I0623 15:19:19.801597 10365 solver.cpp:406] Test net output #147: total_accuracy = 0.385 | |
| I0623 15:19:19.801609 10365 solver.cpp:406] Test net output #148: total_accuracy_not_rec = 0.2 | |
| I0623 15:19:19.801620 10365 solver.cpp:406] Test net output #149: total_confidence = 0.213189 | |
| I0623 15:19:19.801640 10365 solver.cpp:406] Test net output #150: total_confidence_not_rec = 0.14209 | |
| I0623 15:19:20.161268 10365 solver.cpp:229] Iteration 5000, loss = 4.7188 | |
| I0623 15:19:20.161355 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.417391 | |
| I0623 15:19:20.161373 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 15:19:20.161386 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.375 | |
| I0623 15:19:20.161401 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 15:19:20.161412 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.25 | |
| I0623 15:19:20.161424 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.625 | |
| I0623 15:19:20.161437 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 15:19:20.161448 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.25 | |
| I0623 15:19:20.161460 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 15:19:20.161473 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 15:19:20.161484 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.125 | |
| I0623 15:19:20.161495 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.125 | |
| I0623 15:19:20.161507 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 15:19:20.161519 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.75 | |
| I0623 15:19:20.161530 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 15:19:20.161541 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.5 | |
| I0623 15:19:20.161553 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.625 | |
| I0623 15:19:20.161566 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 15:19:20.161576 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 15:19:20.161588 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 0.875 | |
| I0623 15:19:20.161600 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:19:20.161612 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:19:20.161624 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:19:20.161635 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.613636 | |
| I0623 15:19:20.161648 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.713043 | |
| I0623 15:19:20.161664 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.69963 (* 0.3 = 0.509889 loss) | |
| I0623 15:19:20.161684 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.16432 (* 0.3 = 0.349296 loss) | |
| I0623 15:19:20.161700 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.333522 (* 0.0272727 = 0.00909606 loss) | |
| I0623 15:19:20.161715 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.92532 (* 0.0272727 = 0.0525086 loss) | |
| I0623 15:19:20.161727 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.49203 (* 0.0272727 = 0.0679644 loss) | |
| I0623 15:19:20.161741 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.56014 (* 0.0272727 = 0.0698221 loss) | |
| I0623 15:19:20.161756 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.89854 (* 0.0272727 = 0.0517783 loss) | |
| I0623 15:19:20.161769 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.01699 (* 0.0272727 = 0.0550088 loss) | |
| I0623 15:19:20.161783 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.04414 (* 0.0272727 = 0.0557494 loss) | |
| I0623 15:19:20.161798 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.41501 (* 0.0272727 = 0.0658638 loss) | |
| I0623 15:19:20.161811 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.8097 (* 0.0272727 = 0.0493555 loss) | |
| I0623 15:19:20.161825 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.15007 (* 0.0272727 = 0.0586384 loss) | |
| I0623 15:19:20.161839 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.60141 (* 0.0272727 = 0.0709474 loss) | |
| I0623 15:19:20.161888 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.81612 (* 0.0272727 = 0.0495306 loss) | |
| I0623 15:19:20.161903 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.40165 (* 0.0272727 = 0.0382268 loss) | |
| I0623 15:19:20.161917 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.50374 (* 0.0272727 = 0.0410111 loss) | |
| I0623 15:19:20.161931 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.21501 (* 0.0272727 = 0.0331366 loss) | |
| I0623 15:19:20.161944 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.25864 (* 0.0272727 = 0.0343265 loss) | |
| I0623 15:19:20.161957 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.498472 (* 0.0272727 = 0.0135947 loss) | |
| I0623 15:19:20.161972 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.30975 (* 0.0272727 = 0.00844773 loss) | |
| I0623 15:19:20.161985 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.495 (* 0.0272727 = 0.0135 loss) | |
| I0623 15:19:20.162000 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.0340772 (* 0.0272727 = 0.000929379 loss) | |
| I0623 15:19:20.162014 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.00990222 (* 0.0272727 = 0.00027006 loss) | |
| I0623 15:19:20.162029 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.00381439 (* 0.0272727 = 0.000104029 loss) | |
| I0623 15:19:20.162041 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.504348 | |
| I0623 15:19:20.162053 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 15:19:20.162065 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.625 | |
| I0623 15:19:20.162076 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 15:19:20.162088 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.375 | |
| I0623 15:19:20.162099 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 15:19:20.162114 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 15:19:20.162127 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0 | |
| I0623 15:19:20.162138 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 15:19:20.162150 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 15:19:20.162161 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 15:19:20.162173 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.125 | |
| I0623 15:19:20.162184 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 15:19:20.162196 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 15:19:20.162209 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.5 | |
| I0623 15:19:20.162220 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.375 | |
| I0623 15:19:20.162231 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.625 | |
| I0623 15:19:20.162243 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 15:19:20.162256 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 15:19:20.162267 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 0.875 | |
| I0623 15:19:20.162279 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:19:20.162292 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:19:20.162302 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:19:20.162314 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.664773 | |
| I0623 15:19:20.162327 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.8 | |
| I0623 15:19:20.162340 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.44305 (* 0.3 = 0.432914 loss) | |
| I0623 15:19:20.162354 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.983509 (* 0.3 = 0.295053 loss) | |
| I0623 15:19:20.162379 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.337775 (* 0.0272727 = 0.00921204 loss) | |
| I0623 15:19:20.162395 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.874984 (* 0.0272727 = 0.0238632 loss) | |
| I0623 15:19:20.162410 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.03524 (* 0.0272727 = 0.0282338 loss) | |
| I0623 15:19:20.162423 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.93825 (* 0.0272727 = 0.0528613 loss) | |
| I0623 15:19:20.162437 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.64474 (* 0.0272727 = 0.0448565 loss) | |
| I0623 15:19:20.162451 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.36846 (* 0.0272727 = 0.0373217 loss) | |
| I0623 15:19:20.162464 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.89831 (* 0.0272727 = 0.0517721 loss) | |
| I0623 15:19:20.162478 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.76296 (* 0.0272727 = 0.0480807 loss) | |
| I0623 15:19:20.162492 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.60928 (* 0.0272727 = 0.0438894 loss) | |
| I0623 15:19:20.162506 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.14028 (* 0.0272727 = 0.0583712 loss) | |
| I0623 15:19:20.162519 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.25287 (* 0.0272727 = 0.0614418 loss) | |
| I0623 15:19:20.162533 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.34344 (* 0.0272727 = 0.0366394 loss) | |
| I0623 15:19:20.162547 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.60849 (* 0.0272727 = 0.0438678 loss) | |
| I0623 15:19:20.162557 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.58687 (* 0.0272727 = 0.0432782 loss) | |
| I0623 15:19:20.162566 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.48817 (* 0.0272727 = 0.0405866 loss) | |
| I0623 15:19:20.162581 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 1.08494 (* 0.0272727 = 0.0295894 loss) | |
| I0623 15:19:20.162595 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.363457 (* 0.0272727 = 0.00991245 loss) | |
| I0623 15:19:20.162611 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.544343 (* 0.0272727 = 0.0148457 loss) | |
| I0623 15:19:20.162624 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.581936 (* 0.0272727 = 0.015871 loss) | |
| I0623 15:19:20.162638 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00933229 (* 0.0272727 = 0.000254517 loss) | |
| I0623 15:19:20.162652 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.0101916 (* 0.0272727 = 0.000277952 loss) | |
| I0623 15:19:20.162667 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.00301476 (* 0.0272727 = 8.22207e-05 loss) | |
| I0623 15:19:20.162678 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.808696 | |
| I0623 15:19:20.162690 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 15:19:20.162703 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 15:19:20.162714 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 15:19:20.162729 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 15:19:20.162740 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 15:19:20.162752 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.75 | |
| I0623 15:19:20.162765 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 15:19:20.162775 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 15:19:20.162787 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 15:19:20.162798 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.25 | |
| I0623 15:19:20.162811 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.375 | |
| I0623 15:19:20.162822 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.375 | |
| I0623 15:19:20.162833 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.5 | |
| I0623 15:19:20.162844 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.375 | |
| I0623 15:19:20.162866 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 15:19:20.162879 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 15:19:20.162891 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 15:19:20.162902 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.875 | |
| I0623 15:19:20.162914 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 0.875 | |
| I0623 15:19:20.162926 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:19:20.162938 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:19:20.162950 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:19:20.162961 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.869318 | |
| I0623 15:19:20.162972 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.973913 | |
| I0623 15:19:20.162986 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.608941 (* 1 = 0.608941 loss) | |
| I0623 15:19:20.163000 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.417189 (* 1 = 0.417189 loss) | |
| I0623 15:19:20.163014 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0335909 (* 0.0909091 = 0.00305372 loss) | |
| I0623 15:19:20.163029 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.327839 (* 0.0909091 = 0.0298035 loss) | |
| I0623 15:19:20.163043 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.098568 (* 0.0909091 = 0.00896073 loss) | |
| I0623 15:19:20.163058 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.171783 (* 0.0909091 = 0.0156167 loss) | |
| I0623 15:19:20.163072 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.202277 (* 0.0909091 = 0.0183888 loss) | |
| I0623 15:19:20.163086 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 1.02522 (* 0.0909091 = 0.0932015 loss) | |
| I0623 15:19:20.163100 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.751781 (* 0.0909091 = 0.0683437 loss) | |
| I0623 15:19:20.163113 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.60902 (* 0.0909091 = 0.0553654 loss) | |
| I0623 15:19:20.163127 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.81373 (* 0.0909091 = 0.0739754 loss) | |
| I0623 15:19:20.163141 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.33284 (* 0.0909091 = 0.121167 loss) | |
| I0623 15:19:20.163154 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.41742 (* 0.0909091 = 0.128857 loss) | |
| I0623 15:19:20.163172 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.43507 (* 0.0909091 = 0.130461 loss) | |
| I0623 15:19:20.163185 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.30581 (* 0.0909091 = 0.11871 loss) | |
| I0623 15:19:20.163199 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.15319 (* 0.0909091 = 0.104836 loss) | |
| I0623 15:19:20.163213 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.839881 (* 0.0909091 = 0.0763528 loss) | |
| I0623 15:19:20.163226 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.665887 (* 0.0909091 = 0.0605352 loss) | |
| I0623 15:19:20.163240 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.2318 (* 0.0909091 = 0.0210727 loss) | |
| I0623 15:19:20.163254 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.184697 (* 0.0909091 = 0.0167907 loss) | |
| I0623 15:19:20.163269 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.304962 (* 0.0909091 = 0.0277238 loss) | |
| I0623 15:19:20.163283 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.0105208 (* 0.0909091 = 0.00095644 loss) | |
| I0623 15:19:20.163297 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000712472 (* 0.0909091 = 6.47702e-05 loss) | |
| I0623 15:19:20.163311 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 8.0147e-05 (* 0.0909091 = 7.28609e-06 loss) | |
| I0623 15:19:20.163323 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 15:19:20.163346 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 15:19:20.163358 10365 solver.cpp:245] Train net output #149: total_confidence = 0.044765 | |
| I0623 15:19:20.163370 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.00965546 | |
| I0623 15:19:20.163383 10365 sgd_solver.cpp:106] Iteration 5000, lr = 0.001 | |
| I0623 15:21:12.172736 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.7262 > 30) by scale factor 0.863902 | |
| I0623 15:21:15.999522 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.2059 > 30) by scale factor 0.903454 | |
| I0623 15:23:49.033006 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.9108 > 30) by scale factor 0.911555 | |
| I0623 15:23:58.210072 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.3036 > 30) by scale factor 0.92869 | |
| I0623 15:25:42.724093 10365 solver.cpp:229] Iteration 5500, loss = 4.75394 | |
| I0623 15:25:42.724231 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.391753 | |
| I0623 15:25:42.724251 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 15:25:42.724267 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 15:25:42.724279 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 15:25:42.724290 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 15:25:42.724303 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.625 | |
| I0623 15:25:42.724314 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.125 | |
| I0623 15:25:42.724326 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 15:25:42.724339 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 15:25:42.724350 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 15:25:42.724361 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 15:25:42.724373 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 15:25:42.724385 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.5 | |
| I0623 15:25:42.724397 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 15:25:42.724408 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 15:25:42.724421 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 15:25:42.724432 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 15:25:42.724444 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.75 | |
| I0623 15:25:42.724455 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 15:25:42.724467 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 15:25:42.724479 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:25:42.724490 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:25:42.724503 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:25:42.724514 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.659091 | |
| I0623 15:25:42.724525 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.721649 | |
| I0623 15:25:42.724541 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.74601 (* 0.3 = 0.523803 loss) | |
| I0623 15:25:42.724556 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.993634 (* 0.3 = 0.29809 loss) | |
| I0623 15:25:42.724570 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 1.19863 (* 0.0272727 = 0.0326899 loss) | |
| I0623 15:25:42.724584 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 2.85807 (* 0.0272727 = 0.0779475 loss) | |
| I0623 15:25:42.724598 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.6554 (* 0.0272727 = 0.0451472 loss) | |
| I0623 15:25:42.724612 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.55574 (* 0.0272727 = 0.0697019 loss) | |
| I0623 15:25:42.724625 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.33869 (* 0.0272727 = 0.0365098 loss) | |
| I0623 15:25:42.724639 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.61065 (* 0.0272727 = 0.0711996 loss) | |
| I0623 15:25:42.724653 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.61111 (* 0.0272727 = 0.0439393 loss) | |
| I0623 15:25:42.724668 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.53333 (* 0.0272727 = 0.041818 loss) | |
| I0623 15:25:42.724680 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.63054 (* 0.0272727 = 0.0444693 loss) | |
| I0623 15:25:42.724694 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.31601 (* 0.0272727 = 0.0358913 loss) | |
| I0623 15:25:42.724707 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.48303 (* 0.0272727 = 0.0404461 loss) | |
| I0623 15:25:42.724721 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.31371 (* 0.0272727 = 0.0358283 loss) | |
| I0623 15:25:42.724753 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.91021 (* 0.0272727 = 0.0248239 loss) | |
| I0623 15:25:42.724768 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.997999 (* 0.0272727 = 0.0272182 loss) | |
| I0623 15:25:42.724782 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.764995 (* 0.0272727 = 0.0208635 loss) | |
| I0623 15:25:42.724795 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.905079 (* 0.0272727 = 0.024684 loss) | |
| I0623 15:25:42.724808 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 1.32977 (* 0.0272727 = 0.0362664 loss) | |
| I0623 15:25:42.724822 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.626482 (* 0.0272727 = 0.0170859 loss) | |
| I0623 15:25:42.724836 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00485043 (* 0.0272727 = 0.000132284 loss) | |
| I0623 15:25:42.724850 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00105324 (* 0.0272727 = 2.87248e-05 loss) | |
| I0623 15:25:42.724864 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 5.25747e-05 (* 0.0272727 = 1.43385e-06 loss) | |
| I0623 15:25:42.724879 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 6.21394e-06 (* 0.0272727 = 1.69471e-07 loss) | |
| I0623 15:25:42.724890 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.494845 | |
| I0623 15:25:42.724902 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.625 | |
| I0623 15:25:42.724915 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 15:25:42.724925 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 15:25:42.724937 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 15:25:42.724948 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 15:25:42.724959 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.125 | |
| I0623 15:25:42.724972 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 15:25:42.724982 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 15:25:42.724993 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 15:25:42.725005 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 15:25:42.725016 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 15:25:42.725028 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 15:25:42.725040 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 15:25:42.725051 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 15:25:42.725062 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 15:25:42.725075 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 15:25:42.725085 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.75 | |
| I0623 15:25:42.725097 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 15:25:42.725108 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 15:25:42.725119 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:25:42.725131 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:25:42.725142 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:25:42.725153 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.715909 | |
| I0623 15:25:42.725164 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.845361 | |
| I0623 15:25:42.725178 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.41664 (* 0.3 = 0.424992 loss) | |
| I0623 15:25:42.725191 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.817349 (* 0.3 = 0.245205 loss) | |
| I0623 15:25:42.725205 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.7052 (* 0.0272727 = 0.0192327 loss) | |
| I0623 15:25:42.725219 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.918984 (* 0.0272727 = 0.0250632 loss) | |
| I0623 15:25:42.725246 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.08882 (* 0.0272727 = 0.0296952 loss) | |
| I0623 15:25:42.725262 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.09458 (* 0.0272727 = 0.0298522 loss) | |
| I0623 15:25:42.725276 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.52452 (* 0.0272727 = 0.0415779 loss) | |
| I0623 15:25:42.725289 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.33411 (* 0.0272727 = 0.0636576 loss) | |
| I0623 15:25:42.725303 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.68542 (* 0.0272727 = 0.0459661 loss) | |
| I0623 15:25:42.725319 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.51067 (* 0.0272727 = 0.0412002 loss) | |
| I0623 15:25:42.725333 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.55532 (* 0.0272727 = 0.0424179 loss) | |
| I0623 15:25:42.725347 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.65683 (* 0.0272727 = 0.0451863 loss) | |
| I0623 15:25:42.725360 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.21799 (* 0.0272727 = 0.033218 loss) | |
| I0623 15:25:42.725373 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.31052 (* 0.0272727 = 0.0357415 loss) | |
| I0623 15:25:42.725388 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.930739 (* 0.0272727 = 0.0253838 loss) | |
| I0623 15:25:42.725401 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.00695 (* 0.0272727 = 0.0274622 loss) | |
| I0623 15:25:42.725414 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.761906 (* 0.0272727 = 0.0207792 loss) | |
| I0623 15:25:42.725428 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.550763 (* 0.0272727 = 0.0150208 loss) | |
| I0623 15:25:42.725442 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.931605 (* 0.0272727 = 0.0254074 loss) | |
| I0623 15:25:42.725456 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.59707 (* 0.0272727 = 0.0162837 loss) | |
| I0623 15:25:42.725471 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0175272 (* 0.0272727 = 0.000478015 loss) | |
| I0623 15:25:42.725483 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.0059995 (* 0.0272727 = 0.000163623 loss) | |
| I0623 15:25:42.725497 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00450868 (* 0.0272727 = 0.000122964 loss) | |
| I0623 15:25:42.725512 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.00486746 (* 0.0272727 = 0.000132749 loss) | |
| I0623 15:25:42.725524 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.804124 | |
| I0623 15:25:42.725536 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.875 | |
| I0623 15:25:42.725548 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 15:25:42.725560 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 15:25:42.725571 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 15:25:42.725584 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 15:25:42.725594 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 15:25:42.725605 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 15:25:42.725617 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.625 | |
| I0623 15:25:42.725628 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 15:25:42.725641 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 15:25:42.725651 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 15:25:42.725663 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.75 | |
| I0623 15:25:42.725674 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.875 | |
| I0623 15:25:42.725689 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 15:25:42.725697 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 15:25:42.725709 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 15:25:42.725731 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.75 | |
| I0623 15:25:42.725744 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.875 | |
| I0623 15:25:42.725755 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 15:25:42.725767 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:25:42.725778 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:25:42.725790 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:25:42.725800 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.886364 | |
| I0623 15:25:42.725811 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.958763 | |
| I0623 15:25:42.725826 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.697048 (* 1 = 0.697048 loss) | |
| I0623 15:25:42.725838 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.408282 (* 1 = 0.408282 loss) | |
| I0623 15:25:42.725852 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.432122 (* 0.0909091 = 0.0392838 loss) | |
| I0623 15:25:42.725865 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.617548 (* 0.0909091 = 0.0561408 loss) | |
| I0623 15:25:42.725879 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0431997 (* 0.0909091 = 0.00392724 loss) | |
| I0623 15:25:42.725893 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0898517 (* 0.0909091 = 0.00816834 loss) | |
| I0623 15:25:42.725908 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.165389 (* 0.0909091 = 0.0150354 loss) | |
| I0623 15:25:42.725920 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.202094 (* 0.0909091 = 0.0183722 loss) | |
| I0623 15:25:42.725934 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.0942208 (* 0.0909091 = 0.00856552 loss) | |
| I0623 15:25:42.725947 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.03226 (* 0.0909091 = 0.0938422 loss) | |
| I0623 15:25:42.725960 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.917506 (* 0.0909091 = 0.0834096 loss) | |
| I0623 15:25:42.725975 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.892831 (* 0.0909091 = 0.0811665 loss) | |
| I0623 15:25:42.725987 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.885709 (* 0.0909091 = 0.080519 loss) | |
| I0623 15:25:42.726001 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.740749 (* 0.0909091 = 0.0673408 loss) | |
| I0623 15:25:42.726013 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.479557 (* 0.0909091 = 0.0435961 loss) | |
| I0623 15:25:42.726027 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.548719 (* 0.0909091 = 0.0498835 loss) | |
| I0623 15:25:42.726040 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.798879 (* 0.0909091 = 0.0726253 loss) | |
| I0623 15:25:42.726053 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.444022 (* 0.0909091 = 0.0403656 loss) | |
| I0623 15:25:42.726066 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.679818 (* 0.0909091 = 0.0618016 loss) | |
| I0623 15:25:42.726079 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.385992 (* 0.0909091 = 0.0350902 loss) | |
| I0623 15:25:42.726094 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0057015 (* 0.0909091 = 0.000518318 loss) | |
| I0623 15:25:42.726106 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000726757 (* 0.0909091 = 6.60688e-05 loss) | |
| I0623 15:25:42.726120 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 9.58675e-05 (* 0.0909091 = 8.71523e-06 loss) | |
| I0623 15:25:42.726135 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.02335e-06 (* 0.0909091 = 3.65759e-07 loss) | |
| I0623 15:25:42.726146 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.5 | |
| I0623 15:25:42.726157 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 15:25:42.726178 10365 solver.cpp:245] Train net output #149: total_confidence = 0.201012 | |
| I0623 15:25:42.726191 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.141955 | |
| I0623 15:25:42.726204 10365 sgd_solver.cpp:106] Iteration 5500, lr = 0.001 | |
| I0623 15:26:17.526067 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 45.5883 > 30) by scale factor 0.658063 | |
| I0623 15:29:06.677770 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.0608 > 30) by scale factor 0.935722 | |
| I0623 15:30:02.555660 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.3234 > 30) by scale factor 0.989334 | |
| I0623 15:32:05.430155 10365 solver.cpp:229] Iteration 6000, loss = 4.79713 | |
| I0623 15:32:05.430241 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.428571 | |
| I0623 15:32:05.430260 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 15:32:05.430274 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 15:32:05.430285 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 15:32:05.430297 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 15:32:05.430310 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 15:32:05.430321 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 15:32:05.430333 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 15:32:05.430344 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.75 | |
| I0623 15:32:05.430356 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 15:32:05.430368 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 15:32:05.430379 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 15:32:05.430392 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 15:32:05.430402 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.25 | |
| I0623 15:32:05.430418 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.375 | |
| I0623 15:32:05.430431 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 15:32:05.430443 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 15:32:05.430454 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.75 | |
| I0623 15:32:05.430465 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 15:32:05.430477 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 15:32:05.430488 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:32:05.430500 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:32:05.430511 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:32:05.430523 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.636364 | |
| I0623 15:32:05.430536 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.758929 | |
| I0623 15:32:05.430552 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.71509 (* 0.3 = 0.514527 loss) | |
| I0623 15:32:05.430567 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.10384 (* 0.3 = 0.331152 loss) | |
| I0623 15:32:05.430582 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.703296 (* 0.0272727 = 0.0191808 loss) | |
| I0623 15:32:05.430594 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.11372 (* 0.0272727 = 0.0303742 loss) | |
| I0623 15:32:05.430608 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.50726 (* 0.0272727 = 0.0411071 loss) | |
| I0623 15:32:05.430622 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.96125 (* 0.0272727 = 0.0534887 loss) | |
| I0623 15:32:05.430635 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.82822 (* 0.0272727 = 0.0498604 loss) | |
| I0623 15:32:05.430649 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.40689 (* 0.0272727 = 0.0656424 loss) | |
| I0623 15:32:05.430663 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.65458 (* 0.0272727 = 0.0451248 loss) | |
| I0623 15:32:05.430676 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.06152 (* 0.0272727 = 0.0289506 loss) | |
| I0623 15:32:05.430690 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.57807 (* 0.0272727 = 0.0430383 loss) | |
| I0623 15:32:05.430703 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.44656 (* 0.0272727 = 0.0667244 loss) | |
| I0623 15:32:05.430717 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.92636 (* 0.0272727 = 0.0525371 loss) | |
| I0623 15:32:05.430730 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.14511 (* 0.0272727 = 0.058503 loss) | |
| I0623 15:32:05.430762 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.92399 (* 0.0272727 = 0.0524725 loss) | |
| I0623 15:32:05.430776 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.79701 (* 0.0272727 = 0.0490094 loss) | |
| I0623 15:32:05.430789 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.62368 (* 0.0272727 = 0.0442823 loss) | |
| I0623 15:32:05.430804 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.74415 (* 0.0272727 = 0.0475679 loss) | |
| I0623 15:32:05.430816 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.796531 (* 0.0272727 = 0.0217236 loss) | |
| I0623 15:32:05.430830 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.418922 (* 0.0272727 = 0.0114251 loss) | |
| I0623 15:32:05.430845 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0154965 (* 0.0272727 = 0.000422631 loss) | |
| I0623 15:32:05.430857 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.0060456 (* 0.0272727 = 0.00016488 loss) | |
| I0623 15:32:05.430871 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.00124953 (* 0.0272727 = 3.4078e-05 loss) | |
| I0623 15:32:05.430886 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000187931 (* 0.0272727 = 5.12539e-06 loss) | |
| I0623 15:32:05.430897 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.544643 | |
| I0623 15:32:05.430909 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 15:32:05.430922 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 15:32:05.430932 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 15:32:05.430943 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.125 | |
| I0623 15:32:05.430954 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 15:32:05.430966 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.25 | |
| I0623 15:32:05.430977 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 15:32:05.430989 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 15:32:05.431000 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 15:32:05.431011 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 15:32:05.431022 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 15:32:05.431033 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.25 | |
| I0623 15:32:05.431044 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 15:32:05.431056 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.375 | |
| I0623 15:32:05.431066 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.5 | |
| I0623 15:32:05.431078 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.625 | |
| I0623 15:32:05.431089 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.75 | |
| I0623 15:32:05.431100 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 15:32:05.431113 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 15:32:05.431126 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:32:05.431138 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:32:05.431149 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:32:05.431161 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.704545 | |
| I0623 15:32:05.431172 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.830357 | |
| I0623 15:32:05.431186 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.31315 (* 0.3 = 0.393944 loss) | |
| I0623 15:32:05.431200 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.865611 (* 0.3 = 0.259683 loss) | |
| I0623 15:32:05.431215 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.0621485 (* 0.0272727 = 0.00169496 loss) | |
| I0623 15:32:05.431228 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.258923 (* 0.0272727 = 0.00706154 loss) | |
| I0623 15:32:05.431253 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.469636 (* 0.0272727 = 0.0128083 loss) | |
| I0623 15:32:05.431264 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.42965 (* 0.0272727 = 0.0389906 loss) | |
| I0623 15:32:05.431280 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.17808 (* 0.0272727 = 0.0321295 loss) | |
| I0623 15:32:05.431294 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.16226 (* 0.0272727 = 0.0589706 loss) | |
| I0623 15:32:05.431308 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.28974 (* 0.0272727 = 0.0351748 loss) | |
| I0623 15:32:05.431321 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 0.995711 (* 0.0272727 = 0.0271558 loss) | |
| I0623 15:32:05.431335 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 2.06 (* 0.0272727 = 0.0561818 loss) | |
| I0623 15:32:05.431349 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.72041 (* 0.0272727 = 0.0469203 loss) | |
| I0623 15:32:05.431362 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.90653 (* 0.0272727 = 0.0519963 loss) | |
| I0623 15:32:05.431375 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.90007 (* 0.0272727 = 0.05182 loss) | |
| I0623 15:32:05.431390 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.5692 (* 0.0272727 = 0.0427965 loss) | |
| I0623 15:32:05.431402 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.57978 (* 0.0272727 = 0.0430849 loss) | |
| I0623 15:32:05.431416 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.44494 (* 0.0272727 = 0.0394075 loss) | |
| I0623 15:32:05.431429 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 1.45359 (* 0.0272727 = 0.0396433 loss) | |
| I0623 15:32:05.431443 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 1.13185 (* 0.0272727 = 0.0308687 loss) | |
| I0623 15:32:05.431459 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.428844 (* 0.0272727 = 0.0116957 loss) | |
| I0623 15:32:05.431474 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.110426 (* 0.0272727 = 0.00301162 loss) | |
| I0623 15:32:05.431488 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.0208824 (* 0.0272727 = 0.000569521 loss) | |
| I0623 15:32:05.431502 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000454058 (* 0.0272727 = 1.23834e-05 loss) | |
| I0623 15:32:05.431516 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000301367 (* 0.0272727 = 8.21911e-06 loss) | |
| I0623 15:32:05.431529 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.785714 | |
| I0623 15:32:05.431540 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 15:32:05.431552 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 15:32:05.431563 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 15:32:05.431574 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 15:32:05.431586 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 15:32:05.431608 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 15:32:05.431623 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 15:32:05.431634 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 15:32:05.431645 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 15:32:05.431656 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 15:32:05.431668 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 15:32:05.431679 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 15:32:05.431689 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.375 | |
| I0623 15:32:05.431700 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 15:32:05.431711 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 15:32:05.431722 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 15:32:05.431746 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.75 | |
| I0623 15:32:05.431758 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 15:32:05.431769 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 15:32:05.431782 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:32:05.431792 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:32:05.431804 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:32:05.431815 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.857955 | |
| I0623 15:32:05.431828 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.928571 | |
| I0623 15:32:05.431841 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.845292 (* 1 = 0.845292 loss) | |
| I0623 15:32:05.431854 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.550699 (* 1 = 0.550699 loss) | |
| I0623 15:32:05.431869 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.065958 (* 0.0909091 = 0.00599619 loss) | |
| I0623 15:32:05.431884 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.091803 (* 0.0909091 = 0.00834573 loss) | |
| I0623 15:32:05.431897 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.116386 (* 0.0909091 = 0.0105805 loss) | |
| I0623 15:32:05.431911 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.954307 (* 0.0909091 = 0.0867552 loss) | |
| I0623 15:32:05.431924 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 1.33051 (* 0.0909091 = 0.120955 loss) | |
| I0623 15:32:05.431938 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.567433 (* 0.0909091 = 0.0515848 loss) | |
| I0623 15:32:05.431951 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.757357 (* 0.0909091 = 0.0688507 loss) | |
| I0623 15:32:05.431965 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.905916 (* 0.0909091 = 0.082356 loss) | |
| I0623 15:32:05.431978 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.70195 (* 0.0909091 = 0.0638136 loss) | |
| I0623 15:32:05.431993 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.30862 (* 0.0909091 = 0.118965 loss) | |
| I0623 15:32:05.432005 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.729234 (* 0.0909091 = 0.066294 loss) | |
| I0623 15:32:05.432019 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.49972 (* 0.0909091 = 0.136338 loss) | |
| I0623 15:32:05.432032 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.28786 (* 0.0909091 = 0.117078 loss) | |
| I0623 15:32:05.432045 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.974184 (* 0.0909091 = 0.0885622 loss) | |
| I0623 15:32:05.432060 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.1472 (* 0.0909091 = 0.104291 loss) | |
| I0623 15:32:05.432072 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 1.11932 (* 0.0909091 = 0.101756 loss) | |
| I0623 15:32:05.432085 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.644907 (* 0.0909091 = 0.0586279 loss) | |
| I0623 15:32:05.432099 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.119513 (* 0.0909091 = 0.0108648 loss) | |
| I0623 15:32:05.432112 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0298295 (* 0.0909091 = 0.00271177 loss) | |
| I0623 15:32:05.432126 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00511299 (* 0.0909091 = 0.000464817 loss) | |
| I0623 15:32:05.432140 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000878784 (* 0.0909091 = 7.98895e-05 loss) | |
| I0623 15:32:05.432154 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 0.000179678 (* 0.0909091 = 1.63344e-05 loss) | |
| I0623 15:32:05.432166 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 15:32:05.432180 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 15:32:05.432193 10365 solver.cpp:245] Train net output #149: total_confidence = 0.129099 | |
| I0623 15:32:05.432214 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.122155 | |
| I0623 15:32:05.432229 10365 sgd_solver.cpp:106] Iteration 6000, lr = 0.001 | |
| I0623 15:35:28.628430 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.0452 > 30) by scale factor 0.936178 | |
| I0623 15:36:37.461820 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.0939 > 30) by scale factor 0.99688 | |
| I0623 15:37:35.604140 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.0115 > 30) by scale factor 0.749784 | |
| I0623 15:38:10.033852 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.0766 > 30) by scale factor 0.935263 | |
| I0623 15:38:22.287458 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 74.6614 > 30) by scale factor 0.401814 | |
| I0623 15:38:27.642467 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.6114 > 30) by scale factor 0.980027 | |
| I0623 15:38:28.045498 10365 solver.cpp:229] Iteration 6500, loss = 4.72525 | |
| I0623 15:38:28.045560 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.476744 | |
| I0623 15:38:28.045578 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.5 | |
| I0623 15:38:28.045591 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 15:38:28.045603 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 15:38:28.045615 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 15:38:28.045627 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 15:38:28.045639 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.75 | |
| I0623 15:38:28.045650 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 15:38:28.045662 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 15:38:28.045675 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 15:38:28.045686 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 15:38:28.045698 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 15:38:28.045709 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 15:38:28.045722 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.875 | |
| I0623 15:38:28.045733 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 15:38:28.045745 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 15:38:28.045756 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 15:38:28.045775 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 15:38:28.045789 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 15:38:28.045800 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 15:38:28.045812 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:38:28.045824 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:38:28.045835 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:38:28.045846 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.721591 | |
| I0623 15:38:28.045858 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.790698 | |
| I0623 15:38:28.045874 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.70567 (* 0.3 = 0.511702 loss) | |
| I0623 15:38:28.045888 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.91993 (* 0.3 = 0.275979 loss) | |
| I0623 15:38:28.045903 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.982489 (* 0.0272727 = 0.0267951 loss) | |
| I0623 15:38:28.045917 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.917617 (* 0.0272727 = 0.0250259 loss) | |
| I0623 15:38:28.045930 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.64848 (* 0.0272727 = 0.0449585 loss) | |
| I0623 15:38:28.045943 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.94969 (* 0.0272727 = 0.0531734 loss) | |
| I0623 15:38:28.045958 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.68912 (* 0.0272727 = 0.0460669 loss) | |
| I0623 15:38:28.045971 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 0.956929 (* 0.0272727 = 0.0260981 loss) | |
| I0623 15:38:28.045985 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.23358 (* 0.0272727 = 0.0336431 loss) | |
| I0623 15:38:28.045999 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.0695 (* 0.0272727 = 0.0291682 loss) | |
| I0623 15:38:28.046012 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.98409 (* 0.0272727 = 0.0541115 loss) | |
| I0623 15:38:28.046025 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.51929 (* 0.0272727 = 0.068708 loss) | |
| I0623 15:38:28.046069 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.2397 (* 0.0272727 = 0.0338101 loss) | |
| I0623 15:38:28.046088 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.02618 (* 0.0272727 = 0.0552594 loss) | |
| I0623 15:38:28.046103 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.689284 (* 0.0272727 = 0.0187986 loss) | |
| I0623 15:38:28.046118 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.761297 (* 0.0272727 = 0.0207627 loss) | |
| I0623 15:38:28.046131 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.847566 (* 0.0272727 = 0.0231154 loss) | |
| I0623 15:38:28.046150 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.567552 (* 0.0272727 = 0.0154787 loss) | |
| I0623 15:38:28.046165 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.397152 (* 0.0272727 = 0.0108314 loss) | |
| I0623 15:38:28.046178 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.611525 (* 0.0272727 = 0.016678 loss) | |
| I0623 15:38:28.046193 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0095666 (* 0.0272727 = 0.000260907 loss) | |
| I0623 15:38:28.046207 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00132539 (* 0.0272727 = 3.61471e-05 loss) | |
| I0623 15:38:28.046221 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000109502 (* 0.0272727 = 2.98643e-06 loss) | |
| I0623 15:38:28.046236 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 9.05853e-05 (* 0.0272727 = 2.47051e-06 loss) | |
| I0623 15:38:28.046247 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.546512 | |
| I0623 15:38:28.046260 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 15:38:28.046272 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 15:38:28.046283 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 15:38:28.046293 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.875 | |
| I0623 15:38:28.046305 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 15:38:28.046316 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.75 | |
| I0623 15:38:28.046327 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 15:38:28.046339 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.875 | |
| I0623 15:38:28.046350 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 15:38:28.046361 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 15:38:28.046372 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 15:38:28.046383 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 15:38:28.046394 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.875 | |
| I0623 15:38:28.046406 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 15:38:28.046421 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 15:38:28.046432 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 15:38:28.046443 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 15:38:28.046454 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 15:38:28.046466 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 15:38:28.046478 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:38:28.046489 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:38:28.046499 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:38:28.046511 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.75 | |
| I0623 15:38:28.046522 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.813953 | |
| I0623 15:38:28.046536 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.63633 (* 0.3 = 0.490899 loss) | |
| I0623 15:38:28.046561 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.905752 (* 0.3 = 0.271726 loss) | |
| I0623 15:38:28.046576 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.415692 (* 0.0272727 = 0.0113371 loss) | |
| I0623 15:38:28.046589 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.859269 (* 0.0272727 = 0.0234346 loss) | |
| I0623 15:38:28.046603 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.430459 (* 0.0272727 = 0.0117398 loss) | |
| I0623 15:38:28.046617 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.775417 (* 0.0272727 = 0.0211477 loss) | |
| I0623 15:38:28.046629 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 0.979529 (* 0.0272727 = 0.0267144 loss) | |
| I0623 15:38:28.046643 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 0.98069 (* 0.0272727 = 0.0267461 loss) | |
| I0623 15:38:28.046656 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.48037 (* 0.0272727 = 0.0403737 loss) | |
| I0623 15:38:28.046669 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.13236 (* 0.0272727 = 0.0308826 loss) | |
| I0623 15:38:28.046684 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.46582 (* 0.0272727 = 0.039977 loss) | |
| I0623 15:38:28.046697 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.7236 (* 0.0272727 = 0.0742801 loss) | |
| I0623 15:38:28.046710 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.05898 (* 0.0272727 = 0.0561539 loss) | |
| I0623 15:38:28.046723 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.78303 (* 0.0272727 = 0.048628 loss) | |
| I0623 15:38:28.046736 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.876809 (* 0.0272727 = 0.023913 loss) | |
| I0623 15:38:28.046751 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.840859 (* 0.0272727 = 0.0229325 loss) | |
| I0623 15:38:28.046763 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.68487 (* 0.0272727 = 0.0186783 loss) | |
| I0623 15:38:28.046777 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.666831 (* 0.0272727 = 0.0181863 loss) | |
| I0623 15:38:28.046792 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.237709 (* 0.0272727 = 0.00648297 loss) | |
| I0623 15:38:28.046804 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.245159 (* 0.0272727 = 0.00668616 loss) | |
| I0623 15:38:28.046818 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.11026 (* 0.0272727 = 0.00300709 loss) | |
| I0623 15:38:28.046833 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.0100985 (* 0.0272727 = 0.000275414 loss) | |
| I0623 15:38:28.046845 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000180117 (* 0.0272727 = 4.91229e-06 loss) | |
| I0623 15:38:28.046859 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000177411 (* 0.0272727 = 4.83848e-06 loss) | |
| I0623 15:38:28.046871 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.848837 | |
| I0623 15:38:28.046883 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 15:38:28.046895 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 15:38:28.046905 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 15:38:28.046917 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 15:38:28.046928 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 15:38:28.046939 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 15:38:28.046952 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 15:38:28.046962 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 15:38:28.046974 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 15:38:28.046985 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 15:38:28.046996 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.5 | |
| I0623 15:38:28.047008 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 15:38:28.047029 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 15:38:28.047042 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 15:38:28.047055 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 15:38:28.047065 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 15:38:28.047076 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 15:38:28.047088 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.875 | |
| I0623 15:38:28.047099 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 15:38:28.047111 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:38:28.047122 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:38:28.047137 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:38:28.047149 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.909091 | |
| I0623 15:38:28.047161 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.930233 | |
| I0623 15:38:28.047174 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.7538 (* 1 = 0.7538 loss) | |
| I0623 15:38:28.047188 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.45852 (* 1 = 0.45852 loss) | |
| I0623 15:38:28.047202 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0874193 (* 0.0909091 = 0.00794721 loss) | |
| I0623 15:38:28.047221 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0993055 (* 0.0909091 = 0.00902777 loss) | |
| I0623 15:38:28.047231 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.04321 (* 0.0909091 = 0.00392818 loss) | |
| I0623 15:38:28.047246 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.161609 (* 0.0909091 = 0.0146918 loss) | |
| I0623 15:38:28.047260 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.0896466 (* 0.0909091 = 0.00814969 loss) | |
| I0623 15:38:28.047273 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.0699376 (* 0.0909091 = 0.00635796 loss) | |
| I0623 15:38:28.047286 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.117124 (* 0.0909091 = 0.0106477 loss) | |
| I0623 15:38:28.047299 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.4086 (* 0.0909091 = 0.128054 loss) | |
| I0623 15:38:28.047312 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.24603 (* 0.0909091 = 0.113276 loss) | |
| I0623 15:38:28.047327 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.25002 (* 0.0909091 = 0.113638 loss) | |
| I0623 15:38:28.047339 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.86153 (* 0.0909091 = 0.16923 loss) | |
| I0623 15:38:28.047353 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.6854 (* 0.0909091 = 0.153218 loss) | |
| I0623 15:38:28.047365 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.00365 (* 0.0909091 = 0.0912407 loss) | |
| I0623 15:38:28.047379 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.03137 (* 0.0909091 = 0.0937607 loss) | |
| I0623 15:38:28.047391 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.690088 (* 0.0909091 = 0.0627353 loss) | |
| I0623 15:38:28.047405 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.261861 (* 0.0909091 = 0.0238056 loss) | |
| I0623 15:38:28.047417 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.144741 (* 0.0909091 = 0.0131582 loss) | |
| I0623 15:38:28.047430 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.209772 (* 0.0909091 = 0.0190702 loss) | |
| I0623 15:38:28.047444 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0526559 (* 0.0909091 = 0.0047869 loss) | |
| I0623 15:38:28.047457 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00464296 (* 0.0909091 = 0.000422087 loss) | |
| I0623 15:38:28.047474 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.00103849 (* 0.0909091 = 9.44086e-05 loss) | |
| I0623 15:38:28.047499 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 0.000162309 (* 0.0909091 = 1.47553e-05 loss) | |
| I0623 15:38:28.047513 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 15:38:28.047523 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 15:38:28.047534 10365 solver.cpp:245] Train net output #149: total_confidence = 0.293246 | |
| I0623 15:38:28.047546 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.277908 | |
| I0623 15:38:28.047559 10365 sgd_solver.cpp:106] Iteration 6500, lr = 0.001 | |
| I0623 15:39:11.242823 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.2046 > 30) by scale factor 0.993226 | |
| I0623 15:41:10.601804 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.3447 > 30) by scale factor 0.708472 | |
| I0623 15:44:50.546919 10365 solver.cpp:229] Iteration 7000, loss = 4.75963 | |
| I0623 15:44:50.547081 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.526316 | |
| I0623 15:44:50.547102 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 15:44:50.547116 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 15:44:50.547129 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 15:44:50.547142 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.75 | |
| I0623 15:44:50.547153 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 15:44:50.547165 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 15:44:50.547178 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.25 | |
| I0623 15:44:50.547189 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.25 | |
| I0623 15:44:50.547201 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 15:44:50.547212 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 15:44:50.547224 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 15:44:50.547236 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 15:44:50.547248 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.875 | |
| I0623 15:44:50.547262 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 1 | |
| I0623 15:44:50.547276 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 1 | |
| I0623 15:44:50.547287 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 1 | |
| I0623 15:44:50.547299 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 15:44:50.547312 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 15:44:50.547322 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 15:44:50.547334 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:44:50.547346 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:44:50.547358 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:44:50.547370 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.727273 | |
| I0623 15:44:50.547381 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.776316 | |
| I0623 15:44:50.547399 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.39913 (* 0.3 = 0.419738 loss) | |
| I0623 15:44:50.547413 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.93372 (* 0.3 = 0.280116 loss) | |
| I0623 15:44:50.547427 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.399251 (* 0.0272727 = 0.0108887 loss) | |
| I0623 15:44:50.547441 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.27266 (* 0.0272727 = 0.0347089 loss) | |
| I0623 15:44:50.547456 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.57527 (* 0.0272727 = 0.042962 loss) | |
| I0623 15:44:50.547469 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.38877 (* 0.0272727 = 0.0378756 loss) | |
| I0623 15:44:50.547482 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.68884 (* 0.0272727 = 0.0733319 loss) | |
| I0623 15:44:50.547497 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.13637 (* 0.0272727 = 0.0582646 loss) | |
| I0623 15:44:50.547510 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.24127 (* 0.0272727 = 0.0611256 loss) | |
| I0623 15:44:50.547523 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.08357 (* 0.0272727 = 0.0568247 loss) | |
| I0623 15:44:50.547538 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.97273 (* 0.0272727 = 0.0538018 loss) | |
| I0623 15:44:50.547551 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.17084 (* 0.0272727 = 0.0592048 loss) | |
| I0623 15:44:50.547565 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.60199 (* 0.0272727 = 0.0436908 loss) | |
| I0623 15:44:50.547579 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.00566 (* 0.0272727 = 0.0274271 loss) | |
| I0623 15:44:50.547592 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.441624 (* 0.0272727 = 0.0120443 loss) | |
| I0623 15:44:50.547643 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.247733 (* 0.0272727 = 0.00675635 loss) | |
| I0623 15:44:50.547659 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.12045 (* 0.0272727 = 0.00328499 loss) | |
| I0623 15:44:50.547673 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.0395577 (* 0.0272727 = 0.00107885 loss) | |
| I0623 15:44:50.547688 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00874344 (* 0.0272727 = 0.000238457 loss) | |
| I0623 15:44:50.547703 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00227453 (* 0.0272727 = 6.20326e-05 loss) | |
| I0623 15:44:50.547715 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.000765658 (* 0.0272727 = 2.08816e-05 loss) | |
| I0623 15:44:50.547729 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00105247 (* 0.0272727 = 2.87037e-05 loss) | |
| I0623 15:44:50.547744 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000452772 (* 0.0272727 = 1.23483e-05 loss) | |
| I0623 15:44:50.547757 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 1.70182e-05 (* 0.0272727 = 4.64133e-07 loss) | |
| I0623 15:44:50.547770 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.618421 | |
| I0623 15:44:50.547782 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 15:44:50.547794 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 15:44:50.547806 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.5 | |
| I0623 15:44:50.547817 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.875 | |
| I0623 15:44:50.547829 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 15:44:50.547842 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.25 | |
| I0623 15:44:50.547853 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.25 | |
| I0623 15:44:50.547864 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.5 | |
| I0623 15:44:50.547876 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 15:44:50.547888 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.25 | |
| I0623 15:44:50.547899 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 15:44:50.547912 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.75 | |
| I0623 15:44:50.547924 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.875 | |
| I0623 15:44:50.547935 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 1 | |
| I0623 15:44:50.547947 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 1 | |
| I0623 15:44:50.547960 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 15:44:50.547971 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 15:44:50.547981 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 15:44:50.547993 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 15:44:50.548005 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:44:50.548017 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:44:50.548028 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:44:50.548040 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.784091 | |
| I0623 15:44:50.548053 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.907895 | |
| I0623 15:44:50.548066 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.06263 (* 0.3 = 0.318788 loss) | |
| I0623 15:44:50.548080 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.759796 (* 0.3 = 0.227939 loss) | |
| I0623 15:44:50.548099 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.397989 (* 0.0272727 = 0.0108542 loss) | |
| I0623 15:44:50.548110 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.140483 (* 0.0272727 = 0.00383136 loss) | |
| I0623 15:44:50.548135 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.938287 (* 0.0272727 = 0.0255896 loss) | |
| I0623 15:44:50.548151 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.340482 (* 0.0272727 = 0.00928587 loss) | |
| I0623 15:44:50.548166 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.12352 (* 0.0272727 = 0.0579143 loss) | |
| I0623 15:44:50.548179 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.0913 (* 0.0272727 = 0.0570355 loss) | |
| I0623 15:44:50.548193 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 2.00061 (* 0.0272727 = 0.0545622 loss) | |
| I0623 15:44:50.548207 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.91423 (* 0.0272727 = 0.0522063 loss) | |
| I0623 15:44:50.548220 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.89495 (* 0.0272727 = 0.0516804 loss) | |
| I0623 15:44:50.548234 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.58408 (* 0.0272727 = 0.070475 loss) | |
| I0623 15:44:50.548249 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.04669 (* 0.0272727 = 0.0558188 loss) | |
| I0623 15:44:50.548261 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 0.70846 (* 0.0272727 = 0.0193216 loss) | |
| I0623 15:44:50.548275 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.330112 (* 0.0272727 = 0.00900306 loss) | |
| I0623 15:44:50.548290 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.103528 (* 0.0272727 = 0.0028235 loss) | |
| I0623 15:44:50.548305 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.0163748 (* 0.0272727 = 0.000446585 loss) | |
| I0623 15:44:50.548321 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.00242281 (* 0.0272727 = 6.60766e-05 loss) | |
| I0623 15:44:50.548336 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.000236806 (* 0.0272727 = 6.45834e-06 loss) | |
| I0623 15:44:50.548349 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 1.94917e-05 (* 0.0272727 = 5.31592e-07 loss) | |
| I0623 15:44:50.548363 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 4.18996e-05 (* 0.0272727 = 1.14272e-06 loss) | |
| I0623 15:44:50.548377 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 4.47041e-06 (* 0.0272727 = 1.2192e-07 loss) | |
| I0623 15:44:50.548391 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 2.65242e-06 (* 0.0272727 = 7.23388e-08 loss) | |
| I0623 15:44:50.548404 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 5.51344e-07 (* 0.0272727 = 1.50366e-08 loss) | |
| I0623 15:44:50.548418 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.921053 | |
| I0623 15:44:50.548429 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 15:44:50.548441 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 15:44:50.548452 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 15:44:50.548463 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 15:44:50.548475 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 15:44:50.548486 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 15:44:50.548498 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 15:44:50.548509 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 15:44:50.548521 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 15:44:50.548532 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 15:44:50.548544 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.5 | |
| I0623 15:44:50.548557 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.875 | |
| I0623 15:44:50.548568 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.875 | |
| I0623 15:44:50.548579 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 1 | |
| I0623 15:44:50.548591 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 1 | |
| I0623 15:44:50.548602 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 15:44:50.548624 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 15:44:50.548637 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 15:44:50.548650 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 15:44:50.548661 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:44:50.548672 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:44:50.548684 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:44:50.548696 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.914773 | |
| I0623 15:44:50.548708 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 1 | |
| I0623 15:44:50.548722 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.262818 (* 1 = 0.262818 loss) | |
| I0623 15:44:50.548737 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.331497 (* 1 = 0.331497 loss) | |
| I0623 15:44:50.548750 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.028078 (* 0.0909091 = 0.00255255 loss) | |
| I0623 15:44:50.548765 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.370313 (* 0.0909091 = 0.0336648 loss) | |
| I0623 15:44:50.548779 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.210992 (* 0.0909091 = 0.0191811 loss) | |
| I0623 15:44:50.548794 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.140233 (* 0.0909091 = 0.0127484 loss) | |
| I0623 15:44:50.548807 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.637109 (* 0.0909091 = 0.057919 loss) | |
| I0623 15:44:50.548821 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.577113 (* 0.0909091 = 0.0524648 loss) | |
| I0623 15:44:50.548835 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.757726 (* 0.0909091 = 0.0688842 loss) | |
| I0623 15:44:50.548848 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.772375 (* 0.0909091 = 0.0702159 loss) | |
| I0623 15:44:50.548862 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.20879 (* 0.0909091 = 0.10989 loss) | |
| I0623 15:44:50.548877 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.26506 (* 0.0909091 = 0.115005 loss) | |
| I0623 15:44:50.548890 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.27729 (* 0.0909091 = 0.116117 loss) | |
| I0623 15:44:50.548904 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.648553 (* 0.0909091 = 0.0589593 loss) | |
| I0623 15:44:50.548918 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.180039 (* 0.0909091 = 0.0163672 loss) | |
| I0623 15:44:50.548933 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.0124664 (* 0.0909091 = 0.00113331 loss) | |
| I0623 15:44:50.548946 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.000258479 (* 0.0909091 = 2.34981e-05 loss) | |
| I0623 15:44:50.548960 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 7.77852e-06 (* 0.0909091 = 7.07138e-07 loss) | |
| I0623 15:44:50.548975 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 3.35278e-06 (* 0.0909091 = 3.04798e-07 loss) | |
| I0623 15:44:50.548990 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 7.15256e-07 (* 0.0909091 = 6.50233e-08 loss) | |
| I0623 15:44:50.549003 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 1.78814e-07 (* 0.0909091 = 1.62558e-08 loss) | |
| I0623 15:44:50.549018 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 4.61937e-07 (* 0.0909091 = 4.19942e-08 loss) | |
| I0623 15:44:50.549032 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 3.12924e-07 (* 0.0909091 = 2.84477e-08 loss) | |
| I0623 15:44:50.549047 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 2.23517e-07 (* 0.0909091 = 2.03198e-08 loss) | |
| I0623 15:44:50.549059 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.5 | |
| I0623 15:44:50.549072 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 15:44:50.549093 10365 solver.cpp:245] Train net output #149: total_confidence = 0.221147 | |
| I0623 15:44:50.549105 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.158233 | |
| I0623 15:44:50.549119 10365 sgd_solver.cpp:106] Iteration 7000, lr = 0.001 | |
| I0623 15:48:56.617259 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.5383 > 30) by scale factor 0.778447 | |
| I0623 15:48:58.147083 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.8699 > 30) by scale factor 0.860342 | |
| I0623 15:51:13.311424 10365 solver.cpp:229] Iteration 7500, loss = 4.62319 | |
| I0623 15:51:13.311558 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.414141 | |
| I0623 15:51:13.311578 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 15:51:13.311591 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 15:51:13.311604 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 15:51:13.311615 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 15:51:13.311627 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 15:51:13.311638 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 15:51:13.311650 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.125 | |
| I0623 15:51:13.311662 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.875 | |
| I0623 15:51:13.311687 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 15:51:13.311702 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 15:51:13.311714 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.125 | |
| I0623 15:51:13.311725 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 15:51:13.311738 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 15:51:13.311748 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 15:51:13.311760 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 15:51:13.311772 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 15:51:13.311784 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 15:51:13.311795 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 15:51:13.311807 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 15:51:13.311818 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:51:13.311830 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:51:13.311841 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:51:13.311852 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.664773 | |
| I0623 15:51:13.311863 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.727273 | |
| I0623 15:51:13.311879 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.74015 (* 0.3 = 0.522046 loss) | |
| I0623 15:51:13.311894 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.01783 (* 0.3 = 0.30535 loss) | |
| I0623 15:51:13.311908 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.628164 (* 0.0272727 = 0.0171318 loss) | |
| I0623 15:51:13.311923 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.27755 (* 0.0272727 = 0.0348423 loss) | |
| I0623 15:51:13.311936 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.49126 (* 0.0272727 = 0.0406707 loss) | |
| I0623 15:51:13.311950 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.84121 (* 0.0272727 = 0.0502148 loss) | |
| I0623 15:51:13.311964 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.66026 (* 0.0272727 = 0.0452798 loss) | |
| I0623 15:51:13.311977 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 3.036 (* 0.0272727 = 0.0828001 loss) | |
| I0623 15:51:13.311991 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.66301 (* 0.0272727 = 0.0726276 loss) | |
| I0623 15:51:13.312005 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.11827 (* 0.0272727 = 0.0304982 loss) | |
| I0623 15:51:13.312018 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.2488 (* 0.0272727 = 0.061331 loss) | |
| I0623 15:51:13.312032 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.81383 (* 0.0272727 = 0.0494682 loss) | |
| I0623 15:51:13.312046 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.20901 (* 0.0272727 = 0.0602458 loss) | |
| I0623 15:51:13.312059 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.62495 (* 0.0272727 = 0.0443168 loss) | |
| I0623 15:51:13.312091 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.96823 (* 0.0272727 = 0.0264063 loss) | |
| I0623 15:51:13.312108 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.0812 (* 0.0272727 = 0.0294873 loss) | |
| I0623 15:51:13.312121 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.615381 (* 0.0272727 = 0.0167831 loss) | |
| I0623 15:51:13.312135 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.677687 (* 0.0272727 = 0.0184824 loss) | |
| I0623 15:51:13.312150 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0286284 (* 0.0272727 = 0.000780774 loss) | |
| I0623 15:51:13.312163 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00223901 (* 0.0272727 = 6.1064e-05 loss) | |
| I0623 15:51:13.312177 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.000793321 (* 0.0272727 = 2.1636e-05 loss) | |
| I0623 15:51:13.312191 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000120309 (* 0.0272727 = 3.28115e-06 loss) | |
| I0623 15:51:13.312206 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 3.95428e-05 (* 0.0272727 = 1.07844e-06 loss) | |
| I0623 15:51:13.312219 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 1.64667e-05 (* 0.0272727 = 4.49092e-07 loss) | |
| I0623 15:51:13.312232 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.545455 | |
| I0623 15:51:13.312243 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.75 | |
| I0623 15:51:13.312255 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 15:51:13.312269 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 15:51:13.312281 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 15:51:13.312293 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 15:51:13.312304 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 15:51:13.312314 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.125 | |
| I0623 15:51:13.312326 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 15:51:13.312337 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.125 | |
| I0623 15:51:13.312348 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 15:51:13.312360 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 15:51:13.312371 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 15:51:13.312382 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 15:51:13.312393 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.5 | |
| I0623 15:51:13.312404 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 15:51:13.312415 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 15:51:13.312427 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 15:51:13.312438 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 15:51:13.312448 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 15:51:13.312459 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:51:13.312471 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:51:13.312482 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:51:13.312492 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.732955 | |
| I0623 15:51:13.312505 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.808081 | |
| I0623 15:51:13.312517 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.30774 (* 0.3 = 0.392321 loss) | |
| I0623 15:51:13.312531 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.778157 (* 0.3 = 0.233447 loss) | |
| I0623 15:51:13.312544 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.400737 (* 0.0272727 = 0.0109292 loss) | |
| I0623 15:51:13.312558 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.301631 (* 0.0272727 = 0.00822629 loss) | |
| I0623 15:51:13.312587 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.54892 (* 0.0272727 = 0.0149705 loss) | |
| I0623 15:51:13.312602 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.909195 (* 0.0272727 = 0.0247962 loss) | |
| I0623 15:51:13.312616 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.93779 (* 0.0272727 = 0.0528487 loss) | |
| I0623 15:51:13.312630 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.02357 (* 0.0272727 = 0.0551883 loss) | |
| I0623 15:51:13.312644 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 2.25887 (* 0.0272727 = 0.0616054 loss) | |
| I0623 15:51:13.312657 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 0.959831 (* 0.0272727 = 0.0261772 loss) | |
| I0623 15:51:13.312670 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 2.77152 (* 0.0272727 = 0.0755869 loss) | |
| I0623 15:51:13.312685 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.487 (* 0.0272727 = 0.0405546 loss) | |
| I0623 15:51:13.312697 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.47962 (* 0.0272727 = 0.0403534 loss) | |
| I0623 15:51:13.312711 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.31943 (* 0.0272727 = 0.0359844 loss) | |
| I0623 15:51:13.312724 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.11823 (* 0.0272727 = 0.0304972 loss) | |
| I0623 15:51:13.312737 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.16624 (* 0.0272727 = 0.0318064 loss) | |
| I0623 15:51:13.312752 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.637171 (* 0.0272727 = 0.0173774 loss) | |
| I0623 15:51:13.312764 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.824463 (* 0.0272727 = 0.0224854 loss) | |
| I0623 15:51:13.312778 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0592322 (* 0.0272727 = 0.00161542 loss) | |
| I0623 15:51:13.312793 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0218118 (* 0.0272727 = 0.000594868 loss) | |
| I0623 15:51:13.312806 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0082962 (* 0.0272727 = 0.00022626 loss) | |
| I0623 15:51:13.312820 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00178129 (* 0.0272727 = 4.85806e-05 loss) | |
| I0623 15:51:13.312834 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000726518 (* 0.0272727 = 1.98141e-05 loss) | |
| I0623 15:51:13.312849 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.23385e-05 (* 0.0272727 = 3.36504e-07 loss) | |
| I0623 15:51:13.312861 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.848485 | |
| I0623 15:51:13.312873 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 15:51:13.312885 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 15:51:13.312896 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 15:51:13.312907 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 15:51:13.312918 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 15:51:13.312929 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 15:51:13.312942 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 15:51:13.312952 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 15:51:13.312963 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.375 | |
| I0623 15:51:13.312975 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 15:51:13.312986 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 15:51:13.312997 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 15:51:13.313009 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 1 | |
| I0623 15:51:13.313020 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 15:51:13.313031 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 15:51:13.313043 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 15:51:13.313065 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 15:51:13.313077 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 15:51:13.313088 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 15:51:13.313100 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:51:13.313112 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:51:13.313123 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:51:13.313134 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.903409 | |
| I0623 15:51:13.313146 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.979798 | |
| I0623 15:51:13.313159 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.575714 (* 1 = 0.575714 loss) | |
| I0623 15:51:13.313174 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.347227 (* 1 = 0.347227 loss) | |
| I0623 15:51:13.313186 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.110935 (* 0.0909091 = 0.010085 loss) | |
| I0623 15:51:13.313200 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0613462 (* 0.0909091 = 0.00557693 loss) | |
| I0623 15:51:13.313215 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.09217 (* 0.0909091 = 0.00837909 loss) | |
| I0623 15:51:13.313228 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.186683 (* 0.0909091 = 0.0169712 loss) | |
| I0623 15:51:13.313242 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.224123 (* 0.0909091 = 0.0203748 loss) | |
| I0623 15:51:13.313256 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.10579 (* 0.0909091 = 0.00961724 loss) | |
| I0623 15:51:13.313271 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 1.57188 (* 0.0909091 = 0.142898 loss) | |
| I0623 15:51:13.313284 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.393054 (* 0.0909091 = 0.0357322 loss) | |
| I0623 15:51:13.313303 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 2.16067 (* 0.0909091 = 0.196425 loss) | |
| I0623 15:51:13.313320 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.33495 (* 0.0909091 = 0.121359 loss) | |
| I0623 15:51:13.313334 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.24061 (* 0.0909091 = 0.112783 loss) | |
| I0623 15:51:13.313349 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.0873 (* 0.0909091 = 0.0988459 loss) | |
| I0623 15:51:13.313361 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.323039 (* 0.0909091 = 0.0293672 loss) | |
| I0623 15:51:13.313374 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.925735 (* 0.0909091 = 0.0841577 loss) | |
| I0623 15:51:13.313388 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.550536 (* 0.0909091 = 0.0500487 loss) | |
| I0623 15:51:13.313401 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.25512 (* 0.0909091 = 0.0231927 loss) | |
| I0623 15:51:13.313415 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.166457 (* 0.0909091 = 0.0151324 loss) | |
| I0623 15:51:13.313428 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0199539 (* 0.0909091 = 0.00181399 loss) | |
| I0623 15:51:13.313442 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00644312 (* 0.0909091 = 0.000585738 loss) | |
| I0623 15:51:13.313457 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000230736 (* 0.0909091 = 2.0976e-05 loss) | |
| I0623 15:51:13.313469 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 5.16352e-05 (* 0.0909091 = 4.69411e-06 loss) | |
| I0623 15:51:13.313483 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 8.50867e-06 (* 0.0909091 = 7.73516e-07 loss) | |
| I0623 15:51:13.313495 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 15:51:13.313506 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 15:51:13.313518 10365 solver.cpp:245] Train net output #149: total_confidence = 0.159252 | |
| I0623 15:51:13.313539 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.128584 | |
| I0623 15:51:13.313552 10365 sgd_solver.cpp:106] Iteration 7500, lr = 0.001 | |
| I0623 15:57:35.845955 10365 solver.cpp:229] Iteration 8000, loss = 4.68155 | |
| I0623 15:57:35.846097 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.5 | |
| I0623 15:57:35.846117 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 15:57:35.846129 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 15:57:35.846141 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.625 | |
| I0623 15:57:35.846153 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 15:57:35.846164 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 15:57:35.846176 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.625 | |
| I0623 15:57:35.846189 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 15:57:35.846200 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 15:57:35.846211 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.625 | |
| I0623 15:57:35.846223 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 15:57:35.846235 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 15:57:35.846246 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.5 | |
| I0623 15:57:35.846258 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.375 | |
| I0623 15:57:35.846272 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 15:57:35.846284 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 15:57:35.846297 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 15:57:35.846307 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 15:57:35.846326 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 15:57:35.846339 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 15:57:35.846351 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 15:57:35.846362 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 15:57:35.846374 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 15:57:35.846385 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.715909 | |
| I0623 15:57:35.846397 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.81 | |
| I0623 15:57:35.846415 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.46102 (* 0.3 = 0.438305 loss) | |
| I0623 15:57:35.846429 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.846506 (* 0.3 = 0.253952 loss) | |
| I0623 15:57:35.846443 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.53186 (* 0.0272727 = 0.0145053 loss) | |
| I0623 15:57:35.846457 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.72384 (* 0.0272727 = 0.0470139 loss) | |
| I0623 15:57:35.846470 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.0939 (* 0.0272727 = 0.0298335 loss) | |
| I0623 15:57:35.846483 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.44572 (* 0.0272727 = 0.0394286 loss) | |
| I0623 15:57:35.846498 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.51999 (* 0.0272727 = 0.0414543 loss) | |
| I0623 15:57:35.846511 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.74619 (* 0.0272727 = 0.0476233 loss) | |
| I0623 15:57:35.846524 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.79543 (* 0.0272727 = 0.0489662 loss) | |
| I0623 15:57:35.846539 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.31697 (* 0.0272727 = 0.0359174 loss) | |
| I0623 15:57:35.846551 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.80473 (* 0.0272727 = 0.0492198 loss) | |
| I0623 15:57:35.846565 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.86194 (* 0.0272727 = 0.0507802 loss) | |
| I0623 15:57:35.846580 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.64077 (* 0.0272727 = 0.0447484 loss) | |
| I0623 15:57:35.846592 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.54266 (* 0.0272727 = 0.0420726 loss) | |
| I0623 15:57:35.846624 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.0383 (* 0.0272727 = 0.0555899 loss) | |
| I0623 15:57:35.846639 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.55534 (* 0.0272727 = 0.0424183 loss) | |
| I0623 15:57:35.846653 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.761938 (* 0.0272727 = 0.0207801 loss) | |
| I0623 15:57:35.846673 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.92381 (* 0.0272727 = 0.0251948 loss) | |
| I0623 15:57:35.846686 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.378178 (* 0.0272727 = 0.0103139 loss) | |
| I0623 15:57:35.846700 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.726886 (* 0.0272727 = 0.0198242 loss) | |
| I0623 15:57:35.846714 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00698432 (* 0.0272727 = 0.000190481 loss) | |
| I0623 15:57:35.846727 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00275545 (* 0.0272727 = 7.51486e-05 loss) | |
| I0623 15:57:35.846742 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000389427 (* 0.0272727 = 1.06207e-05 loss) | |
| I0623 15:57:35.846756 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 2.77773e-05 (* 0.0272727 = 7.57563e-07 loss) | |
| I0623 15:57:35.846768 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.6 | |
| I0623 15:57:35.846781 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 15:57:35.846791 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 15:57:35.846802 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 15:57:35.846813 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 15:57:35.846824 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 15:57:35.846837 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 15:57:35.846848 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.625 | |
| I0623 15:57:35.846858 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.625 | |
| I0623 15:57:35.846869 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.625 | |
| I0623 15:57:35.846881 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 15:57:35.846892 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.75 | |
| I0623 15:57:35.846904 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 15:57:35.846915 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 15:57:35.846926 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 15:57:35.846937 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 15:57:35.846948 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 15:57:35.846959 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 15:57:35.846971 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 15:57:35.846982 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 15:57:35.846993 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 15:57:35.847004 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 15:57:35.847015 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 15:57:35.847026 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.767045 | |
| I0623 15:57:35.847038 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.8 | |
| I0623 15:57:35.847050 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.26179 (* 0.3 = 0.378538 loss) | |
| I0623 15:57:35.847064 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.731649 (* 0.3 = 0.219495 loss) | |
| I0623 15:57:35.847079 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.165604 (* 0.0272727 = 0.00451647 loss) | |
| I0623 15:57:35.847091 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.22812 (* 0.0272727 = 0.00622145 loss) | |
| I0623 15:57:35.847121 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.290041 (* 0.0272727 = 0.0079102 loss) | |
| I0623 15:57:35.847136 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.697065 (* 0.0272727 = 0.0190109 loss) | |
| I0623 15:57:35.847149 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.88426 (* 0.0272727 = 0.051389 loss) | |
| I0623 15:57:35.847162 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 0.92622 (* 0.0272727 = 0.0252606 loss) | |
| I0623 15:57:35.847177 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.11846 (* 0.0272727 = 0.0305034 loss) | |
| I0623 15:57:35.847189 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.0575 (* 0.0272727 = 0.0288409 loss) | |
| I0623 15:57:35.847203 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.33321 (* 0.0272727 = 0.0363602 loss) | |
| I0623 15:57:35.847218 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.61566 (* 0.0272727 = 0.0440635 loss) | |
| I0623 15:57:35.847230 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.07024 (* 0.0272727 = 0.0291885 loss) | |
| I0623 15:57:35.847244 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.33093 (* 0.0272727 = 0.036298 loss) | |
| I0623 15:57:35.847257 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.47105 (* 0.0272727 = 0.0401196 loss) | |
| I0623 15:57:35.847270 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.0876 (* 0.0272727 = 0.0296617 loss) | |
| I0623 15:57:35.847283 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.891502 (* 0.0272727 = 0.0243137 loss) | |
| I0623 15:57:35.847297 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.924082 (* 0.0272727 = 0.0252022 loss) | |
| I0623 15:57:35.847314 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.404324 (* 0.0272727 = 0.011027 loss) | |
| I0623 15:57:35.847328 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.619371 (* 0.0272727 = 0.0168919 loss) | |
| I0623 15:57:35.847342 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0407015 (* 0.0272727 = 0.00111004 loss) | |
| I0623 15:57:35.847355 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.010894 (* 0.0272727 = 0.00029711 loss) | |
| I0623 15:57:35.847369 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00543741 (* 0.0272727 = 0.000148293 loss) | |
| I0623 15:57:35.847383 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000397865 (* 0.0272727 = 1.08509e-05 loss) | |
| I0623 15:57:35.847395 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.82 | |
| I0623 15:57:35.847407 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 15:57:35.847419 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 15:57:35.847430 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 15:57:35.847441 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 15:57:35.847452 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 15:57:35.847463 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 15:57:35.847475 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 15:57:35.847486 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 15:57:35.847496 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 1 | |
| I0623 15:57:35.847507 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 15:57:35.847519 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 15:57:35.847530 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.375 | |
| I0623 15:57:35.847542 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 15:57:35.847553 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 15:57:35.847563 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 15:57:35.847575 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.625 | |
| I0623 15:57:35.847607 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 15:57:35.847622 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 15:57:35.847635 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 15:57:35.847645 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 15:57:35.847656 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 15:57:35.847667 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 15:57:35.847678 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.897727 | |
| I0623 15:57:35.847690 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.98 | |
| I0623 15:57:35.847704 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.530456 (* 1 = 0.530456 loss) | |
| I0623 15:57:35.847718 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.305773 (* 1 = 0.305773 loss) | |
| I0623 15:57:35.847733 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.143345 (* 0.0909091 = 0.0130313 loss) | |
| I0623 15:57:35.847745 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.045043 (* 0.0909091 = 0.00409482 loss) | |
| I0623 15:57:35.847759 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0404457 (* 0.0909091 = 0.00367688 loss) | |
| I0623 15:57:35.847772 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0420678 (* 0.0909091 = 0.00382435 loss) | |
| I0623 15:57:35.847786 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.367247 (* 0.0909091 = 0.0333861 loss) | |
| I0623 15:57:35.847800 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.112264 (* 0.0909091 = 0.0102058 loss) | |
| I0623 15:57:35.847813 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.198991 (* 0.0909091 = 0.0180901 loss) | |
| I0623 15:57:35.847831 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.247533 (* 0.0909091 = 0.022503 loss) | |
| I0623 15:57:35.847846 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.317606 (* 0.0909091 = 0.0288733 loss) | |
| I0623 15:57:35.847858 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.19004 (* 0.0909091 = 0.108185 loss) | |
| I0623 15:57:35.847872 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.991041 (* 0.0909091 = 0.0900946 loss) | |
| I0623 15:57:35.847892 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.23722 (* 0.0909091 = 0.112474 loss) | |
| I0623 15:57:35.847910 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.24999 (* 0.0909091 = 0.113635 loss) | |
| I0623 15:57:35.847924 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.04352 (* 0.0909091 = 0.094865 loss) | |
| I0623 15:57:35.847937 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.881719 (* 0.0909091 = 0.0801563 loss) | |
| I0623 15:57:35.847950 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.915944 (* 0.0909091 = 0.0832677 loss) | |
| I0623 15:57:35.847965 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.343161 (* 0.0909091 = 0.0311964 loss) | |
| I0623 15:57:35.847977 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.12338 (* 0.0909091 = 0.0112163 loss) | |
| I0623 15:57:35.847991 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0504273 (* 0.0909091 = 0.0045843 loss) | |
| I0623 15:57:35.848004 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00599441 (* 0.0909091 = 0.000544946 loss) | |
| I0623 15:57:35.848018 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000208473 (* 0.0909091 = 1.89521e-05 loss) | |
| I0623 15:57:35.848031 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 3.61161e-05 (* 0.0909091 = 3.28328e-06 loss) | |
| I0623 15:57:35.848043 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 15:57:35.848054 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 15:57:35.848065 10365 solver.cpp:245] Train net output #149: total_confidence = 0.260431 | |
| I0623 15:57:35.848088 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.240707 | |
| I0623 15:57:35.848103 10365 sgd_solver.cpp:106] Iteration 8000, lr = 0.001 | |
| I0623 15:59:24.895273 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.8639 > 30) by scale factor 0.836496 | |
| I0623 16:03:58.516383 10365 solver.cpp:229] Iteration 8500, loss = 4.72581 | |
| I0623 16:03:58.516505 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.412844 | |
| I0623 16:03:58.516525 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 16:03:58.516537 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 16:03:58.516549 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 16:03:58.516562 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.125 | |
| I0623 16:03:58.516574 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 16:03:58.516587 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 16:03:58.516598 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.125 | |
| I0623 16:03:58.516610 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 16:03:58.516623 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 16:03:58.516633 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 16:03:58.516645 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 16:03:58.516657 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 16:03:58.516669 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.25 | |
| I0623 16:03:58.516680 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.375 | |
| I0623 16:03:58.516692 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.375 | |
| I0623 16:03:58.516705 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.5 | |
| I0623 16:03:58.516716 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 16:03:58.516728 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 16:03:58.516741 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:03:58.516752 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:03:58.516763 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:03:58.516774 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:03:58.516787 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.607955 | |
| I0623 16:03:58.516798 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.752294 | |
| I0623 16:03:58.516813 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.81143 (* 0.3 = 0.54343 loss) | |
| I0623 16:03:58.516827 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.22757 (* 0.3 = 0.368271 loss) | |
| I0623 16:03:58.516842 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.159736 (* 0.0272727 = 0.00435643 loss) | |
| I0623 16:03:58.516855 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.57357 (* 0.0272727 = 0.0429155 loss) | |
| I0623 16:03:58.516870 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.46015 (* 0.0272727 = 0.067095 loss) | |
| I0623 16:03:58.516882 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.0689 (* 0.0272727 = 0.0564245 loss) | |
| I0623 16:03:58.516897 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 3.34504 (* 0.0272727 = 0.0912283 loss) | |
| I0623 16:03:58.516912 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.74757 (* 0.0272727 = 0.0749337 loss) | |
| I0623 16:03:58.516926 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.80093 (* 0.0272727 = 0.076389 loss) | |
| I0623 16:03:58.516940 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.53418 (* 0.0272727 = 0.0418412 loss) | |
| I0623 16:03:58.516954 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.06577 (* 0.0272727 = 0.0563393 loss) | |
| I0623 16:03:58.516968 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.73879 (* 0.0272727 = 0.0474216 loss) | |
| I0623 16:03:58.516981 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.8282 (* 0.0272727 = 0.0498599 loss) | |
| I0623 16:03:58.516994 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 3.13639 (* 0.0272727 = 0.085538 loss) | |
| I0623 16:03:58.517026 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.2457 (* 0.0272727 = 0.0612465 loss) | |
| I0623 16:03:58.517041 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 2.17106 (* 0.0272727 = 0.0592107 loss) | |
| I0623 16:03:58.517055 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.48922 (* 0.0272727 = 0.0406151 loss) | |
| I0623 16:03:58.517068 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.22992 (* 0.0272727 = 0.0335432 loss) | |
| I0623 16:03:58.517082 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.517913 (* 0.0272727 = 0.0141249 loss) | |
| I0623 16:03:58.517096 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.684784 (* 0.0272727 = 0.0186759 loss) | |
| I0623 16:03:58.517109 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0020206 (* 0.0272727 = 5.51073e-05 loss) | |
| I0623 16:03:58.517124 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00119467 (* 0.0272727 = 3.2582e-05 loss) | |
| I0623 16:03:58.517138 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000768324 (* 0.0272727 = 2.09543e-05 loss) | |
| I0623 16:03:58.517153 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 5.32022e-05 (* 0.0272727 = 1.45097e-06 loss) | |
| I0623 16:03:58.517164 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.504587 | |
| I0623 16:03:58.517176 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 16:03:58.517187 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 16:03:58.517199 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.375 | |
| I0623 16:03:58.517210 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 16:03:58.517222 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 16:03:58.517233 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 16:03:58.517244 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 16:03:58.517256 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 16:03:58.517271 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 16:03:58.517282 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 16:03:58.517293 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.25 | |
| I0623 16:03:58.517304 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.25 | |
| I0623 16:03:58.517315 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 16:03:58.517328 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.5 | |
| I0623 16:03:58.517338 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.5 | |
| I0623 16:03:58.517349 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.625 | |
| I0623 16:03:58.517361 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 16:03:58.517372 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 16:03:58.517384 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:03:58.517395 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:03:58.517406 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:03:58.517417 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:03:58.517428 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.670455 | |
| I0623 16:03:58.517439 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.752294 | |
| I0623 16:03:58.517453 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.67751 (* 0.3 = 0.503253 loss) | |
| I0623 16:03:58.517467 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 1.13514 (* 0.3 = 0.340543 loss) | |
| I0623 16:03:58.517482 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.295679 (* 0.0272727 = 0.00806396 loss) | |
| I0623 16:03:58.517495 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.793564 (* 0.0272727 = 0.0216427 loss) | |
| I0623 16:03:58.517524 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.63619 (* 0.0272727 = 0.0446235 loss) | |
| I0623 16:03:58.517539 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.48718 (* 0.0272727 = 0.0405594 loss) | |
| I0623 16:03:58.517554 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.13427 (* 0.0272727 = 0.0582073 loss) | |
| I0623 16:03:58.517567 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.26047 (* 0.0272727 = 0.0616491 loss) | |
| I0623 16:03:58.517580 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 2.27143 (* 0.0272727 = 0.061948 loss) | |
| I0623 16:03:58.517595 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.56141 (* 0.0272727 = 0.042584 loss) | |
| I0623 16:03:58.517607 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.7155 (* 0.0272727 = 0.0467864 loss) | |
| I0623 16:03:58.517621 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.04313 (* 0.0272727 = 0.0557217 loss) | |
| I0623 16:03:58.517635 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.22732 (* 0.0272727 = 0.0607452 loss) | |
| I0623 16:03:58.517648 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 2.72389 (* 0.0272727 = 0.0742879 loss) | |
| I0623 16:03:58.517662 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.9101 (* 0.0272727 = 0.0520936 loss) | |
| I0623 16:03:58.517675 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.66185 (* 0.0272727 = 0.0453233 loss) | |
| I0623 16:03:58.517688 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.1148 (* 0.0272727 = 0.0304037 loss) | |
| I0623 16:03:58.517702 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 1.40222 (* 0.0272727 = 0.0382424 loss) | |
| I0623 16:03:58.517716 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.335867 (* 0.0272727 = 0.00916 loss) | |
| I0623 16:03:58.517729 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.501614 (* 0.0272727 = 0.0136804 loss) | |
| I0623 16:03:58.517743 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.044311 (* 0.0272727 = 0.00120848 loss) | |
| I0623 16:03:58.517757 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00855597 (* 0.0272727 = 0.000233345 loss) | |
| I0623 16:03:58.517771 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00484023 (* 0.0272727 = 0.000132006 loss) | |
| I0623 16:03:58.517786 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.00117267 (* 0.0272727 = 3.19819e-05 loss) | |
| I0623 16:03:58.517798 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.733945 | |
| I0623 16:03:58.517812 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 16:03:58.517820 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 16:03:58.517828 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:03:58.517840 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 16:03:58.517853 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 16:03:58.517863 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.75 | |
| I0623 16:03:58.517875 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.625 | |
| I0623 16:03:58.517886 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.625 | |
| I0623 16:03:58.517897 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 16:03:58.517909 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 16:03:58.517920 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 16:03:58.517931 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.25 | |
| I0623 16:03:58.517942 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 16:03:58.517953 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 16:03:58.517964 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 16:03:58.517976 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 16:03:58.517997 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 16:03:58.518009 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.875 | |
| I0623 16:03:58.518021 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:03:58.518033 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:03:58.518043 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:03:58.518054 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:03:58.518065 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.8125 | |
| I0623 16:03:58.518077 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.908257 | |
| I0623 16:03:58.518090 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.909569 (* 1 = 0.909569 loss) | |
| I0623 16:03:58.518105 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.665732 (* 1 = 0.665732 loss) | |
| I0623 16:03:58.518118 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.069955 (* 0.0909091 = 0.00635954 loss) | |
| I0623 16:03:58.518132 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0516536 (* 0.0909091 = 0.00469579 loss) | |
| I0623 16:03:58.518146 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.209125 (* 0.0909091 = 0.0190114 loss) | |
| I0623 16:03:58.518159 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.531319 (* 0.0909091 = 0.0483017 loss) | |
| I0623 16:03:58.518172 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.928384 (* 0.0909091 = 0.0843985 loss) | |
| I0623 16:03:58.518185 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 1.42315 (* 0.0909091 = 0.129377 loss) | |
| I0623 16:03:58.518199 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.767156 (* 0.0909091 = 0.0697414 loss) | |
| I0623 16:03:58.518213 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.90489 (* 0.0909091 = 0.173172 loss) | |
| I0623 16:03:58.518225 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.876104 (* 0.0909091 = 0.0796458 loss) | |
| I0623 16:03:58.518239 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.17596 (* 0.0909091 = 0.106906 loss) | |
| I0623 16:03:58.518252 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.30905 (* 0.0909091 = 0.119004 loss) | |
| I0623 16:03:58.518265 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 2.05956 (* 0.0909091 = 0.187232 loss) | |
| I0623 16:03:58.518280 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.45365 (* 0.0909091 = 0.13215 loss) | |
| I0623 16:03:58.518293 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.21017 (* 0.0909091 = 0.110016 loss) | |
| I0623 16:03:58.518306 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.71372 (* 0.0909091 = 0.0648836 loss) | |
| I0623 16:03:58.518323 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.423355 (* 0.0909091 = 0.0384868 loss) | |
| I0623 16:03:58.518337 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.176054 (* 0.0909091 = 0.0160049 loss) | |
| I0623 16:03:58.518352 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.350088 (* 0.0909091 = 0.0318262 loss) | |
| I0623 16:03:58.518365 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0219449 (* 0.0909091 = 0.001995 loss) | |
| I0623 16:03:58.518379 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00597292 (* 0.0909091 = 0.000542993 loss) | |
| I0623 16:03:58.518393 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000903855 (* 0.0909091 = 8.21686e-05 loss) | |
| I0623 16:03:58.518407 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 0.000128208 (* 0.0909091 = 1.16553e-05 loss) | |
| I0623 16:03:58.518419 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 16:03:58.518431 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 16:03:58.518442 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0911826 | |
| I0623 16:03:58.518463 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0790213 | |
| I0623 16:03:58.518478 10365 sgd_solver.cpp:106] Iteration 8500, lr = 0.001 | |
| I0623 16:10:21.061769 10365 solver.cpp:229] Iteration 9000, loss = 4.6969 | |
| I0623 16:10:21.061908 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.43 | |
| I0623 16:10:21.061928 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 16:10:21.061939 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 16:10:21.061952 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.125 | |
| I0623 16:10:21.061964 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.25 | |
| I0623 16:10:21.061976 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.125 | |
| I0623 16:10:21.061987 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 16:10:21.062000 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.25 | |
| I0623 16:10:21.062011 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 16:10:21.062022 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 16:10:21.062034 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 16:10:21.062047 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 16:10:21.062057 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 16:10:21.062069 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 16:10:21.062082 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 16:10:21.062093 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.875 | |
| I0623 16:10:21.062104 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 16:10:21.062115 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 16:10:21.062126 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 16:10:21.062139 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:10:21.062150 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:10:21.062160 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:10:21.062171 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:10:21.062183 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.664773 | |
| I0623 16:10:21.062194 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.76 | |
| I0623 16:10:21.062211 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.67442 (* 0.3 = 0.502327 loss) | |
| I0623 16:10:21.062225 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.00564 (* 0.3 = 0.301692 loss) | |
| I0623 16:10:21.062240 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.333739 (* 0.0272727 = 0.00910197 loss) | |
| I0623 16:10:21.062255 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.991449 (* 0.0272727 = 0.0270395 loss) | |
| I0623 16:10:21.062273 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.96557 (* 0.0272727 = 0.0536064 loss) | |
| I0623 16:10:21.062286 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.61437 (* 0.0272727 = 0.0440283 loss) | |
| I0623 16:10:21.062300 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.05065 (* 0.0272727 = 0.0559268 loss) | |
| I0623 16:10:21.062314 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.30687 (* 0.0272727 = 0.0629147 loss) | |
| I0623 16:10:21.062327 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.24649 (* 0.0272727 = 0.0612679 loss) | |
| I0623 16:10:21.062340 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.79593 (* 0.0272727 = 0.0489798 loss) | |
| I0623 16:10:21.062355 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.77877 (* 0.0272727 = 0.0757847 loss) | |
| I0623 16:10:21.062367 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.84256 (* 0.0272727 = 0.0502516 loss) | |
| I0623 16:10:21.062381 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.56808 (* 0.0272727 = 0.0427659 loss) | |
| I0623 16:10:21.062396 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.40566 (* 0.0272727 = 0.0383362 loss) | |
| I0623 16:10:21.062425 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.975662 (* 0.0272727 = 0.026609 loss) | |
| I0623 16:10:21.062440 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.78961 (* 0.0272727 = 0.0488077 loss) | |
| I0623 16:10:21.062454 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.740884 (* 0.0272727 = 0.0202059 loss) | |
| I0623 16:10:21.062469 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.28068 (* 0.0272727 = 0.00765492 loss) | |
| I0623 16:10:21.062482 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.681794 (* 0.0272727 = 0.0185944 loss) | |
| I0623 16:10:21.062496 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00203049 (* 0.0272727 = 5.53771e-05 loss) | |
| I0623 16:10:21.062510 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00103173 (* 0.0272727 = 2.81382e-05 loss) | |
| I0623 16:10:21.062525 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000237043 (* 0.0272727 = 6.4648e-06 loss) | |
| I0623 16:10:21.062538 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000187803 (* 0.0272727 = 5.12189e-06 loss) | |
| I0623 16:10:21.062552 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 7.48904e-05 (* 0.0272727 = 2.04247e-06 loss) | |
| I0623 16:10:21.062564 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.55 | |
| I0623 16:10:21.062577 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 16:10:21.062588 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 16:10:21.062599 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 16:10:21.062610 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 16:10:21.062623 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 16:10:21.062633 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 16:10:21.062644 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 16:10:21.062655 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 16:10:21.062667 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.25 | |
| I0623 16:10:21.062679 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 16:10:21.062690 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 16:10:21.062700 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 16:10:21.062711 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 16:10:21.062722 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 16:10:21.062733 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 16:10:21.062744 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 16:10:21.062755 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 16:10:21.062767 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 16:10:21.062778 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:10:21.062789 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:10:21.062800 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:10:21.062813 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:10:21.062824 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.727273 | |
| I0623 16:10:21.062835 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.86 | |
| I0623 16:10:21.062849 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.29535 (* 0.3 = 0.388606 loss) | |
| I0623 16:10:21.062862 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.790853 (* 0.3 = 0.237256 loss) | |
| I0623 16:10:21.062876 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.289238 (* 0.0272727 = 0.00788831 loss) | |
| I0623 16:10:21.062891 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.455389 (* 0.0272727 = 0.0124197 loss) | |
| I0623 16:10:21.062914 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.599016 (* 0.0272727 = 0.0163368 loss) | |
| I0623 16:10:21.062934 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.03947 (* 0.0272727 = 0.0283491 loss) | |
| I0623 16:10:21.062948 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.42266 (* 0.0272727 = 0.0387999 loss) | |
| I0623 16:10:21.062963 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.00589 (* 0.0272727 = 0.0547061 loss) | |
| I0623 16:10:21.062975 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.59493 (* 0.0272727 = 0.043498 loss) | |
| I0623 16:10:21.062989 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.7714 (* 0.0272727 = 0.0483108 loss) | |
| I0623 16:10:21.063004 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.73712 (* 0.0272727 = 0.047376 loss) | |
| I0623 16:10:21.063020 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.87536 (* 0.0272727 = 0.0511462 loss) | |
| I0623 16:10:21.063030 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.39363 (* 0.0272727 = 0.0380082 loss) | |
| I0623 16:10:21.063045 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.96765 (* 0.0272727 = 0.0536633 loss) | |
| I0623 16:10:21.063057 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.16817 (* 0.0272727 = 0.0318593 loss) | |
| I0623 16:10:21.063071 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.47961 (* 0.0272727 = 0.040353 loss) | |
| I0623 16:10:21.063084 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.459394 (* 0.0272727 = 0.0125289 loss) | |
| I0623 16:10:21.063098 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.171985 (* 0.0272727 = 0.00469051 loss) | |
| I0623 16:10:21.063112 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.342981 (* 0.0272727 = 0.00935403 loss) | |
| I0623 16:10:21.063125 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00243756 (* 0.0272727 = 6.64788e-05 loss) | |
| I0623 16:10:21.063139 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000827083 (* 0.0272727 = 2.25568e-05 loss) | |
| I0623 16:10:21.063153 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000291766 (* 0.0272727 = 7.95725e-06 loss) | |
| I0623 16:10:21.063166 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 9.69804e-05 (* 0.0272727 = 2.64492e-06 loss) | |
| I0623 16:10:21.063180 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 2.39322e-05 (* 0.0272727 = 6.52695e-07 loss) | |
| I0623 16:10:21.063192 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.78 | |
| I0623 16:10:21.063205 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 16:10:21.063215 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 16:10:21.063226 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:10:21.063237 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 16:10:21.063249 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 16:10:21.063261 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 16:10:21.063271 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 16:10:21.063282 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.5 | |
| I0623 16:10:21.063293 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 16:10:21.063304 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 16:10:21.063318 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 16:10:21.063330 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 16:10:21.063341 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.5 | |
| I0623 16:10:21.063352 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 16:10:21.063364 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 16:10:21.063375 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 16:10:21.063401 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 16:10:21.063415 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 16:10:21.063426 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:10:21.063436 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:10:21.063447 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:10:21.063458 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:10:21.063469 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.857955 | |
| I0623 16:10:21.063482 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.97 | |
| I0623 16:10:21.063494 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.693111 (* 1 = 0.693111 loss) | |
| I0623 16:10:21.063508 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.470906 (* 1 = 0.470906 loss) | |
| I0623 16:10:21.063522 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0307177 (* 0.0909091 = 0.00279252 loss) | |
| I0623 16:10:21.063536 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0446406 (* 0.0909091 = 0.00405824 loss) | |
| I0623 16:10:21.063550 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.095436 (* 0.0909091 = 0.008676 loss) | |
| I0623 16:10:21.063565 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.269438 (* 0.0909091 = 0.0244943 loss) | |
| I0623 16:10:21.063577 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.15458 (* 0.0909091 = 0.0140528 loss) | |
| I0623 16:10:21.063591 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.389128 (* 0.0909091 = 0.0353753 loss) | |
| I0623 16:10:21.063619 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.923048 (* 0.0909091 = 0.0839135 loss) | |
| I0623 16:10:21.063634 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.57666 (* 0.0909091 = 0.143333 loss) | |
| I0623 16:10:21.063648 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.33221 (* 0.0909091 = 0.12111 loss) | |
| I0623 16:10:21.063662 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.1767 (* 0.0909091 = 0.106973 loss) | |
| I0623 16:10:21.063675 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.701452 (* 0.0909091 = 0.0637684 loss) | |
| I0623 16:10:21.063689 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.47664 (* 0.0909091 = 0.13424 loss) | |
| I0623 16:10:21.063704 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.10597 (* 0.0909091 = 0.100543 loss) | |
| I0623 16:10:21.063716 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.728813 (* 0.0909091 = 0.0662557 loss) | |
| I0623 16:10:21.063730 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.469843 (* 0.0909091 = 0.042713 loss) | |
| I0623 16:10:21.063743 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.154088 (* 0.0909091 = 0.014008 loss) | |
| I0623 16:10:21.063757 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.613136 (* 0.0909091 = 0.0557397 loss) | |
| I0623 16:10:21.063771 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00167606 (* 0.0909091 = 0.000152369 loss) | |
| I0623 16:10:21.063786 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000315785 (* 0.0909091 = 2.87078e-05 loss) | |
| I0623 16:10:21.063801 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000234458 (* 0.0909091 = 2.13144e-05 loss) | |
| I0623 16:10:21.063814 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000119388 (* 0.0909091 = 1.08535e-05 loss) | |
| I0623 16:10:21.063828 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 3.2039e-05 (* 0.0909091 = 2.91263e-06 loss) | |
| I0623 16:10:21.063840 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 16:10:21.063853 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 16:10:21.063863 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0670325 | |
| I0623 16:10:21.063886 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0499258 | |
| I0623 16:10:21.063901 10365 sgd_solver.cpp:106] Iteration 9000, lr = 0.001 | |
| I0623 16:10:47.465718 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.6642 > 30) by scale factor 0.818238 | |
| I0623 16:12:59.889624 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.2974 > 30) by scale factor 0.900971 | |
| I0623 16:16:33.703344 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.6692 > 30) by scale factor 0.918295 | |
| I0623 16:16:44.059309 10365 solver.cpp:229] Iteration 9500, loss = 4.5458 | |
| I0623 16:16:44.059370 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.542857 | |
| I0623 16:16:44.059387 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 16:16:44.059399 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 1 | |
| I0623 16:16:44.059412 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 16:16:44.059423 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 16:16:44.059435 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.5 | |
| I0623 16:16:44.059448 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.625 | |
| I0623 16:16:44.059460 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.625 | |
| I0623 16:16:44.059473 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.75 | |
| I0623 16:16:44.059484 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.875 | |
| I0623 16:16:44.059495 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 16:16:44.059506 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.75 | |
| I0623 16:16:44.059519 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.875 | |
| I0623 16:16:44.059530 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.75 | |
| I0623 16:16:44.059541 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.875 | |
| I0623 16:16:44.059553 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 1 | |
| I0623 16:16:44.059564 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 1 | |
| I0623 16:16:44.059576 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 16:16:44.059587 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 16:16:44.059613 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:16:44.059628 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:16:44.059639 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:16:44.059650 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:16:44.059662 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.795455 | |
| I0623 16:16:44.059674 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.857143 | |
| I0623 16:16:44.059689 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.27754 (* 0.3 = 0.383262 loss) | |
| I0623 16:16:44.059705 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.572564 (* 0.3 = 0.171769 loss) | |
| I0623 16:16:44.059718 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.255881 (* 0.0272727 = 0.00697858 loss) | |
| I0623 16:16:44.059732 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.363017 (* 0.0272727 = 0.00990047 loss) | |
| I0623 16:16:44.059746 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.98508 (* 0.0272727 = 0.0541385 loss) | |
| I0623 16:16:44.059761 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.9615 (* 0.0272727 = 0.0534954 loss) | |
| I0623 16:16:44.059774 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.05241 (* 0.0272727 = 0.0559748 loss) | |
| I0623 16:16:44.059788 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.07021 (* 0.0272727 = 0.0291876 loss) | |
| I0623 16:16:44.059808 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 0.813809 (* 0.0272727 = 0.0221948 loss) | |
| I0623 16:16:44.059823 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 0.720888 (* 0.0272727 = 0.0196606 loss) | |
| I0623 16:16:44.059835 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 0.965097 (* 0.0272727 = 0.0263208 loss) | |
| I0623 16:16:44.059849 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 0.903991 (* 0.0272727 = 0.0246543 loss) | |
| I0623 16:16:44.059864 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.08253 (* 0.0272727 = 0.0295234 loss) | |
| I0623 16:16:44.059876 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 0.820828 (* 0.0272727 = 0.0223862 loss) | |
| I0623 16:16:44.059929 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.622284 (* 0.0272727 = 0.0169714 loss) | |
| I0623 16:16:44.059945 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.353993 (* 0.0272727 = 0.00965434 loss) | |
| I0623 16:16:44.059959 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.0327933 (* 0.0272727 = 0.000894364 loss) | |
| I0623 16:16:44.059973 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.00621857 (* 0.0272727 = 0.000169597 loss) | |
| I0623 16:16:44.059988 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.000489933 (* 0.0272727 = 1.33618e-05 loss) | |
| I0623 16:16:44.060001 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 5.08111e-05 (* 0.0272727 = 1.38576e-06 loss) | |
| I0623 16:16:44.060015 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 5.84131e-06 (* 0.0272727 = 1.59309e-07 loss) | |
| I0623 16:16:44.060029 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 1.53482e-06 (* 0.0272727 = 4.18588e-08 loss) | |
| I0623 16:16:44.060045 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 2.5332e-07 (* 0.0272727 = 6.90872e-09 loss) | |
| I0623 16:16:44.060058 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 1.49012e-08 (* 0.0272727 = 4.06395e-10 loss) | |
| I0623 16:16:44.060070 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.628571 | |
| I0623 16:16:44.060082 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 16:16:44.060094 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 16:16:44.060106 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 16:16:44.060117 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 16:16:44.060128 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 16:16:44.060139 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.875 | |
| I0623 16:16:44.060153 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 16:16:44.060165 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 16:16:44.060176 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.625 | |
| I0623 16:16:44.060189 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.75 | |
| I0623 16:16:44.060199 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 16:16:44.060210 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.75 | |
| I0623 16:16:44.060221 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 16:16:44.060232 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 16:16:44.060245 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 1 | |
| I0623 16:16:44.060256 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 16:16:44.060266 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 16:16:44.060278 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 16:16:44.060289 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:16:44.060300 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:16:44.060312 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:16:44.060322 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:16:44.060333 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.852273 | |
| I0623 16:16:44.060344 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.985714 | |
| I0623 16:16:44.060359 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 0.851176 (* 0.3 = 0.255353 loss) | |
| I0623 16:16:44.060371 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.3561 (* 0.3 = 0.10683 loss) | |
| I0623 16:16:44.060385 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.370202 (* 0.0272727 = 0.0100964 loss) | |
| I0623 16:16:44.060411 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.12406 (* 0.0272727 = 0.00338345 loss) | |
| I0623 16:16:44.060427 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.552757 (* 0.0272727 = 0.0150752 loss) | |
| I0623 16:16:44.060441 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.633004 (* 0.0272727 = 0.0172637 loss) | |
| I0623 16:16:44.060454 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.05418 (* 0.0272727 = 0.0287504 loss) | |
| I0623 16:16:44.060467 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 0.557308 (* 0.0272727 = 0.0151993 loss) | |
| I0623 16:16:44.060482 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 0.984795 (* 0.0272727 = 0.026858 loss) | |
| I0623 16:16:44.060494 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 0.937023 (* 0.0272727 = 0.0255552 loss) | |
| I0623 16:16:44.060508 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 0.724883 (* 0.0272727 = 0.0197695 loss) | |
| I0623 16:16:44.060521 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 0.754706 (* 0.0272727 = 0.0205829 loss) | |
| I0623 16:16:44.060534 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 0.844284 (* 0.0272727 = 0.0230259 loss) | |
| I0623 16:16:44.060547 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 0.823833 (* 0.0272727 = 0.0224682 loss) | |
| I0623 16:16:44.060561 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.602939 (* 0.0272727 = 0.0164438 loss) | |
| I0623 16:16:44.060575 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.341434 (* 0.0272727 = 0.00931184 loss) | |
| I0623 16:16:44.060587 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.0568495 (* 0.0272727 = 0.00155044 loss) | |
| I0623 16:16:44.060601 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.00406943 (* 0.0272727 = 0.000110985 loss) | |
| I0623 16:16:44.060616 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.000447914 (* 0.0272727 = 1.22158e-05 loss) | |
| I0623 16:16:44.060629 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 5.91568e-05 (* 0.0272727 = 1.61337e-06 loss) | |
| I0623 16:16:44.060643 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 4.06861e-05 (* 0.0272727 = 1.10962e-06 loss) | |
| I0623 16:16:44.060657 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 3.44635e-05 (* 0.0272727 = 9.39913e-07 loss) | |
| I0623 16:16:44.060672 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 7.10912e-05 (* 0.0272727 = 1.93885e-06 loss) | |
| I0623 16:16:44.060685 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.08485e-05 (* 0.0272727 = 2.95869e-07 loss) | |
| I0623 16:16:44.060698 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.957143 | |
| I0623 16:16:44.060709 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 16:16:44.060720 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 16:16:44.060732 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:16:44.060745 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 16:16:44.060755 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 16:16:44.060763 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 16:16:44.060770 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 16:16:44.060782 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 16:16:44.060794 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 16:16:44.060806 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 16:16:44.060817 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.875 | |
| I0623 16:16:44.060834 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 16:16:44.060847 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.875 | |
| I0623 16:16:44.060865 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 16:16:44.060878 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 1 | |
| I0623 16:16:44.060899 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 16:16:44.060912 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 16:16:44.060923 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 16:16:44.060935 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:16:44.060945 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:16:44.060956 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:16:44.060967 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:16:44.060979 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.977273 | |
| I0623 16:16:44.060991 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 1 | |
| I0623 16:16:44.061004 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.227967 (* 1 = 0.227967 loss) | |
| I0623 16:16:44.061017 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.0989907 (* 1 = 0.0989907 loss) | |
| I0623 16:16:44.061033 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0202005 (* 0.0909091 = 0.00183641 loss) | |
| I0623 16:16:44.061046 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0227557 (* 0.0909091 = 0.0020687 loss) | |
| I0623 16:16:44.061059 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0495575 (* 0.0909091 = 0.00450522 loss) | |
| I0623 16:16:44.061074 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.132721 (* 0.0909091 = 0.0120656 loss) | |
| I0623 16:16:44.061086 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.200183 (* 0.0909091 = 0.0181985 loss) | |
| I0623 16:16:44.061100 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.0466679 (* 0.0909091 = 0.00424253 loss) | |
| I0623 16:16:44.061115 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.319111 (* 0.0909091 = 0.0290101 loss) | |
| I0623 16:16:44.061127 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.231085 (* 0.0909091 = 0.0210078 loss) | |
| I0623 16:16:44.061141 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.883469 (* 0.0909091 = 0.0803154 loss) | |
| I0623 16:16:44.061154 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.533005 (* 0.0909091 = 0.048455 loss) | |
| I0623 16:16:44.061168 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.395526 (* 0.0909091 = 0.0359569 loss) | |
| I0623 16:16:44.061182 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.73791 (* 0.0909091 = 0.0670827 loss) | |
| I0623 16:16:44.061197 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.221911 (* 0.0909091 = 0.0201737 loss) | |
| I0623 16:16:44.061211 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.332089 (* 0.0909091 = 0.0301899 loss) | |
| I0623 16:16:44.061225 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.0632447 (* 0.0909091 = 0.00574952 loss) | |
| I0623 16:16:44.061239 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.000708894 (* 0.0909091 = 6.44449e-05 loss) | |
| I0623 16:16:44.061252 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.000163486 (* 0.0909091 = 1.48624e-05 loss) | |
| I0623 16:16:44.061266 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 3.63552e-05 (* 0.0909091 = 3.30502e-06 loss) | |
| I0623 16:16:44.061280 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 6.95892e-06 (* 0.0909091 = 6.32629e-07 loss) | |
| I0623 16:16:44.061295 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 6.25855e-06 (* 0.0909091 = 5.68959e-07 loss) | |
| I0623 16:16:44.061308 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 2.51831e-06 (* 0.0909091 = 2.28937e-07 loss) | |
| I0623 16:16:44.061323 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 8.7917e-07 (* 0.0909091 = 7.99246e-08 loss) | |
| I0623 16:16:44.061336 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.625 | |
| I0623 16:16:44.061357 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.5 | |
| I0623 16:16:44.061369 10365 solver.cpp:245] Train net output #149: total_confidence = 0.465604 | |
| I0623 16:16:44.061381 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.457815 | |
| I0623 16:16:44.061394 10365 sgd_solver.cpp:106] Iteration 9500, lr = 0.001 | |
| I0623 16:17:29.647814 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 45.032 > 30) by scale factor 0.666193 | |
| I0623 16:18:04.892212 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.9309 > 30) by scale factor 0.858839 | |
| I0623 16:19:03.893671 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 43.015 > 30) by scale factor 0.697432 | |
| I0623 16:19:21.527099 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.0719 > 30) by scale factor 0.713064 | |
| I0623 16:21:20.329627 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.7698 > 30) by scale factor 0.944292 | |
| I0623 16:21:27.226963 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 39.9294 > 30) by scale factor 0.751325 | |
| I0623 16:23:06.900810 10365 solver.cpp:456] Snapshotting to binary proto file /mnt/snapshots/mixed_lstm22_iter_10000.caffemodel | |
| I0623 16:23:07.620676 10365 sgd_solver.cpp:273] Snapshotting solver state to binary proto file /mnt/snapshots/mixed_lstm22_iter_10000.solverstate | |
| I0623 16:23:08.036142 10365 solver.cpp:338] Iteration 10000, Testing net (#0) | |
| I0623 16:24:05.061308 10365 solver.cpp:393] Test loss: 3.98854 | |
| I0623 16:24:05.061450 10365 solver.cpp:406] Test net output #0: loss1/accuracy = 0.532218 | |
| I0623 16:24:05.061470 10365 solver.cpp:406] Test net output #1: loss1/accuracy01 = 0.937 | |
| I0623 16:24:05.061483 10365 solver.cpp:406] Test net output #2: loss1/accuracy02 = 0.786 | |
| I0623 16:24:05.061496 10365 solver.cpp:406] Test net output #3: loss1/accuracy03 = 0.545 | |
| I0623 16:24:05.061508 10365 solver.cpp:406] Test net output #4: loss1/accuracy04 = 0.47 | |
| I0623 16:24:05.061520 10365 solver.cpp:406] Test net output #5: loss1/accuracy05 = 0.407 | |
| I0623 16:24:05.061533 10365 solver.cpp:406] Test net output #6: loss1/accuracy06 = 0.439 | |
| I0623 16:24:05.061545 10365 solver.cpp:406] Test net output #7: loss1/accuracy07 = 0.387 | |
| I0623 16:24:05.061558 10365 solver.cpp:406] Test net output #8: loss1/accuracy08 = 0.482 | |
| I0623 16:24:05.061569 10365 solver.cpp:406] Test net output #9: loss1/accuracy09 = 0.427 | |
| I0623 16:24:05.061580 10365 solver.cpp:406] Test net output #10: loss1/accuracy10 = 0.406 | |
| I0623 16:24:05.061592 10365 solver.cpp:406] Test net output #11: loss1/accuracy11 = 0.407 | |
| I0623 16:24:05.061604 10365 solver.cpp:406] Test net output #12: loss1/accuracy12 = 0.479 | |
| I0623 16:24:05.061616 10365 solver.cpp:406] Test net output #13: loss1/accuracy13 = 0.609 | |
| I0623 16:24:05.061627 10365 solver.cpp:406] Test net output #14: loss1/accuracy14 = 0.684 | |
| I0623 16:24:05.061640 10365 solver.cpp:406] Test net output #15: loss1/accuracy15 = 0.777 | |
| I0623 16:24:05.061650 10365 solver.cpp:406] Test net output #16: loss1/accuracy16 = 0.83 | |
| I0623 16:24:05.061662 10365 solver.cpp:406] Test net output #17: loss1/accuracy17 = 0.906 | |
| I0623 16:24:05.061673 10365 solver.cpp:406] Test net output #18: loss1/accuracy18 = 0.952 | |
| I0623 16:24:05.061684 10365 solver.cpp:406] Test net output #19: loss1/accuracy19 = 0.972 | |
| I0623 16:24:05.061696 10365 solver.cpp:406] Test net output #20: loss1/accuracy20 = 0.987 | |
| I0623 16:24:05.061707 10365 solver.cpp:406] Test net output #21: loss1/accuracy21 = 0.999 | |
| I0623 16:24:05.061718 10365 solver.cpp:406] Test net output #22: loss1/accuracy22 = 1 | |
| I0623 16:24:05.061729 10365 solver.cpp:406] Test net output #23: loss1/accuracy_incl_empty = 0.712499 | |
| I0623 16:24:05.061741 10365 solver.cpp:406] Test net output #24: loss1/accuracy_top3 = 0.851939 | |
| I0623 16:24:05.061756 10365 solver.cpp:406] Test net output #25: loss1/cross_entropy_loss = 1.37422 (* 0.3 = 0.412265 loss) | |
| I0623 16:24:05.061774 10365 solver.cpp:406] Test net output #26: loss1/cross_entropy_loss_incl_empty = 0.847342 (* 0.3 = 0.254202 loss) | |
| I0623 16:24:05.061802 10365 solver.cpp:406] Test net output #27: loss1/loss01 = 0.319289 (* 0.0272727 = 0.00870787 loss) | |
| I0623 16:24:05.061823 10365 solver.cpp:406] Test net output #28: loss1/loss02 = 0.738935 (* 0.0272727 = 0.0201528 loss) | |
| I0623 16:24:05.061838 10365 solver.cpp:406] Test net output #29: loss1/loss03 = 1.39047 (* 0.0272727 = 0.037922 loss) | |
| I0623 16:24:05.061851 10365 solver.cpp:406] Test net output #30: loss1/loss04 = 1.5678 (* 0.0272727 = 0.0427581 loss) | |
| I0623 16:24:05.061866 10365 solver.cpp:406] Test net output #31: loss1/loss05 = 1.71954 (* 0.0272727 = 0.0468965 loss) | |
| I0623 16:24:05.061879 10365 solver.cpp:406] Test net output #32: loss1/loss06 = 1.79468 (* 0.0272727 = 0.0489457 loss) | |
| I0623 16:24:05.061892 10365 solver.cpp:406] Test net output #33: loss1/loss07 = 1.82852 (* 0.0272727 = 0.0498688 loss) | |
| I0623 16:24:05.061906 10365 solver.cpp:406] Test net output #34: loss1/loss08 = 1.64946 (* 0.0272727 = 0.0449853 loss) | |
| I0623 16:24:05.061919 10365 solver.cpp:406] Test net output #35: loss1/loss09 = 1.73591 (* 0.0272727 = 0.047343 loss) | |
| I0623 16:24:05.061933 10365 solver.cpp:406] Test net output #36: loss1/loss10 = 1.77777 (* 0.0272727 = 0.0484846 loss) | |
| I0623 16:24:05.061946 10365 solver.cpp:406] Test net output #37: loss1/loss11 = 1.83573 (* 0.0272727 = 0.0500655 loss) | |
| I0623 16:24:05.061959 10365 solver.cpp:406] Test net output #38: loss1/loss12 = 1.52782 (* 0.0272727 = 0.0416679 loss) | |
| I0623 16:24:05.061992 10365 solver.cpp:406] Test net output #39: loss1/loss13 = 1.22786 (* 0.0272727 = 0.0334872 loss) | |
| I0623 16:24:05.062007 10365 solver.cpp:406] Test net output #40: loss1/loss14 = 0.940703 (* 0.0272727 = 0.0256555 loss) | |
| I0623 16:24:05.062021 10365 solver.cpp:406] Test net output #41: loss1/loss15 = 0.680672 (* 0.0272727 = 0.0185638 loss) | |
| I0623 16:24:05.062034 10365 solver.cpp:406] Test net output #42: loss1/loss16 = 0.515077 (* 0.0272727 = 0.0140475 loss) | |
| I0623 16:24:05.062048 10365 solver.cpp:406] Test net output #43: loss1/loss17 = 0.319658 (* 0.0272727 = 0.00871796 loss) | |
| I0623 16:24:05.062062 10365 solver.cpp:406] Test net output #44: loss1/loss18 = 0.189945 (* 0.0272727 = 0.00518031 loss) | |
| I0623 16:24:05.062075 10365 solver.cpp:406] Test net output #45: loss1/loss19 = 0.119191 (* 0.0272727 = 0.00325066 loss) | |
| I0623 16:24:05.062089 10365 solver.cpp:406] Test net output #46: loss1/loss20 = 0.0684475 (* 0.0272727 = 0.00186675 loss) | |
| I0623 16:24:05.062103 10365 solver.cpp:406] Test net output #47: loss1/loss21 = 0.00772684 (* 0.0272727 = 0.000210732 loss) | |
| I0623 16:24:05.062116 10365 solver.cpp:406] Test net output #48: loss1/loss22 = 9.33472e-05 (* 0.0272727 = 2.54583e-06 loss) | |
| I0623 16:24:05.062129 10365 solver.cpp:406] Test net output #49: loss2/accuracy = 0.637264 | |
| I0623 16:24:05.062140 10365 solver.cpp:406] Test net output #50: loss2/accuracy01 = 0.972 | |
| I0623 16:24:05.062152 10365 solver.cpp:406] Test net output #51: loss2/accuracy02 = 0.94 | |
| I0623 16:24:05.062163 10365 solver.cpp:406] Test net output #52: loss2/accuracy03 = 0.858 | |
| I0623 16:24:05.062175 10365 solver.cpp:406] Test net output #53: loss2/accuracy04 = 0.721 | |
| I0623 16:24:05.062186 10365 solver.cpp:406] Test net output #54: loss2/accuracy05 = 0.546 | |
| I0623 16:24:05.062197 10365 solver.cpp:406] Test net output #55: loss2/accuracy06 = 0.51 | |
| I0623 16:24:05.062209 10365 solver.cpp:406] Test net output #56: loss2/accuracy07 = 0.491 | |
| I0623 16:24:05.062221 10365 solver.cpp:406] Test net output #57: loss2/accuracy08 = 0.517 | |
| I0623 16:24:05.062232 10365 solver.cpp:406] Test net output #58: loss2/accuracy09 = 0.484 | |
| I0623 16:24:05.062242 10365 solver.cpp:406] Test net output #59: loss2/accuracy10 = 0.435 | |
| I0623 16:24:05.062253 10365 solver.cpp:406] Test net output #60: loss2/accuracy11 = 0.427 | |
| I0623 16:24:05.062268 10365 solver.cpp:406] Test net output #61: loss2/accuracy12 = 0.516 | |
| I0623 16:24:05.062279 10365 solver.cpp:406] Test net output #62: loss2/accuracy13 = 0.615 | |
| I0623 16:24:05.062291 10365 solver.cpp:406] Test net output #63: loss2/accuracy14 = 0.707 | |
| I0623 16:24:05.062302 10365 solver.cpp:406] Test net output #64: loss2/accuracy15 = 0.788 | |
| I0623 16:24:05.062314 10365 solver.cpp:406] Test net output #65: loss2/accuracy16 = 0.842 | |
| I0623 16:24:05.062325 10365 solver.cpp:406] Test net output #66: loss2/accuracy17 = 0.906 | |
| I0623 16:24:05.062336 10365 solver.cpp:406] Test net output #67: loss2/accuracy18 = 0.95 | |
| I0623 16:24:05.062347 10365 solver.cpp:406] Test net output #68: loss2/accuracy19 = 0.972 | |
| I0623 16:24:05.062358 10365 solver.cpp:406] Test net output #69: loss2/accuracy20 = 0.987 | |
| I0623 16:24:05.062371 10365 solver.cpp:406] Test net output #70: loss2/accuracy21 = 0.999 | |
| I0623 16:24:05.062381 10365 solver.cpp:406] Test net output #71: loss2/accuracy22 = 1 | |
| I0623 16:24:05.062392 10365 solver.cpp:406] Test net output #72: loss2/accuracy_incl_empty = 0.770228 | |
| I0623 16:24:05.062403 10365 solver.cpp:406] Test net output #73: loss2/accuracy_top3 = 0.911242 | |
| I0623 16:24:05.062417 10365 solver.cpp:406] Test net output #74: loss2/cross_entropy_loss = 1.05466 (* 0.3 = 0.316399 loss) | |
| I0623 16:24:05.062432 10365 solver.cpp:406] Test net output #75: loss2/cross_entropy_loss_incl_empty = 0.661178 (* 0.3 = 0.198353 loss) | |
| I0623 16:24:05.062445 10365 solver.cpp:406] Test net output #76: loss2/loss01 = 0.19347 (* 0.0272727 = 0.00527646 loss) | |
| I0623 16:24:05.062459 10365 solver.cpp:406] Test net output #77: loss2/loss02 = 0.288499 (* 0.0272727 = 0.00786815 loss) | |
| I0623 16:24:05.062490 10365 solver.cpp:406] Test net output #78: loss2/loss03 = 0.581203 (* 0.0272727 = 0.015851 loss) | |
| I0623 16:24:05.062501 10365 solver.cpp:406] Test net output #79: loss2/loss04 = 0.931626 (* 0.0272727 = 0.025408 loss) | |
| I0623 16:24:05.062511 10365 solver.cpp:406] Test net output #80: loss2/loss05 = 1.22226 (* 0.0272727 = 0.0333343 loss) | |
| I0623 16:24:05.062525 10365 solver.cpp:406] Test net output #81: loss2/loss06 = 1.4445 (* 0.0272727 = 0.0393956 loss) | |
| I0623 16:24:05.062538 10365 solver.cpp:406] Test net output #82: loss2/loss07 = 1.53993 (* 0.0272727 = 0.0419982 loss) | |
| I0623 16:24:05.062551 10365 solver.cpp:406] Test net output #83: loss2/loss08 = 1.45799 (* 0.0272727 = 0.0397633 loss) | |
| I0623 16:24:05.062566 10365 solver.cpp:406] Test net output #84: loss2/loss09 = 1.53595 (* 0.0272727 = 0.0418895 loss) | |
| I0623 16:24:05.062578 10365 solver.cpp:406] Test net output #85: loss2/loss10 = 1.61512 (* 0.0272727 = 0.0440486 loss) | |
| I0623 16:24:05.062592 10365 solver.cpp:406] Test net output #86: loss2/loss11 = 1.65639 (* 0.0272727 = 0.0451743 loss) | |
| I0623 16:24:05.062605 10365 solver.cpp:406] Test net output #87: loss2/loss12 = 1.35708 (* 0.0272727 = 0.0370113 loss) | |
| I0623 16:24:05.062618 10365 solver.cpp:406] Test net output #88: loss2/loss13 = 1.11395 (* 0.0272727 = 0.0303803 loss) | |
| I0623 16:24:05.062633 10365 solver.cpp:406] Test net output #89: loss2/loss14 = 0.849185 (* 0.0272727 = 0.0231596 loss) | |
| I0623 16:24:05.062645 10365 solver.cpp:406] Test net output #90: loss2/loss15 = 0.620886 (* 0.0272727 = 0.0169333 loss) | |
| I0623 16:24:05.062659 10365 solver.cpp:406] Test net output #91: loss2/loss16 = 0.460988 (* 0.0272727 = 0.0125724 loss) | |
| I0623 16:24:05.062671 10365 solver.cpp:406] Test net output #92: loss2/loss17 = 0.303104 (* 0.0272727 = 0.00826646 loss) | |
| I0623 16:24:05.062685 10365 solver.cpp:406] Test net output #93: loss2/loss18 = 0.169109 (* 0.0272727 = 0.00461205 loss) | |
| I0623 16:24:05.062698 10365 solver.cpp:406] Test net output #94: loss2/loss19 = 0.107455 (* 0.0272727 = 0.00293058 loss) | |
| I0623 16:24:05.062711 10365 solver.cpp:406] Test net output #95: loss2/loss20 = 0.0638529 (* 0.0272727 = 0.00174144 loss) | |
| I0623 16:24:05.062726 10365 solver.cpp:406] Test net output #96: loss2/loss21 = 0.00766872 (* 0.0272727 = 0.000209147 loss) | |
| I0623 16:24:05.062738 10365 solver.cpp:406] Test net output #97: loss2/loss22 = 9.43917e-05 (* 0.0272727 = 2.57432e-06 loss) | |
| I0623 16:24:05.062750 10365 solver.cpp:406] Test net output #98: loss3/accuracy = 0.868335 | |
| I0623 16:24:05.062762 10365 solver.cpp:406] Test net output #99: loss3/accuracy01 = 0.978 | |
| I0623 16:24:05.062773 10365 solver.cpp:406] Test net output #100: loss3/accuracy02 = 0.969 | |
| I0623 16:24:05.062784 10365 solver.cpp:406] Test net output #101: loss3/accuracy03 = 0.955 | |
| I0623 16:24:05.062795 10365 solver.cpp:406] Test net output #102: loss3/accuracy04 = 0.941 | |
| I0623 16:24:05.062808 10365 solver.cpp:406] Test net output #103: loss3/accuracy05 = 0.936 | |
| I0623 16:24:05.062819 10365 solver.cpp:406] Test net output #104: loss3/accuracy06 = 0.914 | |
| I0623 16:24:05.062829 10365 solver.cpp:406] Test net output #105: loss3/accuracy07 = 0.904 | |
| I0623 16:24:05.062839 10365 solver.cpp:406] Test net output #106: loss3/accuracy08 = 0.879 | |
| I0623 16:24:05.062851 10365 solver.cpp:406] Test net output #107: loss3/accuracy09 = 0.788 | |
| I0623 16:24:05.062861 10365 solver.cpp:406] Test net output #108: loss3/accuracy10 = 0.663 | |
| I0623 16:24:05.062872 10365 solver.cpp:406] Test net output #109: loss3/accuracy11 = 0.594 | |
| I0623 16:24:05.062883 10365 solver.cpp:406] Test net output #110: loss3/accuracy12 = 0.645 | |
| I0623 16:24:05.062894 10365 solver.cpp:406] Test net output #111: loss3/accuracy13 = 0.68 | |
| I0623 16:24:05.062906 10365 solver.cpp:406] Test net output #112: loss3/accuracy14 = 0.757 | |
| I0623 16:24:05.062916 10365 solver.cpp:406] Test net output #113: loss3/accuracy15 = 0.831 | |
| I0623 16:24:05.062927 10365 solver.cpp:406] Test net output #114: loss3/accuracy16 = 0.876 | |
| I0623 16:24:05.062948 10365 solver.cpp:406] Test net output #115: loss3/accuracy17 = 0.928 | |
| I0623 16:24:05.062961 10365 solver.cpp:406] Test net output #116: loss3/accuracy18 = 0.963 | |
| I0623 16:24:05.062973 10365 solver.cpp:406] Test net output #117: loss3/accuracy19 = 0.975 | |
| I0623 16:24:05.062984 10365 solver.cpp:406] Test net output #118: loss3/accuracy20 = 0.987 | |
| I0623 16:24:05.062995 10365 solver.cpp:406] Test net output #119: loss3/accuracy21 = 0.999 | |
| I0623 16:24:05.063006 10365 solver.cpp:406] Test net output #120: loss3/accuracy22 = 1 | |
| I0623 16:24:05.063017 10365 solver.cpp:406] Test net output #121: loss3/accuracy_incl_empty = 0.911228 | |
| I0623 16:24:05.063030 10365 solver.cpp:406] Test net output #122: loss3/accuracy_top3 = 0.969168 | |
| I0623 16:24:05.063042 10365 solver.cpp:406] Test net output #123: loss3/cross_entropy_loss = 0.530596 (* 1 = 0.530596 loss) | |
| I0623 16:24:05.063056 10365 solver.cpp:406] Test net output #124: loss3/cross_entropy_loss_incl_empty = 0.340269 (* 1 = 0.340269 loss) | |
| I0623 16:24:05.063071 10365 solver.cpp:406] Test net output #125: loss3/loss01 = 0.149585 (* 0.0909091 = 0.0135987 loss) | |
| I0623 16:24:05.063084 10365 solver.cpp:406] Test net output #126: loss3/loss02 = 0.183549 (* 0.0909091 = 0.0166863 loss) | |
| I0623 16:24:05.063098 10365 solver.cpp:406] Test net output #127: loss3/loss03 = 0.3055 (* 0.0909091 = 0.0277728 loss) | |
| I0623 16:24:05.063112 10365 solver.cpp:406] Test net output #128: loss3/loss04 = 0.360835 (* 0.0909091 = 0.0328032 loss) | |
| I0623 16:24:05.063125 10365 solver.cpp:406] Test net output #129: loss3/loss05 = 0.360737 (* 0.0909091 = 0.0327942 loss) | |
| I0623 16:24:05.063139 10365 solver.cpp:406] Test net output #130: loss3/loss06 = 0.467246 (* 0.0909091 = 0.0424769 loss) | |
| I0623 16:24:05.063153 10365 solver.cpp:406] Test net output #131: loss3/loss07 = 0.517786 (* 0.0909091 = 0.0470714 loss) | |
| I0623 16:24:05.063166 10365 solver.cpp:406] Test net output #132: loss3/loss08 = 0.538317 (* 0.0909091 = 0.0489379 loss) | |
| I0623 16:24:05.063179 10365 solver.cpp:406] Test net output #133: loss3/loss09 = 0.72638 (* 0.0909091 = 0.0660346 loss) | |
| I0623 16:24:05.063194 10365 solver.cpp:406] Test net output #134: loss3/loss10 = 0.980019 (* 0.0909091 = 0.0890927 loss) | |
| I0623 16:24:05.063206 10365 solver.cpp:406] Test net output #135: loss3/loss11 = 1.12027 (* 0.0909091 = 0.101843 loss) | |
| I0623 16:24:05.063220 10365 solver.cpp:406] Test net output #136: loss3/loss12 = 0.958663 (* 0.0909091 = 0.0871511 loss) | |
| I0623 16:24:05.063233 10365 solver.cpp:406] Test net output #137: loss3/loss13 = 0.851221 (* 0.0909091 = 0.0773838 loss) | |
| I0623 16:24:05.063247 10365 solver.cpp:406] Test net output #138: loss3/loss14 = 0.649027 (* 0.0909091 = 0.0590024 loss) | |
| I0623 16:24:05.063261 10365 solver.cpp:406] Test net output #139: loss3/loss15 = 0.471805 (* 0.0909091 = 0.0428913 loss) | |
| I0623 16:24:05.063274 10365 solver.cpp:406] Test net output #140: loss3/loss16 = 0.362535 (* 0.0909091 = 0.0329578 loss) | |
| I0623 16:24:05.063288 10365 solver.cpp:406] Test net output #141: loss3/loss17 = 0.20972 (* 0.0909091 = 0.0190655 loss) | |
| I0623 16:24:05.063302 10365 solver.cpp:406] Test net output #142: loss3/loss18 = 0.125947 (* 0.0909091 = 0.0114497 loss) | |
| I0623 16:24:05.063318 10365 solver.cpp:406] Test net output #143: loss3/loss19 = 0.0714735 (* 0.0909091 = 0.00649759 loss) | |
| I0623 16:24:05.063333 10365 solver.cpp:406] Test net output #144: loss3/loss20 = 0.0413603 (* 0.0909091 = 0.00376003 loss) | |
| I0623 16:24:05.063347 10365 solver.cpp:406] Test net output #145: loss3/loss21 = 0.00622251 (* 0.0909091 = 0.000565683 loss) | |
| I0623 16:24:05.063361 10365 solver.cpp:406] Test net output #146: loss3/loss22 = 0.000112198 (* 0.0909091 = 1.01998e-05 loss) | |
| I0623 16:24:05.063374 10365 solver.cpp:406] Test net output #147: total_accuracy = 0.42 | |
| I0623 16:24:05.063385 10365 solver.cpp:406] Test net output #148: total_accuracy_not_rec = 0.223 | |
| I0623 16:24:05.063396 10365 solver.cpp:406] Test net output #149: total_confidence = 0.221943 | |
| I0623 16:24:05.063418 10365 solver.cpp:406] Test net output #150: total_confidence_not_rec = 0.141333 | |
| I0623 16:24:05.422760 10365 solver.cpp:229] Iteration 10000, loss = 4.57648 | |
| I0623 16:24:05.422819 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.458716 | |
| I0623 16:24:05.422837 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 16:24:05.422850 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.375 | |
| I0623 16:24:05.422864 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.625 | |
| I0623 16:24:05.422878 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 16:24:05.422890 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 16:24:05.422904 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 16:24:05.422916 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.625 | |
| I0623 16:24:05.422930 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.25 | |
| I0623 16:24:05.422941 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 16:24:05.422955 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 16:24:05.422966 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 16:24:05.422979 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.125 | |
| I0623 16:24:05.422991 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 16:24:05.423003 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.375 | |
| I0623 16:24:05.423015 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 16:24:05.423027 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 16:24:05.423038 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 16:24:05.423049 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 16:24:05.423061 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:24:05.423074 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:24:05.423085 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:24:05.423097 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:24:05.423110 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.659091 | |
| I0623 16:24:05.423122 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.788991 | |
| I0623 16:24:05.423141 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.74489 (* 0.3 = 0.523467 loss) | |
| I0623 16:24:05.423156 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.11388 (* 0.3 = 0.334163 loss) | |
| I0623 16:24:05.423171 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.450962 (* 0.0272727 = 0.012299 loss) | |
| I0623 16:24:05.423185 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 2.14368 (* 0.0272727 = 0.0584641 loss) | |
| I0623 16:24:05.423199 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.19577 (* 0.0272727 = 0.0326118 loss) | |
| I0623 16:24:05.423213 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.6977 (* 0.0272727 = 0.046301 loss) | |
| I0623 16:24:05.423228 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.16535 (* 0.0272727 = 0.059055 loss) | |
| I0623 16:24:05.423241 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.22471 (* 0.0272727 = 0.060674 loss) | |
| I0623 16:24:05.423255 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.68133 (* 0.0272727 = 0.0458546 loss) | |
| I0623 16:24:05.423269 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.23312 (* 0.0272727 = 0.0609032 loss) | |
| I0623 16:24:05.423283 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.2657 (* 0.0272727 = 0.0617918 loss) | |
| I0623 16:24:05.423296 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.32461 (* 0.0272727 = 0.0633983 loss) | |
| I0623 16:24:05.423310 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.16664 (* 0.0272727 = 0.0590902 loss) | |
| I0623 16:24:05.423351 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.58286 (* 0.0272727 = 0.0704418 loss) | |
| I0623 16:24:05.423365 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.32545 (* 0.0272727 = 0.0361486 loss) | |
| I0623 16:24:05.423379 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.61845 (* 0.0272727 = 0.0441396 loss) | |
| I0623 16:24:05.423393 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.795135 (* 0.0272727 = 0.0216855 loss) | |
| I0623 16:24:05.423408 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.48919 (* 0.0272727 = 0.0133415 loss) | |
| I0623 16:24:05.423421 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.299296 (* 0.0272727 = 0.00816261 loss) | |
| I0623 16:24:05.423435 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.093602 (* 0.0272727 = 0.00255278 loss) | |
| I0623 16:24:05.423449 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0230499 (* 0.0272727 = 0.000628634 loss) | |
| I0623 16:24:05.423463 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00464083 (* 0.0272727 = 0.000126568 loss) | |
| I0623 16:24:05.423477 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.00152757 (* 0.0272727 = 4.16609e-05 loss) | |
| I0623 16:24:05.423492 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000311863 (* 0.0272727 = 8.50536e-06 loss) | |
| I0623 16:24:05.423504 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.53211 | |
| I0623 16:24:05.423517 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 16:24:05.423529 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.375 | |
| I0623 16:24:05.423542 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 16:24:05.423552 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 16:24:05.423564 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 16:24:05.423576 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 16:24:05.423588 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.125 | |
| I0623 16:24:05.423617 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.25 | |
| I0623 16:24:05.423631 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 16:24:05.423643 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.125 | |
| I0623 16:24:05.423655 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.25 | |
| I0623 16:24:05.423666 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.25 | |
| I0623 16:24:05.423678 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 16:24:05.423689 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.375 | |
| I0623 16:24:05.423702 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 16:24:05.423713 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 16:24:05.423725 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 16:24:05.423738 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 16:24:05.423753 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:24:05.423763 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:24:05.423771 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:24:05.423779 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:24:05.423790 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.681818 | |
| I0623 16:24:05.423802 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.834862 | |
| I0623 16:24:05.423816 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.3249 (* 0.3 = 0.397469 loss) | |
| I0623 16:24:05.423830 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.933592 (* 0.3 = 0.280078 loss) | |
| I0623 16:24:05.423857 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.192369 (* 0.0272727 = 0.00524642 loss) | |
| I0623 16:24:05.423872 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 1.35021 (* 0.0272727 = 0.036824 loss) | |
| I0623 16:24:05.423887 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.625674 (* 0.0272727 = 0.0170638 loss) | |
| I0623 16:24:05.423900 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.923492 (* 0.0272727 = 0.0251861 loss) | |
| I0623 16:24:05.423914 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.60636 (* 0.0272727 = 0.0438098 loss) | |
| I0623 16:24:05.423928 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.63755 (* 0.0272727 = 0.0446603 loss) | |
| I0623 16:24:05.423943 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.86431 (* 0.0272727 = 0.0508448 loss) | |
| I0623 16:24:05.423956 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.07205 (* 0.0272727 = 0.0565105 loss) | |
| I0623 16:24:05.423969 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 2.21501 (* 0.0272727 = 0.0604094 loss) | |
| I0623 16:24:05.423984 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.17824 (* 0.0272727 = 0.0594066 loss) | |
| I0623 16:24:05.423996 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.16743 (* 0.0272727 = 0.0591116 loss) | |
| I0623 16:24:05.424010 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 2.01547 (* 0.0272727 = 0.0549673 loss) | |
| I0623 16:24:05.424024 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.24768 (* 0.0272727 = 0.0340278 loss) | |
| I0623 16:24:05.424037 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.63972 (* 0.0272727 = 0.0447196 loss) | |
| I0623 16:24:05.424052 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.01652 (* 0.0272727 = 0.0277234 loss) | |
| I0623 16:24:05.424065 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.477879 (* 0.0272727 = 0.0130331 loss) | |
| I0623 16:24:05.424079 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.271905 (* 0.0272727 = 0.0074156 loss) | |
| I0623 16:24:05.424093 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0926642 (* 0.0272727 = 0.00252721 loss) | |
| I0623 16:24:05.424108 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0121776 (* 0.0272727 = 0.000332116 loss) | |
| I0623 16:24:05.424121 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00288717 (* 0.0272727 = 7.87411e-05 loss) | |
| I0623 16:24:05.424135 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00115313 (* 0.0272727 = 3.14491e-05 loss) | |
| I0623 16:24:05.424149 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000201699 (* 0.0272727 = 5.50088e-06 loss) | |
| I0623 16:24:05.424161 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.807339 | |
| I0623 16:24:05.424173 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 16:24:05.424188 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 16:24:05.424199 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:24:05.424211 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 16:24:05.424222 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 16:24:05.424234 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 16:24:05.424245 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.625 | |
| I0623 16:24:05.424257 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.625 | |
| I0623 16:24:05.424269 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 16:24:05.424280 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 16:24:05.424293 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.375 | |
| I0623 16:24:05.424304 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 16:24:05.424315 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 16:24:05.424326 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 16:24:05.424350 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 16:24:05.424363 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 16:24:05.424376 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 16:24:05.424387 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 16:24:05.424398 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:24:05.424410 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:24:05.424422 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:24:05.424433 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:24:05.424444 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.875 | |
| I0623 16:24:05.424458 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.981651 | |
| I0623 16:24:05.424471 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.555443 (* 1 = 0.555443 loss) | |
| I0623 16:24:05.424485 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.368694 (* 1 = 0.368694 loss) | |
| I0623 16:24:05.424500 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0584147 (* 0.0909091 = 0.00531043 loss) | |
| I0623 16:24:05.424515 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.17155 (* 0.0909091 = 0.0155955 loss) | |
| I0623 16:24:05.424530 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0813092 (* 0.0909091 = 0.00739174 loss) | |
| I0623 16:24:05.424543 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0907369 (* 0.0909091 = 0.00824881 loss) | |
| I0623 16:24:05.424558 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.4546 (* 0.0909091 = 0.0413272 loss) | |
| I0623 16:24:05.424572 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.162305 (* 0.0909091 = 0.014755 loss) | |
| I0623 16:24:05.424587 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.630506 (* 0.0909091 = 0.0573187 loss) | |
| I0623 16:24:05.424600 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.01999 (* 0.0909091 = 0.0927261 loss) | |
| I0623 16:24:05.424613 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.692019 (* 0.0909091 = 0.0629108 loss) | |
| I0623 16:24:05.424628 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.854129 (* 0.0909091 = 0.0776481 loss) | |
| I0623 16:24:05.424641 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.3035 (* 0.0909091 = 0.1185 loss) | |
| I0623 16:24:05.424655 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.8915 (* 0.0909091 = 0.171954 loss) | |
| I0623 16:24:05.424669 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.931036 (* 0.0909091 = 0.0846397 loss) | |
| I0623 16:24:05.424682 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.14917 (* 0.0909091 = 0.10447 loss) | |
| I0623 16:24:05.424696 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.19928 (* 0.0909091 = 0.109025 loss) | |
| I0623 16:24:05.424710 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.234849 (* 0.0909091 = 0.0213499 loss) | |
| I0623 16:24:05.424724 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.21675 (* 0.0909091 = 0.0197046 loss) | |
| I0623 16:24:05.424738 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0395051 (* 0.0909091 = 0.00359137 loss) | |
| I0623 16:24:05.424752 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00157899 (* 0.0909091 = 0.000143545 loss) | |
| I0623 16:24:05.424767 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000821444 (* 0.0909091 = 7.46767e-05 loss) | |
| I0623 16:24:05.424780 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000147607 (* 0.0909091 = 1.34188e-05 loss) | |
| I0623 16:24:05.424799 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 1.20999e-05 (* 0.0909091 = 1.09999e-06 loss) | |
| I0623 16:24:05.424823 10365 solver.cpp:245] Train net output #147: total_accuracy = 0 | |
| I0623 16:24:05.424836 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 16:24:05.424849 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0240825 | |
| I0623 16:24:05.424860 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.00388674 | |
| I0623 16:24:05.424873 10365 sgd_solver.cpp:106] Iteration 10000, lr = 0.001 | |
| I0623 16:25:06.311198 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.5416 > 30) by scale factor 0.820983 | |
| I0623 16:30:28.603117 10365 solver.cpp:229] Iteration 10500, loss = 4.64449 | |
| I0623 16:30:28.603221 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.462366 | |
| I0623 16:30:28.603241 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 16:30:28.603255 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 16:30:28.603267 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 16:30:28.603279 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 16:30:28.603292 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.625 | |
| I0623 16:30:28.603305 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 16:30:28.603318 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 16:30:28.603330 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 16:30:28.603343 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 16:30:28.603355 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 16:30:28.603368 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.125 | |
| I0623 16:30:28.603380 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.75 | |
| I0623 16:30:28.603392 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 16:30:28.603404 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 16:30:28.603415 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 16:30:28.603427 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 16:30:28.603440 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 16:30:28.603451 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 16:30:28.603463 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:30:28.603474 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:30:28.603485 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:30:28.603497 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:30:28.603509 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.6875 | |
| I0623 16:30:28.603520 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.784946 | |
| I0623 16:30:28.603538 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.53317 (* 0.3 = 0.459952 loss) | |
| I0623 16:30:28.603551 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.892533 (* 0.3 = 0.26776 loss) | |
| I0623 16:30:28.603566 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.400655 (* 0.0272727 = 0.010927 loss) | |
| I0623 16:30:28.603580 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.675188 (* 0.0272727 = 0.0184142 loss) | |
| I0623 16:30:28.603593 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.46454 (* 0.0272727 = 0.0399419 loss) | |
| I0623 16:30:28.603623 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.71894 (* 0.0272727 = 0.0468801 loss) | |
| I0623 16:30:28.603639 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.59532 (* 0.0272727 = 0.0435089 loss) | |
| I0623 16:30:28.603654 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.83153 (* 0.0272727 = 0.0499509 loss) | |
| I0623 16:30:28.603668 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.80553 (* 0.0272727 = 0.0492416 loss) | |
| I0623 16:30:28.603682 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.15216 (* 0.0272727 = 0.0586954 loss) | |
| I0623 16:30:28.603696 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.86782 (* 0.0272727 = 0.0509405 loss) | |
| I0623 16:30:28.603710 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.61111 (* 0.0272727 = 0.0439394 loss) | |
| I0623 16:30:28.603724 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.00333 (* 0.0272727 = 0.0546363 loss) | |
| I0623 16:30:28.603739 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.59528 (* 0.0272727 = 0.0435077 loss) | |
| I0623 16:30:28.603771 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.878454 (* 0.0272727 = 0.0239578 loss) | |
| I0623 16:30:28.603786 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.865628 (* 0.0272727 = 0.023608 loss) | |
| I0623 16:30:28.603801 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.04104 (* 0.0272727 = 0.0283919 loss) | |
| I0623 16:30:28.603814 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.612728 (* 0.0272727 = 0.0167108 loss) | |
| I0623 16:30:28.603828 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0348289 (* 0.0272727 = 0.00094988 loss) | |
| I0623 16:30:28.603842 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00743639 (* 0.0272727 = 0.000202811 loss) | |
| I0623 16:30:28.603857 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00197824 (* 0.0272727 = 5.39519e-05 loss) | |
| I0623 16:30:28.603871 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000513801 (* 0.0272727 = 1.40128e-05 loss) | |
| I0623 16:30:28.603886 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000126422 (* 0.0272727 = 3.44788e-06 loss) | |
| I0623 16:30:28.603900 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.6719e-05 (* 0.0272727 = 1.00143e-06 loss) | |
| I0623 16:30:28.603912 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.591398 | |
| I0623 16:30:28.603925 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 16:30:28.603937 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 16:30:28.603948 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 16:30:28.603960 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 16:30:28.603972 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 16:30:28.603983 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 16:30:28.603996 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 16:30:28.604007 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 16:30:28.604018 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 16:30:28.604029 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.25 | |
| I0623 16:30:28.604041 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 16:30:28.604053 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 16:30:28.604064 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 16:30:28.604075 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 16:30:28.604087 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 16:30:28.604099 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 16:30:28.604110 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 16:30:28.604121 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 16:30:28.604133 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:30:28.604145 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:30:28.604156 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:30:28.604167 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:30:28.604178 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.772727 | |
| I0623 16:30:28.604197 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.88172 | |
| I0623 16:30:28.604210 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.19882 (* 0.3 = 0.359645 loss) | |
| I0623 16:30:28.604224 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.665486 (* 0.3 = 0.199646 loss) | |
| I0623 16:30:28.604238 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.284842 (* 0.0272727 = 0.00776843 loss) | |
| I0623 16:30:28.604252 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.0837156 (* 0.0272727 = 0.00228315 loss) | |
| I0623 16:30:28.604277 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.46297 (* 0.0272727 = 0.0126264 loss) | |
| I0623 16:30:28.604292 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.50091 (* 0.0272727 = 0.0409339 loss) | |
| I0623 16:30:28.604306 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.28822 (* 0.0272727 = 0.0351333 loss) | |
| I0623 16:30:28.604321 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.40879 (* 0.0272727 = 0.0384215 loss) | |
| I0623 16:30:28.604334 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.7337 (* 0.0272727 = 0.0472826 loss) | |
| I0623 16:30:28.604348 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.04861 (* 0.0272727 = 0.0558712 loss) | |
| I0623 16:30:28.604362 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.42889 (* 0.0272727 = 0.0389699 loss) | |
| I0623 16:30:28.604375 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.28764 (* 0.0272727 = 0.0351174 loss) | |
| I0623 16:30:28.604389 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.423 (* 0.0272727 = 0.038809 loss) | |
| I0623 16:30:28.604403 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.13419 (* 0.0272727 = 0.0309325 loss) | |
| I0623 16:30:28.604416 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.714474 (* 0.0272727 = 0.0194857 loss) | |
| I0623 16:30:28.604430 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.871151 (* 0.0272727 = 0.0237587 loss) | |
| I0623 16:30:28.604444 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.781869 (* 0.0272727 = 0.0213237 loss) | |
| I0623 16:30:28.604459 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.516202 (* 0.0272727 = 0.0140782 loss) | |
| I0623 16:30:28.604471 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0160319 (* 0.0272727 = 0.000437233 loss) | |
| I0623 16:30:28.604485 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0019126 (* 0.0272727 = 5.21618e-05 loss) | |
| I0623 16:30:28.604501 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000800948 (* 0.0272727 = 2.1844e-05 loss) | |
| I0623 16:30:28.604514 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000231675 (* 0.0272727 = 6.3184e-06 loss) | |
| I0623 16:30:28.604528 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000181545 (* 0.0272727 = 4.95122e-06 loss) | |
| I0623 16:30:28.604539 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000146656 (* 0.0272727 = 3.99971e-06 loss) | |
| I0623 16:30:28.604547 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.827957 | |
| I0623 16:30:28.604560 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 16:30:28.604573 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 16:30:28.604584 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:30:28.604596 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 16:30:28.604607 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 16:30:28.604619 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 16:30:28.604629 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 16:30:28.604640 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 16:30:28.604651 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 16:30:28.604663 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 16:30:28.604674 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 16:30:28.604686 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.75 | |
| I0623 16:30:28.604697 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 1 | |
| I0623 16:30:28.604708 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 1 | |
| I0623 16:30:28.604719 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 16:30:28.604730 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 16:30:28.604751 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 16:30:28.604764 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 16:30:28.604776 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:30:28.604787 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:30:28.604799 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:30:28.604810 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:30:28.604821 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.903409 | |
| I0623 16:30:28.604832 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 1 | |
| I0623 16:30:28.604846 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.424685 (* 1 = 0.424685 loss) | |
| I0623 16:30:28.604859 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.242343 (* 1 = 0.242343 loss) | |
| I0623 16:30:28.604873 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.023922 (* 0.0909091 = 0.00217473 loss) | |
| I0623 16:30:28.604887 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0277651 (* 0.0909091 = 0.0025241 loss) | |
| I0623 16:30:28.604902 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0849593 (* 0.0909091 = 0.00772357 loss) | |
| I0623 16:30:28.604915 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.057655 (* 0.0909091 = 0.00524136 loss) | |
| I0623 16:30:28.604929 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.0723471 (* 0.0909091 = 0.00657701 loss) | |
| I0623 16:30:28.604943 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.127385 (* 0.0909091 = 0.0115805 loss) | |
| I0623 16:30:28.604956 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.179361 (* 0.0909091 = 0.0163056 loss) | |
| I0623 16:30:28.604969 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.292653 (* 0.0909091 = 0.0266048 loss) | |
| I0623 16:30:28.604982 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.809077 (* 0.0909091 = 0.0735524 loss) | |
| I0623 16:30:28.604996 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.705206 (* 0.0909091 = 0.0641097 loss) | |
| I0623 16:30:28.605010 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.786393 (* 0.0909091 = 0.0714903 loss) | |
| I0623 16:30:28.605023 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.581398 (* 0.0909091 = 0.0528544 loss) | |
| I0623 16:30:28.605036 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.338749 (* 0.0909091 = 0.0307954 loss) | |
| I0623 16:30:28.605051 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.292125 (* 0.0909091 = 0.0265568 loss) | |
| I0623 16:30:28.605064 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.411417 (* 0.0909091 = 0.0374015 loss) | |
| I0623 16:30:28.605077 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.204928 (* 0.0909091 = 0.0186298 loss) | |
| I0623 16:30:28.605092 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0108389 (* 0.0909091 = 0.000985358 loss) | |
| I0623 16:30:28.605105 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000966391 (* 0.0909091 = 8.78537e-05 loss) | |
| I0623 16:30:28.605119 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000126899 (* 0.0909091 = 1.15363e-05 loss) | |
| I0623 16:30:28.605134 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 5.63329e-05 (* 0.0909091 = 5.12118e-06 loss) | |
| I0623 16:30:28.605147 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 1.94916e-05 (* 0.0909091 = 1.77196e-06 loss) | |
| I0623 16:30:28.605161 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 2.32024e-05 (* 0.0909091 = 2.10931e-06 loss) | |
| I0623 16:30:28.605175 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 16:30:28.605185 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 16:30:28.605197 10365 solver.cpp:245] Train net output #149: total_confidence = 0.210835 | |
| I0623 16:30:28.605219 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.147144 | |
| I0623 16:30:28.605239 10365 sgd_solver.cpp:106] Iteration 10500, lr = 0.001 | |
| I0623 16:30:48.127117 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.4909 > 30) by scale factor 0.9839 | |
| I0623 16:32:03.246161 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.4885 > 30) by scale factor 0.822176 | |
| I0623 16:36:51.758781 10365 solver.cpp:229] Iteration 11000, loss = 4.52815 | |
| I0623 16:36:51.758922 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.4625 | |
| I0623 16:36:51.758944 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 16:36:51.758956 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 16:36:51.758970 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 16:36:51.758982 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 16:36:51.758994 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 16:36:51.759007 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.75 | |
| I0623 16:36:51.759019 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 16:36:51.759032 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 1 | |
| I0623 16:36:51.759045 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.625 | |
| I0623 16:36:51.759057 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 16:36:51.759070 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 1 | |
| I0623 16:36:51.759083 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 16:36:51.759094 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.875 | |
| I0623 16:36:51.759105 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.875 | |
| I0623 16:36:51.759117 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 16:36:51.759130 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 16:36:51.759141 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 16:36:51.759152 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 16:36:51.759165 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:36:51.759176 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:36:51.759187 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:36:51.759198 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:36:51.759210 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.744318 | |
| I0623 16:36:51.759222 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.8125 | |
| I0623 16:36:51.759239 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.4564 (* 0.3 = 0.436919 loss) | |
| I0623 16:36:51.759253 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.692674 (* 0.3 = 0.207802 loss) | |
| I0623 16:36:51.759271 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.439044 (* 0.0272727 = 0.0119739 loss) | |
| I0623 16:36:51.759286 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.904371 (* 0.0272727 = 0.0246647 loss) | |
| I0623 16:36:51.759299 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.01481 (* 0.0272727 = 0.0276767 loss) | |
| I0623 16:36:51.759313 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.17318 (* 0.0272727 = 0.0592686 loss) | |
| I0623 16:36:51.759327 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.44038 (* 0.0272727 = 0.0392831 loss) | |
| I0623 16:36:51.759341 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.11727 (* 0.0272727 = 0.030471 loss) | |
| I0623 16:36:51.759356 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.25985 (* 0.0272727 = 0.0343597 loss) | |
| I0623 16:36:51.759368 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 0.402099 (* 0.0272727 = 0.0109663 loss) | |
| I0623 16:36:51.759382 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.61591 (* 0.0272727 = 0.0440701 loss) | |
| I0623 16:36:51.759397 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.20607 (* 0.0272727 = 0.0328928 loss) | |
| I0623 16:36:51.759410 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 0.869026 (* 0.0272727 = 0.0237007 loss) | |
| I0623 16:36:51.759423 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.25632 (* 0.0272727 = 0.0342632 loss) | |
| I0623 16:36:51.759455 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.563433 (* 0.0272727 = 0.0153664 loss) | |
| I0623 16:36:51.759470 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.594583 (* 0.0272727 = 0.0162159 loss) | |
| I0623 16:36:51.759485 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.693661 (* 0.0272727 = 0.018918 loss) | |
| I0623 16:36:51.759500 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.867401 (* 0.0272727 = 0.0236564 loss) | |
| I0623 16:36:51.759513 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0174819 (* 0.0272727 = 0.000476778 loss) | |
| I0623 16:36:51.759527 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00554597 (* 0.0272727 = 0.000151254 loss) | |
| I0623 16:36:51.759541 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00145996 (* 0.0272727 = 3.98172e-05 loss) | |
| I0623 16:36:51.759555 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000469959 (* 0.0272727 = 1.28171e-05 loss) | |
| I0623 16:36:51.759569 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000473097 (* 0.0272727 = 1.29027e-05 loss) | |
| I0623 16:36:51.759583 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000149324 (* 0.0272727 = 4.07247e-06 loss) | |
| I0623 16:36:51.759606 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.625 | |
| I0623 16:36:51.759621 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 16:36:51.759634 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 16:36:51.759645 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 16:36:51.759656 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 1 | |
| I0623 16:36:51.759668 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 16:36:51.759680 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 16:36:51.759690 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.75 | |
| I0623 16:36:51.759702 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 1 | |
| I0623 16:36:51.759713 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 16:36:51.759724 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 16:36:51.759737 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 16:36:51.759747 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.875 | |
| I0623 16:36:51.759758 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 16:36:51.759769 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 16:36:51.759780 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 16:36:51.759793 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 16:36:51.759804 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 16:36:51.759814 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 16:36:51.759825 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:36:51.759836 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:36:51.759848 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:36:51.759860 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:36:51.759871 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.8125 | |
| I0623 16:36:51.759881 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.85 | |
| I0623 16:36:51.759896 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.14542 (* 0.3 = 0.343625 loss) | |
| I0623 16:36:51.759908 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.561101 (* 0.3 = 0.16833 loss) | |
| I0623 16:36:51.759922 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.0185273 (* 0.0272727 = 0.000505291 loss) | |
| I0623 16:36:51.759939 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.641501 (* 0.0272727 = 0.0174955 loss) | |
| I0623 16:36:51.759966 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.441964 (* 0.0272727 = 0.0120536 loss) | |
| I0623 16:36:51.759981 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.588651 (* 0.0272727 = 0.0160541 loss) | |
| I0623 16:36:51.759994 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.01523 (* 0.0272727 = 0.027688 loss) | |
| I0623 16:36:51.760009 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.34773 (* 0.0272727 = 0.0367562 loss) | |
| I0623 16:36:51.760021 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.09926 (* 0.0272727 = 0.0299798 loss) | |
| I0623 16:36:51.760035 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 0.311083 (* 0.0272727 = 0.00848408 loss) | |
| I0623 16:36:51.760049 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.3194 (* 0.0272727 = 0.0359836 loss) | |
| I0623 16:36:51.760063 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.60414 (* 0.0272727 = 0.0437492 loss) | |
| I0623 16:36:51.760076 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.07306 (* 0.0272727 = 0.0292652 loss) | |
| I0623 16:36:51.760090 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 0.810996 (* 0.0272727 = 0.0221181 loss) | |
| I0623 16:36:51.760104 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.421194 (* 0.0272727 = 0.0114871 loss) | |
| I0623 16:36:51.760118 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.359911 (* 0.0272727 = 0.00981576 loss) | |
| I0623 16:36:51.760133 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.11798 (* 0.0272727 = 0.0304905 loss) | |
| I0623 16:36:51.760146 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 1.12606 (* 0.0272727 = 0.0307108 loss) | |
| I0623 16:36:51.760160 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0115497 (* 0.0272727 = 0.000314993 loss) | |
| I0623 16:36:51.760174 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00114349 (* 0.0272727 = 3.11862e-05 loss) | |
| I0623 16:36:51.760188 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000185766 (* 0.0272727 = 5.06635e-06 loss) | |
| I0623 16:36:51.760202 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 1.54383e-05 (* 0.0272727 = 4.21045e-07 loss) | |
| I0623 16:36:51.760216 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 1.83285e-06 (* 0.0272727 = 4.99869e-08 loss) | |
| I0623 16:36:51.760231 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.04308e-07 (* 0.0272727 = 2.84477e-09 loss) | |
| I0623 16:36:51.760243 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.85 | |
| I0623 16:36:51.760256 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 16:36:51.760267 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 16:36:51.760279 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:36:51.760290 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 16:36:51.760301 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 16:36:51.760315 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 16:36:51.760327 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 16:36:51.760339 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 16:36:51.760351 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 16:36:51.760362 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 16:36:51.760375 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 1 | |
| I0623 16:36:51.760386 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.875 | |
| I0623 16:36:51.760397 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 16:36:51.760409 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 16:36:51.760421 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 16:36:51.760432 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 16:36:51.760453 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 16:36:51.760467 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 16:36:51.760478 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:36:51.760489 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:36:51.760501 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:36:51.760512 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:36:51.760524 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.926136 | |
| I0623 16:36:51.760535 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.9875 | |
| I0623 16:36:51.760550 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.501931 (* 1 = 0.501931 loss) | |
| I0623 16:36:51.760563 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.245108 (* 1 = 0.245108 loss) | |
| I0623 16:36:51.760577 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0104423 (* 0.0909091 = 0.000949302 loss) | |
| I0623 16:36:51.760592 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0274545 (* 0.0909091 = 0.00249587 loss) | |
| I0623 16:36:51.760606 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0191145 (* 0.0909091 = 0.00173769 loss) | |
| I0623 16:36:51.760619 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0349495 (* 0.0909091 = 0.00317723 loss) | |
| I0623 16:36:51.760633 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.142307 (* 0.0909091 = 0.012937 loss) | |
| I0623 16:36:51.760648 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.916304 (* 0.0909091 = 0.0833003 loss) | |
| I0623 16:36:51.760661 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.242287 (* 0.0909091 = 0.0220261 loss) | |
| I0623 16:36:51.760675 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.0711855 (* 0.0909091 = 0.00647141 loss) | |
| I0623 16:36:51.760689 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.404164 (* 0.0909091 = 0.0367422 loss) | |
| I0623 16:36:51.760704 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.983387 (* 0.0909091 = 0.0893988 loss) | |
| I0623 16:36:51.760717 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.381993 (* 0.0909091 = 0.0347267 loss) | |
| I0623 16:36:51.760727 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.457131 (* 0.0909091 = 0.0415574 loss) | |
| I0623 16:36:51.760736 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.439277 (* 0.0909091 = 0.0399343 loss) | |
| I0623 16:36:51.760751 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.444153 (* 0.0909091 = 0.0403775 loss) | |
| I0623 16:36:51.760764 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.07206 (* 0.0909091 = 0.0974597 loss) | |
| I0623 16:36:51.760778 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.442986 (* 0.0909091 = 0.0402714 loss) | |
| I0623 16:36:51.760792 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0672896 (* 0.0909091 = 0.00611724 loss) | |
| I0623 16:36:51.760805 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00544372 (* 0.0909091 = 0.000494884 loss) | |
| I0623 16:36:51.760818 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00119896 (* 0.0909091 = 0.000108996 loss) | |
| I0623 16:36:51.760833 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000692605 (* 0.0909091 = 6.29641e-05 loss) | |
| I0623 16:36:51.760846 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000128231 (* 0.0909091 = 1.16574e-05 loss) | |
| I0623 16:36:51.760859 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.15747e-06 (* 0.0909091 = 3.77952e-07 loss) | |
| I0623 16:36:51.760871 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.625 | |
| I0623 16:36:51.760884 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.5 | |
| I0623 16:36:51.760895 10365 solver.cpp:245] Train net output #149: total_confidence = 0.374347 | |
| I0623 16:36:51.760915 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.347628 | |
| I0623 16:36:51.760929 10365 sgd_solver.cpp:106] Iteration 11000, lr = 0.001 | |
| I0623 16:38:56.191293 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.9704 > 30) by scale factor 0.938369 | |
| I0623 16:40:07.387783 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.7891 > 30) by scale factor 0.943721 | |
| I0623 16:40:19.644819 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.3621 > 30) by scale factor 0.988073 | |
| I0623 16:40:48.723675 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 49.211 > 30) by scale factor 0.609619 | |
| I0623 16:43:14.692836 10365 solver.cpp:229] Iteration 11500, loss = 4.64708 | |
| I0623 16:43:14.692957 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.423423 | |
| I0623 16:43:14.692978 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 16:43:14.692992 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 16:43:14.693006 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0 | |
| I0623 16:43:14.693019 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 16:43:14.693032 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.5 | |
| I0623 16:43:14.693044 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 16:43:14.693058 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.25 | |
| I0623 16:43:14.693069 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.75 | |
| I0623 16:43:14.693083 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 16:43:14.693094 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 16:43:14.693106 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 16:43:14.693120 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 16:43:14.693131 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 16:43:14.693143 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.375 | |
| I0623 16:43:14.693156 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.375 | |
| I0623 16:43:14.693166 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 16:43:14.693178 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 16:43:14.693191 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 16:43:14.693202 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:43:14.693213 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:43:14.693225 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:43:14.693236 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:43:14.693248 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.630682 | |
| I0623 16:43:14.693262 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.792793 | |
| I0623 16:43:14.693280 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.6412 (* 0.3 = 0.49236 loss) | |
| I0623 16:43:14.693295 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.06558 (* 0.3 = 0.319673 loss) | |
| I0623 16:43:14.693310 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.136911 (* 0.0272727 = 0.00373395 loss) | |
| I0623 16:43:14.693325 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.43704 (* 0.0272727 = 0.039192 loss) | |
| I0623 16:43:14.693339 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.71603 (* 0.0272727 = 0.0740735 loss) | |
| I0623 16:43:14.693353 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.71055 (* 0.0272727 = 0.0466515 loss) | |
| I0623 16:43:14.693368 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.22205 (* 0.0272727 = 0.0606014 loss) | |
| I0623 16:43:14.693382 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.29792 (* 0.0272727 = 0.0626707 loss) | |
| I0623 16:43:14.693395 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.81028 (* 0.0272727 = 0.0493714 loss) | |
| I0623 16:43:14.693409 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.46528 (* 0.0272727 = 0.0399622 loss) | |
| I0623 16:43:14.693423 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.10387 (* 0.0272727 = 0.0573782 loss) | |
| I0623 16:43:14.693437 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.95846 (* 0.0272727 = 0.0534126 loss) | |
| I0623 16:43:14.693450 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.9303 (* 0.0272727 = 0.0526445 loss) | |
| I0623 16:43:14.693464 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.12011 (* 0.0272727 = 0.0578212 loss) | |
| I0623 16:43:14.693478 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.13468 (* 0.0272727 = 0.0582184 loss) | |
| I0623 16:43:14.693509 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 2.13162 (* 0.0272727 = 0.0581351 loss) | |
| I0623 16:43:14.693526 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.50085 (* 0.0272727 = 0.0409322 loss) | |
| I0623 16:43:14.693539 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.15826 (* 0.0272727 = 0.031589 loss) | |
| I0623 16:43:14.693552 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.385213 (* 0.0272727 = 0.0105058 loss) | |
| I0623 16:43:14.693567 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0612515 (* 0.0272727 = 0.00167049 loss) | |
| I0623 16:43:14.693581 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.020963 (* 0.0272727 = 0.000571719 loss) | |
| I0623 16:43:14.693595 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00298179 (* 0.0272727 = 8.13217e-05 loss) | |
| I0623 16:43:14.693609 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000168466 (* 0.0272727 = 4.59453e-06 loss) | |
| I0623 16:43:14.693624 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.41486e-05 (* 0.0272727 = 9.31326e-07 loss) | |
| I0623 16:43:14.693636 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.54955 | |
| I0623 16:43:14.693648 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 16:43:14.693660 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 16:43:14.693671 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 16:43:14.693683 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 16:43:14.693694 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.125 | |
| I0623 16:43:14.693706 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 16:43:14.693717 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.25 | |
| I0623 16:43:14.693728 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 16:43:14.693740 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 16:43:14.693752 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 16:43:14.693763 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 16:43:14.693774 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.125 | |
| I0623 16:43:14.693785 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 16:43:14.693797 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 16:43:14.693809 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 16:43:14.693820 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.625 | |
| I0623 16:43:14.693831 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 16:43:14.693843 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 16:43:14.693855 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:43:14.693866 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:43:14.693876 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:43:14.693888 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:43:14.693899 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.704545 | |
| I0623 16:43:14.693912 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.837838 | |
| I0623 16:43:14.693925 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.40075 (* 0.3 = 0.420226 loss) | |
| I0623 16:43:14.693939 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.924138 (* 0.3 = 0.277242 loss) | |
| I0623 16:43:14.693954 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.264932 (* 0.0272727 = 0.00722542 loss) | |
| I0623 16:43:14.693969 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.553454 (* 0.0272727 = 0.0150942 loss) | |
| I0623 16:43:14.693999 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.384293 (* 0.0272727 = 0.0104807 loss) | |
| I0623 16:43:14.694013 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.914525 (* 0.0272727 = 0.0249416 loss) | |
| I0623 16:43:14.694027 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.80042 (* 0.0272727 = 0.0491023 loss) | |
| I0623 16:43:14.694041 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.31929 (* 0.0272727 = 0.0359807 loss) | |
| I0623 16:43:14.694056 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.50873 (* 0.0272727 = 0.0411472 loss) | |
| I0623 16:43:14.694069 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.2932 (* 0.0272727 = 0.0352692 loss) | |
| I0623 16:43:14.694083 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.91007 (* 0.0272727 = 0.0520929 loss) | |
| I0623 16:43:14.694097 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.55344 (* 0.0272727 = 0.0423665 loss) | |
| I0623 16:43:14.694110 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.96601 (* 0.0272727 = 0.0536184 loss) | |
| I0623 16:43:14.694124 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 2.24199 (* 0.0272727 = 0.0611451 loss) | |
| I0623 16:43:14.694139 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.78208 (* 0.0272727 = 0.0486022 loss) | |
| I0623 16:43:14.694151 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.74805 (* 0.0272727 = 0.047674 loss) | |
| I0623 16:43:14.694165 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.32136 (* 0.0272727 = 0.036037 loss) | |
| I0623 16:43:14.694178 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.885852 (* 0.0272727 = 0.0241596 loss) | |
| I0623 16:43:14.694193 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.342473 (* 0.0272727 = 0.00934017 loss) | |
| I0623 16:43:14.694207 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.035168 (* 0.0272727 = 0.000959128 loss) | |
| I0623 16:43:14.694221 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0114018 (* 0.0272727 = 0.000310958 loss) | |
| I0623 16:43:14.694236 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00491493 (* 0.0272727 = 0.000134044 loss) | |
| I0623 16:43:14.694249 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00426184 (* 0.0272727 = 0.000116232 loss) | |
| I0623 16:43:14.694262 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000318865 (* 0.0272727 = 8.69631e-06 loss) | |
| I0623 16:43:14.694274 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.810811 | |
| I0623 16:43:14.694286 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 16:43:14.694298 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 16:43:14.694309 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:43:14.694324 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 16:43:14.694335 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 16:43:14.694347 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 16:43:14.694358 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 16:43:14.694370 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 16:43:14.694381 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 16:43:14.694393 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 16:43:14.694404 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 16:43:14.694416 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.375 | |
| I0623 16:43:14.694427 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 16:43:14.694438 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.5 | |
| I0623 16:43:14.694450 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 16:43:14.694461 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 16:43:14.694488 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 16:43:14.694500 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 16:43:14.694511 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:43:14.694524 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:43:14.694535 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:43:14.694545 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:43:14.694557 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.863636 | |
| I0623 16:43:14.694568 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.963964 | |
| I0623 16:43:14.694582 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.555946 (* 1 = 0.555946 loss) | |
| I0623 16:43:14.694596 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.385645 (* 1 = 0.385645 loss) | |
| I0623 16:43:14.694609 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.112282 (* 0.0909091 = 0.0102075 loss) | |
| I0623 16:43:14.694623 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0966811 (* 0.0909091 = 0.00878919 loss) | |
| I0623 16:43:14.694638 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0702422 (* 0.0909091 = 0.00638565 loss) | |
| I0623 16:43:14.694650 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.139635 (* 0.0909091 = 0.0126941 loss) | |
| I0623 16:43:14.694664 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.776243 (* 0.0909091 = 0.0705676 loss) | |
| I0623 16:43:14.694677 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.17173 (* 0.0909091 = 0.0156118 loss) | |
| I0623 16:43:14.694691 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.447207 (* 0.0909091 = 0.0406552 loss) | |
| I0623 16:43:14.694705 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.268661 (* 0.0909091 = 0.0244237 loss) | |
| I0623 16:43:14.694717 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.02407 (* 0.0909091 = 0.0930974 loss) | |
| I0623 16:43:14.694731 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.660553 (* 0.0909091 = 0.0600502 loss) | |
| I0623 16:43:14.694744 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.829719 (* 0.0909091 = 0.075429 loss) | |
| I0623 16:43:14.694757 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.44653 (* 0.0909091 = 0.131502 loss) | |
| I0623 16:43:14.694772 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.16527 (* 0.0909091 = 0.105934 loss) | |
| I0623 16:43:14.694784 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.38038 (* 0.0909091 = 0.125489 loss) | |
| I0623 16:43:14.694798 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.694915 (* 0.0909091 = 0.0631741 loss) | |
| I0623 16:43:14.694811 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.481077 (* 0.0909091 = 0.0437343 loss) | |
| I0623 16:43:14.694825 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.420824 (* 0.0909091 = 0.0382567 loss) | |
| I0623 16:43:14.694839 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0525545 (* 0.0909091 = 0.00477768 loss) | |
| I0623 16:43:14.694852 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00294314 (* 0.0909091 = 0.000267558 loss) | |
| I0623 16:43:14.694867 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000599568 (* 0.0909091 = 5.45062e-05 loss) | |
| I0623 16:43:14.694881 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000321309 (* 0.0909091 = 2.92099e-05 loss) | |
| I0623 16:43:14.694895 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 1.82245e-05 (* 0.0909091 = 1.65678e-06 loss) | |
| I0623 16:43:14.694907 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 16:43:14.694919 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 16:43:14.694931 10365 solver.cpp:245] Train net output #149: total_confidence = 0.103027 | |
| I0623 16:43:14.694952 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.118955 | |
| I0623 16:43:14.694967 10365 sgd_solver.cpp:106] Iteration 11500, lr = 0.001 | |
| I0623 16:44:04.047317 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.3448 > 30) by scale factor 0.988637 | |
| I0623 16:44:20.112308 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.102 > 30) by scale factor 0.854652 | |
| I0623 16:45:19.842679 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.2993 > 30) by scale factor 0.990121 | |
| I0623 16:49:37.670214 10365 solver.cpp:229] Iteration 12000, loss = 4.52634 | |
| I0623 16:49:37.670316 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.460177 | |
| I0623 16:49:37.670336 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 16:49:37.670349 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 16:49:37.670362 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 16:49:37.670375 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 16:49:37.670387 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 16:49:37.670399 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.125 | |
| I0623 16:49:37.670413 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 16:49:37.670426 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 16:49:37.670439 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 16:49:37.670450 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 16:49:37.670464 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 16:49:37.670475 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 16:49:37.670488 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.375 | |
| I0623 16:49:37.670500 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 16:49:37.670511 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 16:49:37.670523 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.625 | |
| I0623 16:49:37.670536 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 16:49:37.670547 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 16:49:37.670558 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:49:37.670569 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:49:37.670580 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:49:37.670593 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:49:37.670603 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.642045 | |
| I0623 16:49:37.670615 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.690265 | |
| I0623 16:49:37.670632 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.75543 (* 0.3 = 0.526629 loss) | |
| I0623 16:49:37.670647 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.16629 (* 0.3 = 0.349887 loss) | |
| I0623 16:49:37.670662 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.16921 (* 0.0272727 = 0.00461482 loss) | |
| I0623 16:49:37.670676 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.63819 (* 0.0272727 = 0.044678 loss) | |
| I0623 16:49:37.670691 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.08875 (* 0.0272727 = 0.0569658 loss) | |
| I0623 16:49:37.670704 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.931 (* 0.0272727 = 0.0526637 loss) | |
| I0623 16:49:37.670718 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.5953 (* 0.0272727 = 0.0707809 loss) | |
| I0623 16:49:37.670732 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.48174 (* 0.0272727 = 0.0676839 loss) | |
| I0623 16:49:37.670745 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.67417 (* 0.0272727 = 0.0456592 loss) | |
| I0623 16:49:37.670759 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.32456 (* 0.0272727 = 0.0633971 loss) | |
| I0623 16:49:37.670773 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.99803 (* 0.0272727 = 0.0544919 loss) | |
| I0623 16:49:37.670786 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.03548 (* 0.0272727 = 0.0555131 loss) | |
| I0623 16:49:37.670800 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.03702 (* 0.0272727 = 0.055555 loss) | |
| I0623 16:49:37.670814 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.27031 (* 0.0272727 = 0.0619176 loss) | |
| I0623 16:49:37.670846 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.06598 (* 0.0272727 = 0.056345 loss) | |
| I0623 16:49:37.670861 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.44897 (* 0.0272727 = 0.0395174 loss) | |
| I0623 16:49:37.670874 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.75902 (* 0.0272727 = 0.0479734 loss) | |
| I0623 16:49:37.670888 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.10108 (* 0.0272727 = 0.0300295 loss) | |
| I0623 16:49:37.670902 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.706438 (* 0.0272727 = 0.0192665 loss) | |
| I0623 16:49:37.670915 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0544418 (* 0.0272727 = 0.00148478 loss) | |
| I0623 16:49:37.670929 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.016769 (* 0.0272727 = 0.000457337 loss) | |
| I0623 16:49:37.670943 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00752112 (* 0.0272727 = 0.000205121 loss) | |
| I0623 16:49:37.670958 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.00194064 (* 0.0272727 = 5.29265e-05 loss) | |
| I0623 16:49:37.670971 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000561589 (* 0.0272727 = 1.53161e-05 loss) | |
| I0623 16:49:37.670984 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.433628 | |
| I0623 16:49:37.670995 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.75 | |
| I0623 16:49:37.671007 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 16:49:37.671018 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 16:49:37.671030 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 16:49:37.671041 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 16:49:37.671052 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.125 | |
| I0623 16:49:37.671063 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 16:49:37.671074 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.25 | |
| I0623 16:49:37.671085 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 16:49:37.671097 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.625 | |
| I0623 16:49:37.671108 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 16:49:37.671119 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 16:49:37.671134 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 16:49:37.671146 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 16:49:37.671157 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 16:49:37.671169 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.5 | |
| I0623 16:49:37.671180 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 16:49:37.671191 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 16:49:37.671202 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:49:37.671214 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:49:37.671226 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:49:37.671237 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:49:37.671248 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.636364 | |
| I0623 16:49:37.671260 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.725664 | |
| I0623 16:49:37.671273 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.69091 (* 0.3 = 0.507274 loss) | |
| I0623 16:49:37.671288 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 1.11145 (* 0.3 = 0.333435 loss) | |
| I0623 16:49:37.671301 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.73391 (* 0.0272727 = 0.0200157 loss) | |
| I0623 16:49:37.671315 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.960113 (* 0.0272727 = 0.0261849 loss) | |
| I0623 16:49:37.671345 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.4012 (* 0.0272727 = 0.0382146 loss) | |
| I0623 16:49:37.671361 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.46361 (* 0.0272727 = 0.0399168 loss) | |
| I0623 16:49:37.671375 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.61199 (* 0.0272727 = 0.0712362 loss) | |
| I0623 16:49:37.671388 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.29654 (* 0.0272727 = 0.062633 loss) | |
| I0623 16:49:37.671402 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.81873 (* 0.0272727 = 0.0496017 loss) | |
| I0623 16:49:37.671416 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.14243 (* 0.0272727 = 0.05843 loss) | |
| I0623 16:49:37.671429 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.7842 (* 0.0272727 = 0.04866 loss) | |
| I0623 16:49:37.671442 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.17458 (* 0.0272727 = 0.0593066 loss) | |
| I0623 16:49:37.671455 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.94219 (* 0.0272727 = 0.0529687 loss) | |
| I0623 16:49:37.671469 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.70852 (* 0.0272727 = 0.0465959 loss) | |
| I0623 16:49:37.671479 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 2.04199 (* 0.0272727 = 0.0556906 loss) | |
| I0623 16:49:37.671489 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.18702 (* 0.0272727 = 0.0323733 loss) | |
| I0623 16:49:37.671502 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.84987 (* 0.0272727 = 0.0504509 loss) | |
| I0623 16:49:37.671516 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 1.54393 (* 0.0272727 = 0.0421071 loss) | |
| I0623 16:49:37.671530 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.788862 (* 0.0272727 = 0.0215144 loss) | |
| I0623 16:49:37.671545 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0387418 (* 0.0272727 = 0.00105659 loss) | |
| I0623 16:49:37.671557 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0101359 (* 0.0272727 = 0.000276433 loss) | |
| I0623 16:49:37.671571 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00145732 (* 0.0272727 = 3.97451e-05 loss) | |
| I0623 16:49:37.671586 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000579242 (* 0.0272727 = 1.57975e-05 loss) | |
| I0623 16:49:37.671612 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000187417 (* 0.0272727 = 5.11138e-06 loss) | |
| I0623 16:49:37.671627 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.761062 | |
| I0623 16:49:37.671638 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.875 | |
| I0623 16:49:37.671650 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 16:49:37.671661 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:49:37.671672 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 16:49:37.671684 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 16:49:37.671695 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 16:49:37.671706 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 16:49:37.671717 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 16:49:37.671728 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 16:49:37.671739 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 16:49:37.671751 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 16:49:37.671762 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.875 | |
| I0623 16:49:37.671773 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.5 | |
| I0623 16:49:37.671784 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.5 | |
| I0623 16:49:37.671795 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.5 | |
| I0623 16:49:37.671808 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.625 | |
| I0623 16:49:37.671830 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 16:49:37.671843 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 16:49:37.671854 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:49:37.671865 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:49:37.671876 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:49:37.671887 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:49:37.671898 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.835227 | |
| I0623 16:49:37.671911 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.946903 | |
| I0623 16:49:37.671923 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.806652 (* 1 = 0.806652 loss) | |
| I0623 16:49:37.671937 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.547894 (* 1 = 0.547894 loss) | |
| I0623 16:49:37.671952 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.634492 (* 0.0909091 = 0.0576811 loss) | |
| I0623 16:49:37.671965 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0702586 (* 0.0909091 = 0.00638714 loss) | |
| I0623 16:49:37.671979 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.248632 (* 0.0909091 = 0.0226029 loss) | |
| I0623 16:49:37.671993 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0507668 (* 0.0909091 = 0.00461516 loss) | |
| I0623 16:49:37.672008 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.908621 (* 0.0909091 = 0.0826019 loss) | |
| I0623 16:49:37.672021 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.272386 (* 0.0909091 = 0.0247624 loss) | |
| I0623 16:49:37.672034 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.828241 (* 0.0909091 = 0.0752946 loss) | |
| I0623 16:49:37.672049 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.890719 (* 0.0909091 = 0.0809745 loss) | |
| I0623 16:49:37.672061 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.695628 (* 0.0909091 = 0.0632389 loss) | |
| I0623 16:49:37.672075 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.0149 (* 0.0909091 = 0.0922639 loss) | |
| I0623 16:49:37.672088 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.10254 (* 0.0909091 = 0.100231 loss) | |
| I0623 16:49:37.672102 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.07221 (* 0.0909091 = 0.0974739 loss) | |
| I0623 16:49:37.672116 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.33645 (* 0.0909091 = 0.121496 loss) | |
| I0623 16:49:37.672129 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.47575 (* 0.0909091 = 0.134159 loss) | |
| I0623 16:49:37.672143 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.4838 (* 0.0909091 = 0.134891 loss) | |
| I0623 16:49:37.672158 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 1.15384 (* 0.0909091 = 0.104894 loss) | |
| I0623 16:49:37.672170 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.23616 (* 0.0909091 = 0.0214691 loss) | |
| I0623 16:49:37.672188 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0948308 (* 0.0909091 = 0.00862098 loss) | |
| I0623 16:49:37.672202 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00928817 (* 0.0909091 = 0.000844379 loss) | |
| I0623 16:49:37.672216 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00312469 (* 0.0909091 = 0.000284063 loss) | |
| I0623 16:49:37.672230 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.00101282 (* 0.0909091 = 9.20747e-05 loss) | |
| I0623 16:49:37.672245 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 9.14792e-05 (* 0.0909091 = 8.31629e-06 loss) | |
| I0623 16:49:37.672257 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 16:49:37.672268 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 16:49:37.672281 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0847843 | |
| I0623 16:49:37.672302 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0518543 | |
| I0623 16:49:37.672317 10365 sgd_solver.cpp:106] Iteration 12000, lr = 0.001 | |
| I0623 16:50:39.387656 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.0707 > 30) by scale factor 0.748677 | |
| I0623 16:51:06.195912 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.571 > 30) by scale factor 0.981321 | |
| I0623 16:51:42.217351 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.0581 > 30) by scale factor 0.998068 | |
| I0623 16:52:12.141832 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.3044 > 30) by scale factor 0.826347 | |
| I0623 16:52:50.448083 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.6394 > 30) by scale factor 0.866067 | |
| I0623 16:54:13.208590 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.3364 > 30) by scale factor 0.70861 | |
| I0623 16:54:42.327561 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.0705 > 30) by scale factor 0.997654 | |
| I0623 16:56:00.881510 10365 solver.cpp:229] Iteration 12500, loss = 4.49231 | |
| I0623 16:56:00.881603 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.454545 | |
| I0623 16:56:00.881623 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 16:56:00.881635 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 16:56:00.881650 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 16:56:00.881662 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 16:56:00.881675 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.625 | |
| I0623 16:56:00.881687 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 16:56:00.881700 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.625 | |
| I0623 16:56:00.881713 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 16:56:00.881726 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 16:56:00.881738 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 16:56:00.881750 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 16:56:00.881763 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 16:56:00.881775 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 16:56:00.881788 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 16:56:00.881798 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 16:56:00.881810 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 16:56:00.881821 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.75 | |
| I0623 16:56:00.881834 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.75 | |
| I0623 16:56:00.881845 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 16:56:00.881857 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 16:56:00.881868 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 16:56:00.881880 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 16:56:00.881892 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.681818 | |
| I0623 16:56:00.881903 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.818182 | |
| I0623 16:56:00.881927 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.51407 (* 0.3 = 0.454222 loss) | |
| I0623 16:56:00.881940 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.89882 (* 0.3 = 0.269646 loss) | |
| I0623 16:56:00.881956 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.34974 (* 0.0272727 = 0.00953835 loss) | |
| I0623 16:56:00.881970 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.901743 (* 0.0272727 = 0.024593 loss) | |
| I0623 16:56:00.881984 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.55412 (* 0.0272727 = 0.042385 loss) | |
| I0623 16:56:00.881999 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.96021 (* 0.0272727 = 0.0534602 loss) | |
| I0623 16:56:00.882012 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.29869 (* 0.0272727 = 0.0354189 loss) | |
| I0623 16:56:00.882026 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.13008 (* 0.0272727 = 0.0580931 loss) | |
| I0623 16:56:00.882040 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.18126 (* 0.0272727 = 0.0322161 loss) | |
| I0623 16:56:00.882053 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.03106 (* 0.0272727 = 0.0281197 loss) | |
| I0623 16:56:00.882069 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.6607 (* 0.0272727 = 0.0452919 loss) | |
| I0623 16:56:00.882084 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.46395 (* 0.0272727 = 0.0399258 loss) | |
| I0623 16:56:00.882098 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.29819 (* 0.0272727 = 0.0354053 loss) | |
| I0623 16:56:00.882112 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.60281 (* 0.0272727 = 0.0437131 loss) | |
| I0623 16:56:00.882144 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.22672 (* 0.0272727 = 0.033456 loss) | |
| I0623 16:56:00.882159 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.27059 (* 0.0272727 = 0.0346524 loss) | |
| I0623 16:56:00.882174 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.743815 (* 0.0272727 = 0.0202859 loss) | |
| I0623 16:56:00.882187 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.943332 (* 0.0272727 = 0.0257272 loss) | |
| I0623 16:56:00.882200 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 1.34065 (* 0.0272727 = 0.0365632 loss) | |
| I0623 16:56:00.882213 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 1.28968 (* 0.0272727 = 0.0351731 loss) | |
| I0623 16:56:00.882228 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0114037 (* 0.0272727 = 0.000311011 loss) | |
| I0623 16:56:00.882242 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000826226 (* 0.0272727 = 2.25334e-05 loss) | |
| I0623 16:56:00.882256 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000104424 (* 0.0272727 = 2.84792e-06 loss) | |
| I0623 16:56:00.882271 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.37974e-05 (* 0.0272727 = 9.21747e-07 loss) | |
| I0623 16:56:00.882282 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.525253 | |
| I0623 16:56:00.882294 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 16:56:00.882307 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 16:56:00.882318 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 16:56:00.882329 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.375 | |
| I0623 16:56:00.882340 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 16:56:00.882352 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 16:56:00.882364 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 16:56:00.882376 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.625 | |
| I0623 16:56:00.882387 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.625 | |
| I0623 16:56:00.882400 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.25 | |
| I0623 16:56:00.882411 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.75 | |
| I0623 16:56:00.882422 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 16:56:00.882433 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 16:56:00.882444 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 16:56:00.882457 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 16:56:00.882467 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 16:56:00.882479 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.625 | |
| I0623 16:56:00.882490 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.75 | |
| I0623 16:56:00.882501 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 16:56:00.882513 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 16:56:00.882524 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 16:56:00.882535 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 16:56:00.882547 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.721591 | |
| I0623 16:56:00.882558 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.808081 | |
| I0623 16:56:00.882572 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.37026 (* 0.3 = 0.411078 loss) | |
| I0623 16:56:00.882586 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.805218 (* 0.3 = 0.241565 loss) | |
| I0623 16:56:00.882599 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.34969 (* 0.0272727 = 0.009537 loss) | |
| I0623 16:56:00.882613 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.167659 (* 0.0272727 = 0.00457251 loss) | |
| I0623 16:56:00.882638 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.54512 (* 0.0272727 = 0.0148669 loss) | |
| I0623 16:56:00.882653 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.01428 (* 0.0272727 = 0.0276623 loss) | |
| I0623 16:56:00.882668 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.05698 (* 0.0272727 = 0.0560994 loss) | |
| I0623 16:56:00.882681 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.61018 (* 0.0272727 = 0.0439139 loss) | |
| I0623 16:56:00.882695 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.44854 (* 0.0272727 = 0.0395057 loss) | |
| I0623 16:56:00.882709 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.13734 (* 0.0272727 = 0.0310183 loss) | |
| I0623 16:56:00.882722 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.60361 (* 0.0272727 = 0.0437348 loss) | |
| I0623 16:56:00.882736 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.49372 (* 0.0272727 = 0.0407378 loss) | |
| I0623 16:56:00.882750 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 0.927388 (* 0.0272727 = 0.0252924 loss) | |
| I0623 16:56:00.882763 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.70964 (* 0.0272727 = 0.0466266 loss) | |
| I0623 16:56:00.882776 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.53219 (* 0.0272727 = 0.0417871 loss) | |
| I0623 16:56:00.882791 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.24755 (* 0.0272727 = 0.0340242 loss) | |
| I0623 16:56:00.882803 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.13972 (* 0.0272727 = 0.0310832 loss) | |
| I0623 16:56:00.882817 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.738118 (* 0.0272727 = 0.0201305 loss) | |
| I0623 16:56:00.882830 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 1.26818 (* 0.0272727 = 0.0345868 loss) | |
| I0623 16:56:00.882844 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 1.02597 (* 0.0272727 = 0.0279809 loss) | |
| I0623 16:56:00.882858 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00917371 (* 0.0272727 = 0.000250192 loss) | |
| I0623 16:56:00.882872 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000871216 (* 0.0272727 = 2.37604e-05 loss) | |
| I0623 16:56:00.882886 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 7.9822e-05 (* 0.0272727 = 2.17696e-06 loss) | |
| I0623 16:56:00.882900 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 2.1444e-05 (* 0.0272727 = 5.84837e-07 loss) | |
| I0623 16:56:00.882912 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.818182 | |
| I0623 16:56:00.882925 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 16:56:00.882936 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 16:56:00.882948 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 16:56:00.882964 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 16:56:00.882977 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 16:56:00.882987 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 16:56:00.882999 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 16:56:00.883010 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 16:56:00.883023 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 16:56:00.883033 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 16:56:00.883045 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 16:56:00.883056 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.875 | |
| I0623 16:56:00.883069 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 1 | |
| I0623 16:56:00.883080 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 16:56:00.883090 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 16:56:00.883103 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 16:56:00.883127 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.75 | |
| I0623 16:56:00.883141 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.75 | |
| I0623 16:56:00.883153 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 16:56:00.883164 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 16:56:00.883175 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 16:56:00.883188 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 16:56:00.883198 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.892045 | |
| I0623 16:56:00.883210 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.929293 | |
| I0623 16:56:00.883224 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.626364 (* 1 = 0.626364 loss) | |
| I0623 16:56:00.883237 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.364681 (* 1 = 0.364681 loss) | |
| I0623 16:56:00.883251 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0226147 (* 0.0909091 = 0.00205588 loss) | |
| I0623 16:56:00.883265 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0271061 (* 0.0909091 = 0.00246419 loss) | |
| I0623 16:56:00.883280 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0585525 (* 0.0909091 = 0.00532296 loss) | |
| I0623 16:56:00.883293 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.067429 (* 0.0909091 = 0.00612991 loss) | |
| I0623 16:56:00.883307 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 1.46406 (* 0.0909091 = 0.133096 loss) | |
| I0623 16:56:00.883321 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.091514 (* 0.0909091 = 0.00831945 loss) | |
| I0623 16:56:00.883334 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.0902642 (* 0.0909091 = 0.00820583 loss) | |
| I0623 16:56:00.883348 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.197236 (* 0.0909091 = 0.0179306 loss) | |
| I0623 16:56:00.883363 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.474311 (* 0.0909091 = 0.0431192 loss) | |
| I0623 16:56:00.883376 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.64602 (* 0.0909091 = 0.0587291 loss) | |
| I0623 16:56:00.883389 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.866394 (* 0.0909091 = 0.0787631 loss) | |
| I0623 16:56:00.883404 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.807648 (* 0.0909091 = 0.0734226 loss) | |
| I0623 16:56:00.883417 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.35331 (* 0.0909091 = 0.0321191 loss) | |
| I0623 16:56:00.883430 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.950096 (* 0.0909091 = 0.0863724 loss) | |
| I0623 16:56:00.883445 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.876865 (* 0.0909091 = 0.079715 loss) | |
| I0623 16:56:00.883457 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 1.04649 (* 0.0909091 = 0.0951358 loss) | |
| I0623 16:56:00.883471 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.803938 (* 0.0909091 = 0.0730853 loss) | |
| I0623 16:56:00.883486 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.861179 (* 0.0909091 = 0.078289 loss) | |
| I0623 16:56:00.883499 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0226982 (* 0.0909091 = 0.00206347 loss) | |
| I0623 16:56:00.883513 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00575558 (* 0.0909091 = 0.000523234 loss) | |
| I0623 16:56:00.883527 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.00100878 (* 0.0909091 = 9.17071e-05 loss) | |
| I0623 16:56:00.883541 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 0.000111299 (* 0.0909091 = 1.01181e-05 loss) | |
| I0623 16:56:00.883553 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 16:56:00.883565 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.25 | |
| I0623 16:56:00.883576 10365 solver.cpp:245] Train net output #149: total_confidence = 0.260875 | |
| I0623 16:56:00.883611 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.251353 | |
| I0623 16:56:00.883630 10365 sgd_solver.cpp:106] Iteration 12500, lr = 0.001 | |
| I0623 16:58:30.644220 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.0449 > 30) by scale factor 0.856045 | |
| I0623 17:02:23.938454 10365 solver.cpp:229] Iteration 13000, loss = 4.56353 | |
| I0623 17:02:23.938544 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.356436 | |
| I0623 17:02:23.938562 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 17:02:23.938575 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 17:02:23.938588 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 17:02:23.938601 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.125 | |
| I0623 17:02:23.938614 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 17:02:23.938627 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 17:02:23.938639 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 17:02:23.938652 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 17:02:23.938663 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 17:02:23.938676 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 17:02:23.938688 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 17:02:23.938701 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 17:02:23.938714 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 17:02:23.938724 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 17:02:23.938736 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.5 | |
| I0623 17:02:23.938748 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 17:02:23.938760 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 17:02:23.938771 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 17:02:23.938783 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 0.875 | |
| I0623 17:02:23.938796 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 0.875 | |
| I0623 17:02:23.938807 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:02:23.938818 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:02:23.938829 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.590909 | |
| I0623 17:02:23.938841 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.712871 | |
| I0623 17:02:23.938858 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.85997 (* 0.3 = 0.557991 loss) | |
| I0623 17:02:23.938873 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.18275 (* 0.3 = 0.354824 loss) | |
| I0623 17:02:23.938887 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.734833 (* 0.0272727 = 0.0200409 loss) | |
| I0623 17:02:23.938900 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.09891 (* 0.0272727 = 0.0299704 loss) | |
| I0623 17:02:23.938915 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.26292 (* 0.0272727 = 0.0617159 loss) | |
| I0623 17:02:23.938927 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.54657 (* 0.0272727 = 0.0421791 loss) | |
| I0623 17:02:23.938941 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.60146 (* 0.0272727 = 0.0709489 loss) | |
| I0623 17:02:23.938956 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.60786 (* 0.0272727 = 0.0438507 loss) | |
| I0623 17:02:23.938969 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.93875 (* 0.0272727 = 0.052875 loss) | |
| I0623 17:02:23.938982 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.65408 (* 0.0272727 = 0.0451113 loss) | |
| I0623 17:02:23.938997 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.88198 (* 0.0272727 = 0.0513268 loss) | |
| I0623 17:02:23.939009 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.45347 (* 0.0272727 = 0.0396402 loss) | |
| I0623 17:02:23.939023 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.5872 (* 0.0272727 = 0.0432873 loss) | |
| I0623 17:02:23.939038 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.63028 (* 0.0272727 = 0.0444622 loss) | |
| I0623 17:02:23.939069 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.35205 (* 0.0272727 = 0.036874 loss) | |
| I0623 17:02:23.939085 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.65626 (* 0.0272727 = 0.0451706 loss) | |
| I0623 17:02:23.939097 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.59847 (* 0.0272727 = 0.0435946 loss) | |
| I0623 17:02:23.939111 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.770556 (* 0.0272727 = 0.0210152 loss) | |
| I0623 17:02:23.939136 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.425662 (* 0.0272727 = 0.011609 loss) | |
| I0623 17:02:23.939163 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.2381 (* 0.0272727 = 0.00649365 loss) | |
| I0623 17:02:23.939180 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.489116 (* 0.0272727 = 0.0133395 loss) | |
| I0623 17:02:23.939194 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.383576 (* 0.0272727 = 0.0104612 loss) | |
| I0623 17:02:23.939209 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.00308424 (* 0.0272727 = 8.41158e-05 loss) | |
| I0623 17:02:23.939224 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000516253 (* 0.0272727 = 1.40796e-05 loss) | |
| I0623 17:02:23.939235 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.504951 | |
| I0623 17:02:23.939247 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 17:02:23.939260 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 17:02:23.939271 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.375 | |
| I0623 17:02:23.939282 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 17:02:23.939294 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 17:02:23.939306 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 17:02:23.939316 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 17:02:23.939328 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.625 | |
| I0623 17:02:23.939339 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.625 | |
| I0623 17:02:23.939352 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 17:02:23.939363 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 17:02:23.939374 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.25 | |
| I0623 17:02:23.939385 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 17:02:23.939398 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 17:02:23.939409 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 17:02:23.939420 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 17:02:23.939431 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 17:02:23.939442 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 17:02:23.939453 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 0.875 | |
| I0623 17:02:23.939465 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 0.875 | |
| I0623 17:02:23.939477 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:02:23.939488 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:02:23.939499 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.693182 | |
| I0623 17:02:23.939512 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.811881 | |
| I0623 17:02:23.939525 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.4448 (* 0.3 = 0.43344 loss) | |
| I0623 17:02:23.939539 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.920838 (* 0.3 = 0.276251 loss) | |
| I0623 17:02:23.939553 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.29178 (* 0.0272727 = 0.00795763 loss) | |
| I0623 17:02:23.939568 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.600669 (* 0.0272727 = 0.0163819 loss) | |
| I0623 17:02:23.939594 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 2.57627 (* 0.0272727 = 0.0702618 loss) | |
| I0623 17:02:23.939625 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.865708 (* 0.0272727 = 0.0236102 loss) | |
| I0623 17:02:23.939638 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.94965 (* 0.0272727 = 0.0531722 loss) | |
| I0623 17:02:23.939652 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.68062 (* 0.0272727 = 0.0458352 loss) | |
| I0623 17:02:23.939666 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.49599 (* 0.0272727 = 0.0407997 loss) | |
| I0623 17:02:23.939679 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.24403 (* 0.0272727 = 0.033928 loss) | |
| I0623 17:02:23.939693 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.62951 (* 0.0272727 = 0.0444411 loss) | |
| I0623 17:02:23.939707 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.52166 (* 0.0272727 = 0.0414998 loss) | |
| I0623 17:02:23.939721 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.47785 (* 0.0272727 = 0.0403049 loss) | |
| I0623 17:02:23.939734 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.59805 (* 0.0272727 = 0.0435833 loss) | |
| I0623 17:02:23.939749 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.2795 (* 0.0272727 = 0.0348955 loss) | |
| I0623 17:02:23.939762 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.31954 (* 0.0272727 = 0.0359873 loss) | |
| I0623 17:02:23.939775 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.35727 (* 0.0272727 = 0.0370164 loss) | |
| I0623 17:02:23.939790 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.644564 (* 0.0272727 = 0.017579 loss) | |
| I0623 17:02:23.939803 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.621562 (* 0.0272727 = 0.0169517 loss) | |
| I0623 17:02:23.939817 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.435038 (* 0.0272727 = 0.0118647 loss) | |
| I0623 17:02:23.939831 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.550726 (* 0.0272727 = 0.0150198 loss) | |
| I0623 17:02:23.939844 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.658926 (* 0.0272727 = 0.0179707 loss) | |
| I0623 17:02:23.939859 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.0286293 (* 0.0272727 = 0.000780799 loss) | |
| I0623 17:02:23.939873 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.00713939 (* 0.0272727 = 0.000194711 loss) | |
| I0623 17:02:23.939885 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.851485 | |
| I0623 17:02:23.939898 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 17:02:23.939908 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 17:02:23.939920 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 17:02:23.939932 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 17:02:23.939944 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 17:02:23.939955 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.75 | |
| I0623 17:02:23.939966 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 17:02:23.939977 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 17:02:23.939990 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 17:02:23.940001 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 17:02:23.940012 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 17:02:23.940023 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 17:02:23.940034 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.5 | |
| I0623 17:02:23.940047 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 17:02:23.940057 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 17:02:23.940068 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 17:02:23.940093 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.75 | |
| I0623 17:02:23.940105 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.875 | |
| I0623 17:02:23.940117 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 0.875 | |
| I0623 17:02:23.940129 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 0.875 | |
| I0623 17:02:23.940140 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:02:23.940151 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:02:23.940162 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.903409 | |
| I0623 17:02:23.940174 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.940594 | |
| I0623 17:02:23.940191 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.580624 (* 1 = 0.580624 loss) | |
| I0623 17:02:23.940207 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.371603 (* 1 = 0.371603 loss) | |
| I0623 17:02:23.940220 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0623311 (* 0.0909091 = 0.00566646 loss) | |
| I0623 17:02:23.940234 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.159669 (* 0.0909091 = 0.0145154 loss) | |
| I0623 17:02:23.940248 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.658661 (* 0.0909091 = 0.0598783 loss) | |
| I0623 17:02:23.940261 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0810581 (* 0.0909091 = 0.00736892 loss) | |
| I0623 17:02:23.940275 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.164657 (* 0.0909091 = 0.0149689 loss) | |
| I0623 17:02:23.940289 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.456321 (* 0.0909091 = 0.0414838 loss) | |
| I0623 17:02:23.940302 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.0639302 (* 0.0909091 = 0.00581184 loss) | |
| I0623 17:02:23.940316 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.218851 (* 0.0909091 = 0.0198956 loss) | |
| I0623 17:02:23.940330 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.447448 (* 0.0909091 = 0.0406771 loss) | |
| I0623 17:02:23.940343 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.569987 (* 0.0909091 = 0.051817 loss) | |
| I0623 17:02:23.940356 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.919186 (* 0.0909091 = 0.0835624 loss) | |
| I0623 17:02:23.940371 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.20239 (* 0.0909091 = 0.109309 loss) | |
| I0623 17:02:23.940383 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.18645 (* 0.0909091 = 0.107859 loss) | |
| I0623 17:02:23.940397 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.06881 (* 0.0909091 = 0.0971648 loss) | |
| I0623 17:02:23.940412 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.06483 (* 0.0909091 = 0.0968028 loss) | |
| I0623 17:02:23.940424 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.491227 (* 0.0909091 = 0.044657 loss) | |
| I0623 17:02:23.940438 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.685921 (* 0.0909091 = 0.0623565 loss) | |
| I0623 17:02:23.940451 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.367355 (* 0.0909091 = 0.0333959 loss) | |
| I0623 17:02:23.940465 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.432365 (* 0.0909091 = 0.0393059 loss) | |
| I0623 17:02:23.940479 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.361712 (* 0.0909091 = 0.0328829 loss) | |
| I0623 17:02:23.940492 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.00141603 (* 0.0909091 = 0.00012873 loss) | |
| I0623 17:02:23.940507 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 0.000235236 (* 0.0909091 = 2.13851e-05 loss) | |
| I0623 17:02:23.940520 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 17:02:23.940531 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.25 | |
| I0623 17:02:23.940542 10365 solver.cpp:245] Train net output #149: total_confidence = 0.207359 | |
| I0623 17:02:23.940563 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.203442 | |
| I0623 17:02:23.940578 10365 sgd_solver.cpp:106] Iteration 13000, lr = 0.001 | |
| I0623 17:06:05.680428 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.5556 > 30) by scale factor 0.739726 | |
| I0623 17:07:35.340976 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 44.3485 > 30) by scale factor 0.676461 | |
| I0623 17:07:58.323185 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.0766 > 30) by scale factor 0.965356 | |
| I0623 17:08:07.513934 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.1793 > 30) by scale factor 0.711249 | |
| I0623 17:08:46.991204 10365 solver.cpp:229] Iteration 13500, loss = 4.41842 | |
| I0623 17:08:46.991327 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.493671 | |
| I0623 17:08:46.991346 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 17:08:46.991360 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 17:08:46.991374 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 17:08:46.991386 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 17:08:46.991399 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.625 | |
| I0623 17:08:46.991411 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 17:08:46.991423 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.625 | |
| I0623 17:08:46.991436 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.75 | |
| I0623 17:08:46.991448 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 17:08:46.991461 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 17:08:46.991472 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.625 | |
| I0623 17:08:46.991484 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.75 | |
| I0623 17:08:46.991497 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 17:08:46.991508 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 17:08:46.991519 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.875 | |
| I0623 17:08:46.991531 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 1 | |
| I0623 17:08:46.991544 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 17:08:46.991554 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 17:08:46.991566 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 17:08:46.991577 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 17:08:46.991590 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:08:46.991617 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:08:46.991632 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.767045 | |
| I0623 17:08:46.991644 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.873418 | |
| I0623 17:08:46.991660 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.30131 (* 0.3 = 0.390392 loss) | |
| I0623 17:08:46.991675 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.616494 (* 0.3 = 0.184948 loss) | |
| I0623 17:08:46.991689 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.470662 (* 0.0272727 = 0.0128362 loss) | |
| I0623 17:08:46.991703 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.861369 (* 0.0272727 = 0.0234919 loss) | |
| I0623 17:08:46.991717 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.3584 (* 0.0272727 = 0.0370473 loss) | |
| I0623 17:08:46.991731 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.087 (* 0.0272727 = 0.0296455 loss) | |
| I0623 17:08:46.991745 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.06379 (* 0.0272727 = 0.0562852 loss) | |
| I0623 17:08:46.991758 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.24038 (* 0.0272727 = 0.0338285 loss) | |
| I0623 17:08:46.991772 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.49219 (* 0.0272727 = 0.0406961 loss) | |
| I0623 17:08:46.991786 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 0.949796 (* 0.0272727 = 0.0259035 loss) | |
| I0623 17:08:46.991799 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.53658 (* 0.0272727 = 0.0419067 loss) | |
| I0623 17:08:46.991813 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.5903 (* 0.0272727 = 0.0433717 loss) | |
| I0623 17:08:46.991827 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 0.996937 (* 0.0272727 = 0.0271892 loss) | |
| I0623 17:08:46.991842 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 0.952825 (* 0.0272727 = 0.0259861 loss) | |
| I0623 17:08:46.991873 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.13091 (* 0.0272727 = 0.0308431 loss) | |
| I0623 17:08:46.991888 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.842102 (* 0.0272727 = 0.0229664 loss) | |
| I0623 17:08:46.991902 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.605672 (* 0.0272727 = 0.0165183 loss) | |
| I0623 17:08:46.991916 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.00724254 (* 0.0272727 = 0.000197524 loss) | |
| I0623 17:08:46.991931 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00089125 (* 0.0272727 = 2.43068e-05 loss) | |
| I0623 17:08:46.991945 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 8.60887e-05 (* 0.0272727 = 2.34787e-06 loss) | |
| I0623 17:08:46.991960 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 3.59689e-05 (* 0.0272727 = 9.8097e-07 loss) | |
| I0623 17:08:46.991973 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 3.3975e-06 (* 0.0272727 = 9.26592e-08 loss) | |
| I0623 17:08:46.991987 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 1.92226e-06 (* 0.0272727 = 5.24253e-08 loss) | |
| I0623 17:08:46.992002 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 2.68221e-07 (* 0.0272727 = 7.31512e-09 loss) | |
| I0623 17:08:46.992014 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.632911 | |
| I0623 17:08:46.992027 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 17:08:46.992038 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 17:08:46.992049 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 17:08:46.992060 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.875 | |
| I0623 17:08:46.992072 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 17:08:46.992084 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.875 | |
| I0623 17:08:46.992095 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.625 | |
| I0623 17:08:46.992107 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 17:08:46.992118 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 17:08:46.992130 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 17:08:46.992141 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 17:08:46.992153 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 17:08:46.992164 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.875 | |
| I0623 17:08:46.992177 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 17:08:46.992187 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 17:08:46.992199 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 17:08:46.992210 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 17:08:46.992221 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 17:08:46.992233 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 17:08:46.992244 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 17:08:46.992256 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:08:46.992270 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:08:46.992281 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.818182 | |
| I0623 17:08:46.992293 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.886076 | |
| I0623 17:08:46.992307 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.07734 (* 0.3 = 0.323201 loss) | |
| I0623 17:08:46.992321 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.527907 (* 0.3 = 0.158372 loss) | |
| I0623 17:08:46.992334 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.115498 (* 0.0272727 = 0.00314995 loss) | |
| I0623 17:08:46.992352 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.47737 (* 0.0272727 = 0.0130192 loss) | |
| I0623 17:08:46.992378 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.382497 (* 0.0272727 = 0.0104317 loss) | |
| I0623 17:08:46.992393 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.570017 (* 0.0272727 = 0.0155459 loss) | |
| I0623 17:08:46.992408 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.39008 (* 0.0272727 = 0.0379114 loss) | |
| I0623 17:08:46.992421 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 0.680459 (* 0.0272727 = 0.018558 loss) | |
| I0623 17:08:46.992434 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 0.83461 (* 0.0272727 = 0.0227621 loss) | |
| I0623 17:08:46.992449 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.29737 (* 0.0272727 = 0.0353828 loss) | |
| I0623 17:08:46.992462 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.90324 (* 0.0272727 = 0.0519065 loss) | |
| I0623 17:08:46.992475 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.95626 (* 0.0272727 = 0.0533524 loss) | |
| I0623 17:08:46.992489 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.10226 (* 0.0272727 = 0.0300616 loss) | |
| I0623 17:08:46.992503 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 0.903734 (* 0.0272727 = 0.0246473 loss) | |
| I0623 17:08:46.992516 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.713086 (* 0.0272727 = 0.0194478 loss) | |
| I0623 17:08:46.992530 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.681021 (* 0.0272727 = 0.0185733 loss) | |
| I0623 17:08:46.992543 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.476224 (* 0.0272727 = 0.0129879 loss) | |
| I0623 17:08:46.992558 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.0148979 (* 0.0272727 = 0.000406308 loss) | |
| I0623 17:08:46.992571 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0023977 (* 0.0272727 = 6.53917e-05 loss) | |
| I0623 17:08:46.992585 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.000242876 (* 0.0272727 = 6.62389e-06 loss) | |
| I0623 17:08:46.992599 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 2.51251e-05 (* 0.0272727 = 6.85229e-07 loss) | |
| I0623 17:08:46.992612 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 1.16979e-05 (* 0.0272727 = 3.19034e-07 loss) | |
| I0623 17:08:46.992626 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 4.05318e-06 (* 0.0272727 = 1.10541e-07 loss) | |
| I0623 17:08:46.992640 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 5.36442e-07 (* 0.0272727 = 1.46302e-08 loss) | |
| I0623 17:08:46.992652 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.911392 | |
| I0623 17:08:46.992666 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 17:08:46.992676 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 17:08:46.992688 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 17:08:46.992699 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 17:08:46.992710 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 17:08:46.992722 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 17:08:46.992733 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 17:08:46.992744 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 17:08:46.992756 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 17:08:46.992768 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 17:08:46.992779 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.875 | |
| I0623 17:08:46.992790 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.75 | |
| I0623 17:08:46.992802 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 17:08:46.992813 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 17:08:46.992825 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 17:08:46.992846 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 17:08:46.992859 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 17:08:46.992871 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 17:08:46.992882 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 17:08:46.992894 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 17:08:46.992905 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:08:46.992916 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:08:46.992928 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.960227 | |
| I0623 17:08:46.992939 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.987342 | |
| I0623 17:08:46.992954 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.337426 (* 1 = 0.337426 loss) | |
| I0623 17:08:46.992967 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.155006 (* 1 = 0.155006 loss) | |
| I0623 17:08:46.992981 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0177278 (* 0.0909091 = 0.00161162 loss) | |
| I0623 17:08:46.992995 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0261488 (* 0.0909091 = 0.00237717 loss) | |
| I0623 17:08:46.993010 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0190429 (* 0.0909091 = 0.00173117 loss) | |
| I0623 17:08:46.993023 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0606323 (* 0.0909091 = 0.00551202 loss) | |
| I0623 17:08:46.993036 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.235178 (* 0.0909091 = 0.0213798 loss) | |
| I0623 17:08:46.993051 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.0481128 (* 0.0909091 = 0.0043739 loss) | |
| I0623 17:08:46.993063 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.415632 (* 0.0909091 = 0.0377847 loss) | |
| I0623 17:08:46.993077 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.80269 (* 0.0909091 = 0.0729718 loss) | |
| I0623 17:08:46.993090 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.411909 (* 0.0909091 = 0.0374463 loss) | |
| I0623 17:08:46.993104 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.24926 (* 0.0909091 = 0.113569 loss) | |
| I0623 17:08:46.993118 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.367425 (* 0.0909091 = 0.0334022 loss) | |
| I0623 17:08:46.993131 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.416134 (* 0.0909091 = 0.0378304 loss) | |
| I0623 17:08:46.993144 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.668186 (* 0.0909091 = 0.0607442 loss) | |
| I0623 17:08:46.993158 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.459282 (* 0.0909091 = 0.0417529 loss) | |
| I0623 17:08:46.993171 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.356034 (* 0.0909091 = 0.0323668 loss) | |
| I0623 17:08:46.993185 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.00503363 (* 0.0909091 = 0.000457603 loss) | |
| I0623 17:08:46.993198 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.00020704 (* 0.0909091 = 1.88218e-05 loss) | |
| I0623 17:08:46.993212 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 4.54813e-05 (* 0.0909091 = 4.13467e-06 loss) | |
| I0623 17:08:46.993227 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 8.65767e-06 (* 0.0909091 = 7.87061e-07 loss) | |
| I0623 17:08:46.993240 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 9.03021e-06 (* 0.0909091 = 8.20928e-07 loss) | |
| I0623 17:08:46.993254 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 9.96904e-06 (* 0.0909091 = 9.06276e-07 loss) | |
| I0623 17:08:46.993268 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 1.08779e-06 (* 0.0909091 = 9.88898e-08 loss) | |
| I0623 17:08:46.993279 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.625 | |
| I0623 17:08:46.993291 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 17:08:46.993315 10365 solver.cpp:245] Train net output #149: total_confidence = 0.385994 | |
| I0623 17:08:46.993330 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.388949 | |
| I0623 17:08:46.993342 10365 sgd_solver.cpp:106] Iteration 13500, lr = 0.001 | |
| I0623 17:10:58.324808 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.5131 > 30) by scale factor 0.895173 | |
| I0623 17:13:23.921828 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 37.0036 > 30) by scale factor 0.810731 | |
| I0623 17:14:45.901037 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.0999 > 30) by scale factor 0.906348 | |
| I0623 17:15:10.055840 10365 solver.cpp:229] Iteration 14000, loss = 4.47865 | |
| I0623 17:15:10.055915 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.422018 | |
| I0623 17:15:10.055934 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 17:15:10.055948 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 17:15:10.055961 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.75 | |
| I0623 17:15:10.055974 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.625 | |
| I0623 17:15:10.055986 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 17:15:10.055999 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 17:15:10.056012 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 17:15:10.056025 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.125 | |
| I0623 17:15:10.056038 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 17:15:10.056051 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 17:15:10.056063 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 17:15:10.056077 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 17:15:10.056088 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 17:15:10.056099 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 17:15:10.056114 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 17:15:10.056126 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.5 | |
| I0623 17:15:10.056138 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 17:15:10.056149 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 17:15:10.056161 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 17:15:10.056174 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 17:15:10.056185 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:15:10.056196 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:15:10.056208 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.625 | |
| I0623 17:15:10.056219 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.770642 | |
| I0623 17:15:10.056236 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.74202 (* 0.3 = 0.522605 loss) | |
| I0623 17:15:10.056252 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.12538 (* 0.3 = 0.337615 loss) | |
| I0623 17:15:10.056265 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.490115 (* 0.0272727 = 0.0133668 loss) | |
| I0623 17:15:10.056279 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.32523 (* 0.0272727 = 0.0361426 loss) | |
| I0623 17:15:10.056293 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.22265 (* 0.0272727 = 0.0333449 loss) | |
| I0623 17:15:10.056308 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.39272 (* 0.0272727 = 0.0379834 loss) | |
| I0623 17:15:10.056326 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.59087 (* 0.0272727 = 0.07066 loss) | |
| I0623 17:15:10.056341 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.18838 (* 0.0272727 = 0.059683 loss) | |
| I0623 17:15:10.056354 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.91904 (* 0.0272727 = 0.0523374 loss) | |
| I0623 17:15:10.056368 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.6264 (* 0.0272727 = 0.071629 loss) | |
| I0623 17:15:10.056382 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.70826 (* 0.0272727 = 0.046589 loss) | |
| I0623 17:15:10.056396 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.30424 (* 0.0272727 = 0.0628429 loss) | |
| I0623 17:15:10.056411 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.98266 (* 0.0272727 = 0.0540726 loss) | |
| I0623 17:15:10.056423 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.23324 (* 0.0272727 = 0.0336338 loss) | |
| I0623 17:15:10.056469 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.20013 (* 0.0272727 = 0.0327308 loss) | |
| I0623 17:15:10.056484 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.74951 (* 0.0272727 = 0.0477139 loss) | |
| I0623 17:15:10.056499 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.892686 (* 0.0272727 = 0.024346 loss) | |
| I0623 17:15:10.056512 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.18568 (* 0.0272727 = 0.0323366 loss) | |
| I0623 17:15:10.056526 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.142273 (* 0.0272727 = 0.00388019 loss) | |
| I0623 17:15:10.056541 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.034059 (* 0.0272727 = 0.000928881 loss) | |
| I0623 17:15:10.056555 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00802562 (* 0.0272727 = 0.00021888 loss) | |
| I0623 17:15:10.056569 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00230327 (* 0.0272727 = 6.28164e-05 loss) | |
| I0623 17:15:10.056584 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000523589 (* 0.0272727 = 1.42797e-05 loss) | |
| I0623 17:15:10.056598 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000134901 (* 0.0272727 = 3.67913e-06 loss) | |
| I0623 17:15:10.056610 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.623853 | |
| I0623 17:15:10.056622 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 17:15:10.056634 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.625 | |
| I0623 17:15:10.056645 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 17:15:10.056658 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 17:15:10.056668 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 17:15:10.056680 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.25 | |
| I0623 17:15:10.056692 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 17:15:10.056704 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.125 | |
| I0623 17:15:10.056715 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 17:15:10.056726 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.125 | |
| I0623 17:15:10.056738 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 17:15:10.056749 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.75 | |
| I0623 17:15:10.056761 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 17:15:10.056772 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.5 | |
| I0623 17:15:10.056783 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.5 | |
| I0623 17:15:10.056795 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.625 | |
| I0623 17:15:10.056807 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 17:15:10.056818 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 17:15:10.056829 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 17:15:10.056841 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 17:15:10.056852 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:15:10.056864 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:15:10.056875 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.761364 | |
| I0623 17:15:10.056887 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.825688 | |
| I0623 17:15:10.056900 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.32556 (* 0.3 = 0.397668 loss) | |
| I0623 17:15:10.056915 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.845688 (* 0.3 = 0.253706 loss) | |
| I0623 17:15:10.056928 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.286714 (* 0.0272727 = 0.00781947 loss) | |
| I0623 17:15:10.056942 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 1.50007 (* 0.0272727 = 0.040911 loss) | |
| I0623 17:15:10.056967 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.603316 (* 0.0272727 = 0.0164541 loss) | |
| I0623 17:15:10.056982 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.08203 (* 0.0272727 = 0.0295099 loss) | |
| I0623 17:15:10.056996 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.55233 (* 0.0272727 = 0.0423362 loss) | |
| I0623 17:15:10.057010 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.00194 (* 0.0272727 = 0.0545985 loss) | |
| I0623 17:15:10.057024 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 2.40059 (* 0.0272727 = 0.0654705 loss) | |
| I0623 17:15:10.057037 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.80461 (* 0.0272727 = 0.0492166 loss) | |
| I0623 17:15:10.057051 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.3042 (* 0.0272727 = 0.035569 loss) | |
| I0623 17:15:10.057065 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.18593 (* 0.0272727 = 0.0596162 loss) | |
| I0623 17:15:10.057078 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.53451 (* 0.0272727 = 0.0418504 loss) | |
| I0623 17:15:10.057091 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.14194 (* 0.0272727 = 0.0311438 loss) | |
| I0623 17:15:10.057106 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.17471 (* 0.0272727 = 0.0320376 loss) | |
| I0623 17:15:10.057119 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.13836 (* 0.0272727 = 0.0310461 loss) | |
| I0623 17:15:10.057132 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.872707 (* 0.0272727 = 0.0238011 loss) | |
| I0623 17:15:10.057147 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.922972 (* 0.0272727 = 0.025172 loss) | |
| I0623 17:15:10.057163 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0148237 (* 0.0272727 = 0.000404283 loss) | |
| I0623 17:15:10.057178 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00180797 (* 0.0272727 = 4.93082e-05 loss) | |
| I0623 17:15:10.057193 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000382672 (* 0.0272727 = 1.04365e-05 loss) | |
| I0623 17:15:10.057206 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000174773 (* 0.0272727 = 4.76655e-06 loss) | |
| I0623 17:15:10.057220 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000277712 (* 0.0272727 = 7.57396e-06 loss) | |
| I0623 17:15:10.057235 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000166982 (* 0.0272727 = 4.55406e-06 loss) | |
| I0623 17:15:10.057246 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.807339 | |
| I0623 17:15:10.057258 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 17:15:10.057270 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 17:15:10.057281 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 17:15:10.057293 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 17:15:10.057304 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 17:15:10.057317 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.75 | |
| I0623 17:15:10.057327 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 17:15:10.057339 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 17:15:10.057351 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.5 | |
| I0623 17:15:10.057363 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 17:15:10.057379 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 17:15:10.057390 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.75 | |
| I0623 17:15:10.057402 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 17:15:10.057415 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 17:15:10.057426 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.5 | |
| I0623 17:15:10.057448 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.625 | |
| I0623 17:15:10.057461 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 17:15:10.057472 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 17:15:10.057484 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 17:15:10.057497 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 17:15:10.057509 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:15:10.057520 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:15:10.057533 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.880682 | |
| I0623 17:15:10.057543 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.981651 | |
| I0623 17:15:10.057557 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.580686 (* 1 = 0.580686 loss) | |
| I0623 17:15:10.057571 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.366701 (* 1 = 0.366701 loss) | |
| I0623 17:15:10.057585 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.164812 (* 0.0909091 = 0.0149829 loss) | |
| I0623 17:15:10.057600 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.667757 (* 0.0909091 = 0.0607052 loss) | |
| I0623 17:15:10.057613 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.107605 (* 0.0909091 = 0.00978228 loss) | |
| I0623 17:15:10.057627 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.235332 (* 0.0909091 = 0.0213938 loss) | |
| I0623 17:15:10.057641 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.307031 (* 0.0909091 = 0.0279119 loss) | |
| I0623 17:15:10.057656 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.570599 (* 0.0909091 = 0.0518727 loss) | |
| I0623 17:15:10.057668 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.923937 (* 0.0909091 = 0.0839942 loss) | |
| I0623 17:15:10.057682 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.640571 (* 0.0909091 = 0.0582337 loss) | |
| I0623 17:15:10.057696 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.12916 (* 0.0909091 = 0.102651 loss) | |
| I0623 17:15:10.057709 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.873837 (* 0.0909091 = 0.0794397 loss) | |
| I0623 17:15:10.057723 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.23948 (* 0.0909091 = 0.11268 loss) | |
| I0623 17:15:10.057736 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.754054 (* 0.0909091 = 0.0685503 loss) | |
| I0623 17:15:10.057750 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.13535 (* 0.0909091 = 0.103214 loss) | |
| I0623 17:15:10.057765 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.623961 (* 0.0909091 = 0.0567237 loss) | |
| I0623 17:15:10.057777 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.915952 (* 0.0909091 = 0.0832684 loss) | |
| I0623 17:15:10.057791 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.546888 (* 0.0909091 = 0.0497171 loss) | |
| I0623 17:15:10.057806 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0337383 (* 0.0909091 = 0.00306712 loss) | |
| I0623 17:15:10.057818 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00326406 (* 0.0909091 = 0.000296733 loss) | |
| I0623 17:15:10.057832 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000845686 (* 0.0909091 = 7.68806e-05 loss) | |
| I0623 17:15:10.057847 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000376596 (* 0.0909091 = 3.4236e-05 loss) | |
| I0623 17:15:10.057862 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 9.37831e-05 (* 0.0909091 = 8.52574e-06 loss) | |
| I0623 17:15:10.057874 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.89166e-05 (* 0.0909091 = 4.44696e-06 loss) | |
| I0623 17:15:10.057886 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 17:15:10.057898 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 17:15:10.057920 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0460726 | |
| I0623 17:15:10.057934 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0279589 | |
| I0623 17:15:10.057946 10365 sgd_solver.cpp:106] Iteration 14000, lr = 0.001 | |
| I0623 17:15:19.600751 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 48.8729 > 30) by scale factor 0.613837 | |
| I0623 17:17:30.564775 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 44.6816 > 30) by scale factor 0.671417 | |
| I0623 17:21:33.047780 10365 solver.cpp:229] Iteration 14500, loss = 4.54607 | |
| I0623 17:21:33.047876 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.396396 | |
| I0623 17:21:33.047895 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 17:21:33.047909 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 17:21:33.047922 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.125 | |
| I0623 17:21:33.047935 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 17:21:33.047947 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.125 | |
| I0623 17:21:33.047960 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.125 | |
| I0623 17:21:33.047972 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 17:21:33.047986 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.125 | |
| I0623 17:21:33.047997 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 17:21:33.048009 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 17:21:33.048022 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 17:21:33.048034 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 17:21:33.048046 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.375 | |
| I0623 17:21:33.048058 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 17:21:33.048069 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 17:21:33.048081 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.625 | |
| I0623 17:21:33.048094 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 17:21:33.048105 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 17:21:33.048120 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 17:21:33.048133 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 17:21:33.048146 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:21:33.048156 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:21:33.048168 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.596591 | |
| I0623 17:21:33.048179 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.774775 | |
| I0623 17:21:33.048197 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.75397 (* 0.3 = 0.526192 loss) | |
| I0623 17:21:33.048212 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.18 (* 0.3 = 0.353999 loss) | |
| I0623 17:21:33.048226 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.612364 (* 0.0272727 = 0.0167008 loss) | |
| I0623 17:21:33.048240 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.946152 (* 0.0272727 = 0.0258041 loss) | |
| I0623 17:21:33.048254 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.69753 (* 0.0272727 = 0.0735691 loss) | |
| I0623 17:21:33.048267 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.47203 (* 0.0272727 = 0.067419 loss) | |
| I0623 17:21:33.048281 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 3.06654 (* 0.0272727 = 0.0836329 loss) | |
| I0623 17:21:33.048295 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.00795 (* 0.0272727 = 0.0547624 loss) | |
| I0623 17:21:33.048310 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.01568 (* 0.0272727 = 0.054973 loss) | |
| I0623 17:21:33.048323 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.6923 (* 0.0272727 = 0.0734264 loss) | |
| I0623 17:21:33.048337 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.07022 (* 0.0272727 = 0.0564606 loss) | |
| I0623 17:21:33.048351 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.98509 (* 0.0272727 = 0.0541388 loss) | |
| I0623 17:21:33.048364 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.62614 (* 0.0272727 = 0.0443492 loss) | |
| I0623 17:21:33.048378 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.43982 (* 0.0272727 = 0.0392677 loss) | |
| I0623 17:21:33.048413 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.21834 (* 0.0272727 = 0.0605001 loss) | |
| I0623 17:21:33.048429 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.23821 (* 0.0272727 = 0.0337694 loss) | |
| I0623 17:21:33.048442 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.69569 (* 0.0272727 = 0.0462462 loss) | |
| I0623 17:21:33.048456 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.90872 (* 0.0272727 = 0.0247833 loss) | |
| I0623 17:21:33.048470 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.5742 (* 0.0272727 = 0.01566 loss) | |
| I0623 17:21:33.048485 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.624521 (* 0.0272727 = 0.0170324 loss) | |
| I0623 17:21:33.048498 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00501168 (* 0.0272727 = 0.000136682 loss) | |
| I0623 17:21:33.048512 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00127739 (* 0.0272727 = 3.4838e-05 loss) | |
| I0623 17:21:33.048527 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000937399 (* 0.0272727 = 2.55654e-05 loss) | |
| I0623 17:21:33.048542 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000248698 (* 0.0272727 = 6.78267e-06 loss) | |
| I0623 17:21:33.048553 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.540541 | |
| I0623 17:21:33.048566 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.75 | |
| I0623 17:21:33.048578 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 17:21:33.048589 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.5 | |
| I0623 17:21:33.048601 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 17:21:33.048612 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 17:21:33.048624 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 17:21:33.048635 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.25 | |
| I0623 17:21:33.048647 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.125 | |
| I0623 17:21:33.048660 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.25 | |
| I0623 17:21:33.048671 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 17:21:33.048682 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.75 | |
| I0623 17:21:33.048693 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.125 | |
| I0623 17:21:33.048705 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.25 | |
| I0623 17:21:33.048717 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.5 | |
| I0623 17:21:33.048728 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 17:21:33.048739 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 17:21:33.048751 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 17:21:33.048763 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 17:21:33.048774 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 17:21:33.048785 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 17:21:33.048796 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:21:33.048807 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:21:33.048820 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.704545 | |
| I0623 17:21:33.048830 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.837838 | |
| I0623 17:21:33.048845 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.30683 (* 0.3 = 0.392048 loss) | |
| I0623 17:21:33.048858 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.864131 (* 0.3 = 0.259239 loss) | |
| I0623 17:21:33.048872 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.513898 (* 0.0272727 = 0.0140154 loss) | |
| I0623 17:21:33.048885 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.267233 (* 0.0272727 = 0.00728818 loss) | |
| I0623 17:21:33.048910 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.732343 (* 0.0272727 = 0.019973 loss) | |
| I0623 17:21:33.048925 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.05662 (* 0.0272727 = 0.028817 loss) | |
| I0623 17:21:33.048939 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.47354 (* 0.0272727 = 0.0401874 loss) | |
| I0623 17:21:33.048954 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.9493 (* 0.0272727 = 0.0531628 loss) | |
| I0623 17:21:33.048967 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.83533 (* 0.0272727 = 0.0500544 loss) | |
| I0623 17:21:33.048981 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.44798 (* 0.0272727 = 0.0667632 loss) | |
| I0623 17:21:33.048995 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.91286 (* 0.0272727 = 0.0521688 loss) | |
| I0623 17:21:33.049007 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.83017 (* 0.0272727 = 0.0499137 loss) | |
| I0623 17:21:33.049021 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.60034 (* 0.0272727 = 0.0436457 loss) | |
| I0623 17:21:33.049034 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 2.31992 (* 0.0272727 = 0.0632704 loss) | |
| I0623 17:21:33.049048 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 2.29135 (* 0.0272727 = 0.0624914 loss) | |
| I0623 17:21:33.049062 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.990566 (* 0.0272727 = 0.0270154 loss) | |
| I0623 17:21:33.049075 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.31986 (* 0.0272727 = 0.0359961 loss) | |
| I0623 17:21:33.049088 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.743633 (* 0.0272727 = 0.0202809 loss) | |
| I0623 17:21:33.049103 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.445779 (* 0.0272727 = 0.0121576 loss) | |
| I0623 17:21:33.049116 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.226735 (* 0.0272727 = 0.00618369 loss) | |
| I0623 17:21:33.049129 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00935476 (* 0.0272727 = 0.00025513 loss) | |
| I0623 17:21:33.049144 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00269293 (* 0.0272727 = 7.34435e-05 loss) | |
| I0623 17:21:33.049157 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00051283 (* 0.0272727 = 1.39863e-05 loss) | |
| I0623 17:21:33.049175 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000121358 (* 0.0272727 = 3.30977e-06 loss) | |
| I0623 17:21:33.049188 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.810811 | |
| I0623 17:21:33.049201 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 17:21:33.049211 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 17:21:33.049223 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 17:21:33.049234 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 17:21:33.049247 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 17:21:33.049257 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 17:21:33.049269 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 17:21:33.049280 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.5 | |
| I0623 17:21:33.049291 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 17:21:33.049304 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 17:21:33.049314 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.5 | |
| I0623 17:21:33.049325 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 17:21:33.049337 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.25 | |
| I0623 17:21:33.049348 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 17:21:33.049360 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 17:21:33.049371 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 17:21:33.049392 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 17:21:33.049407 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 17:21:33.049418 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 17:21:33.049430 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 17:21:33.049441 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:21:33.049453 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:21:33.049464 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.880682 | |
| I0623 17:21:33.049476 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.972973 | |
| I0623 17:21:33.049489 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.614184 (* 1 = 0.614184 loss) | |
| I0623 17:21:33.049504 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.39858 (* 1 = 0.39858 loss) | |
| I0623 17:21:33.049517 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0210263 (* 0.0909091 = 0.00191149 loss) | |
| I0623 17:21:33.049531 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0213704 (* 0.0909091 = 0.00194276 loss) | |
| I0623 17:21:33.049546 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0539277 (* 0.0909091 = 0.00490252 loss) | |
| I0623 17:21:33.049559 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.766735 (* 0.0909091 = 0.0697032 loss) | |
| I0623 17:21:33.049573 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.196064 (* 0.0909091 = 0.017824 loss) | |
| I0623 17:21:33.049587 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.490421 (* 0.0909091 = 0.0445838 loss) | |
| I0623 17:21:33.049600 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.313654 (* 0.0909091 = 0.028514 loss) | |
| I0623 17:21:33.049614 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.22119 (* 0.0909091 = 0.111018 loss) | |
| I0623 17:21:33.049628 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.523083 (* 0.0909091 = 0.047553 loss) | |
| I0623 17:21:33.049641 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.28744 (* 0.0909091 = 0.11704 loss) | |
| I0623 17:21:33.049655 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.4624 (* 0.0909091 = 0.132945 loss) | |
| I0623 17:21:33.049669 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.62791 (* 0.0909091 = 0.147992 loss) | |
| I0623 17:21:33.049682 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.49462 (* 0.0909091 = 0.135874 loss) | |
| I0623 17:21:33.049696 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.894309 (* 0.0909091 = 0.0813008 loss) | |
| I0623 17:21:33.049710 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.05277 (* 0.0909091 = 0.0957061 loss) | |
| I0623 17:21:33.049724 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.516391 (* 0.0909091 = 0.0469446 loss) | |
| I0623 17:21:33.049738 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.145954 (* 0.0909091 = 0.0132686 loss) | |
| I0623 17:21:33.049751 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0983883 (* 0.0909091 = 0.00894439 loss) | |
| I0623 17:21:33.049765 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0124972 (* 0.0909091 = 0.00113611 loss) | |
| I0623 17:21:33.049779 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00172688 (* 0.0909091 = 0.000156989 loss) | |
| I0623 17:21:33.049793 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000739196 (* 0.0909091 = 6.71997e-05 loss) | |
| I0623 17:21:33.049808 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 6.29569e-05 (* 0.0909091 = 5.72336e-06 loss) | |
| I0623 17:21:33.049819 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 17:21:33.049831 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 17:21:33.049842 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0357764 | |
| I0623 17:21:33.049865 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0192301 | |
| I0623 17:21:33.049880 10365 sgd_solver.cpp:106] Iteration 14500, lr = 0.001 | |
| I0623 17:22:39.306763 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 39.8523 > 30) by scale factor 0.75278 | |
| I0623 17:23:29.122648 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.9898 > 30) by scale factor 0.909372 | |
| I0623 17:24:05.912299 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 39.7546 > 30) by scale factor 0.75463 | |
| I0623 17:24:38.109985 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 41.247 > 30) by scale factor 0.727326 | |
| I0623 17:25:17.954161 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 106.089 > 30) by scale factor 0.282781 | |
| I0623 17:27:55.729800 10365 solver.cpp:338] Iteration 15000, Testing net (#0) | |
| I0623 17:28:52.903836 10365 solver.cpp:393] Test loss: 3.9476 | |
| I0623 17:28:52.903957 10365 solver.cpp:406] Test net output #0: loss1/accuracy = 0.513129 | |
| I0623 17:28:52.903978 10365 solver.cpp:406] Test net output #1: loss1/accuracy01 = 0.933 | |
| I0623 17:28:52.903991 10365 solver.cpp:406] Test net output #2: loss1/accuracy02 = 0.798 | |
| I0623 17:28:52.904006 10365 solver.cpp:406] Test net output #3: loss1/accuracy03 = 0.562 | |
| I0623 17:28:52.904017 10365 solver.cpp:406] Test net output #4: loss1/accuracy04 = 0.474 | |
| I0623 17:28:52.904029 10365 solver.cpp:406] Test net output #5: loss1/accuracy05 = 0.405 | |
| I0623 17:28:52.904042 10365 solver.cpp:406] Test net output #6: loss1/accuracy06 = 0.434 | |
| I0623 17:28:52.904053 10365 solver.cpp:406] Test net output #7: loss1/accuracy07 = 0.406 | |
| I0623 17:28:52.904065 10365 solver.cpp:406] Test net output #8: loss1/accuracy08 = 0.487 | |
| I0623 17:28:52.904078 10365 solver.cpp:406] Test net output #9: loss1/accuracy09 = 0.449 | |
| I0623 17:28:52.904091 10365 solver.cpp:406] Test net output #10: loss1/accuracy10 = 0.412 | |
| I0623 17:28:52.904103 10365 solver.cpp:406] Test net output #11: loss1/accuracy11 = 0.41 | |
| I0623 17:28:52.904115 10365 solver.cpp:406] Test net output #12: loss1/accuracy12 = 0.486 | |
| I0623 17:28:52.904127 10365 solver.cpp:406] Test net output #13: loss1/accuracy13 = 0.594 | |
| I0623 17:28:52.904139 10365 solver.cpp:406] Test net output #14: loss1/accuracy14 = 0.676 | |
| I0623 17:28:52.904150 10365 solver.cpp:406] Test net output #15: loss1/accuracy15 = 0.768 | |
| I0623 17:28:52.904161 10365 solver.cpp:406] Test net output #16: loss1/accuracy16 = 0.832 | |
| I0623 17:28:52.904172 10365 solver.cpp:406] Test net output #17: loss1/accuracy17 = 0.905 | |
| I0623 17:28:52.904183 10365 solver.cpp:406] Test net output #18: loss1/accuracy18 = 0.95 | |
| I0623 17:28:52.904194 10365 solver.cpp:406] Test net output #19: loss1/accuracy19 = 0.972 | |
| I0623 17:28:52.904206 10365 solver.cpp:406] Test net output #20: loss1/accuracy20 = 0.987 | |
| I0623 17:28:52.904217 10365 solver.cpp:406] Test net output #21: loss1/accuracy21 = 0.999 | |
| I0623 17:28:52.904228 10365 solver.cpp:406] Test net output #22: loss1/accuracy22 = 1 | |
| I0623 17:28:52.904239 10365 solver.cpp:406] Test net output #23: loss1/accuracy_incl_empty = 0.701772 | |
| I0623 17:28:52.904252 10365 solver.cpp:406] Test net output #24: loss1/accuracy_top3 = 0.848627 | |
| I0623 17:28:52.904270 10365 solver.cpp:406] Test net output #25: loss1/cross_entropy_loss = 1.39484 (* 0.3 = 0.418451 loss) | |
| I0623 17:28:52.904285 10365 solver.cpp:406] Test net output #26: loss1/cross_entropy_loss_incl_empty = 0.857432 (* 0.3 = 0.257229 loss) | |
| I0623 17:28:52.904300 10365 solver.cpp:406] Test net output #27: loss1/loss01 = 0.308384 (* 0.0272727 = 0.00841047 loss) | |
| I0623 17:28:52.904314 10365 solver.cpp:406] Test net output #28: loss1/loss02 = 0.718276 (* 0.0272727 = 0.0195893 loss) | |
| I0623 17:28:52.904327 10365 solver.cpp:406] Test net output #29: loss1/loss03 = 1.33856 (* 0.0272727 = 0.0365062 loss) | |
| I0623 17:28:52.904341 10365 solver.cpp:406] Test net output #30: loss1/loss04 = 1.53309 (* 0.0272727 = 0.0418116 loss) | |
| I0623 17:28:52.904355 10365 solver.cpp:406] Test net output #31: loss1/loss05 = 1.67412 (* 0.0272727 = 0.0456578 loss) | |
| I0623 17:28:52.904369 10365 solver.cpp:406] Test net output #32: loss1/loss06 = 1.76757 (* 0.0272727 = 0.0482065 loss) | |
| I0623 17:28:52.904382 10365 solver.cpp:406] Test net output #33: loss1/loss07 = 1.79086 (* 0.0272727 = 0.0488416 loss) | |
| I0623 17:28:52.904397 10365 solver.cpp:406] Test net output #34: loss1/loss08 = 1.6213 (* 0.0272727 = 0.0442172 loss) | |
| I0623 17:28:52.904409 10365 solver.cpp:406] Test net output #35: loss1/loss09 = 1.69879 (* 0.0272727 = 0.0463306 loss) | |
| I0623 17:28:52.904422 10365 solver.cpp:406] Test net output #36: loss1/loss10 = 1.7343 (* 0.0272727 = 0.0472992 loss) | |
| I0623 17:28:52.904436 10365 solver.cpp:406] Test net output #37: loss1/loss11 = 1.81281 (* 0.0272727 = 0.0494402 loss) | |
| I0623 17:28:52.904449 10365 solver.cpp:406] Test net output #38: loss1/loss12 = 1.51392 (* 0.0272727 = 0.0412887 loss) | |
| I0623 17:28:52.904481 10365 solver.cpp:406] Test net output #39: loss1/loss13 = 1.22069 (* 0.0272727 = 0.0332915 loss) | |
| I0623 17:28:52.904496 10365 solver.cpp:406] Test net output #40: loss1/loss14 = 0.941574 (* 0.0272727 = 0.0256793 loss) | |
| I0623 17:28:52.904510 10365 solver.cpp:406] Test net output #41: loss1/loss15 = 0.675854 (* 0.0272727 = 0.0184324 loss) | |
| I0623 17:28:52.904523 10365 solver.cpp:406] Test net output #42: loss1/loss16 = 0.514589 (* 0.0272727 = 0.0140342 loss) | |
| I0623 17:28:52.904537 10365 solver.cpp:406] Test net output #43: loss1/loss17 = 0.324097 (* 0.0272727 = 0.00883901 loss) | |
| I0623 17:28:52.904551 10365 solver.cpp:406] Test net output #44: loss1/loss18 = 0.190667 (* 0.0272727 = 0.0052 loss) | |
| I0623 17:28:52.904566 10365 solver.cpp:406] Test net output #45: loss1/loss19 = 0.120566 (* 0.0272727 = 0.00328816 loss) | |
| I0623 17:28:52.904578 10365 solver.cpp:406] Test net output #46: loss1/loss20 = 0.0691191 (* 0.0272727 = 0.00188507 loss) | |
| I0623 17:28:52.904592 10365 solver.cpp:406] Test net output #47: loss1/loss21 = 0.00695473 (* 0.0272727 = 0.000189675 loss) | |
| I0623 17:28:52.904605 10365 solver.cpp:406] Test net output #48: loss1/loss22 = 6.57153e-05 (* 0.0272727 = 1.79223e-06 loss) | |
| I0623 17:28:52.904618 10365 solver.cpp:406] Test net output #49: loss2/accuracy = 0.625939 | |
| I0623 17:28:52.904629 10365 solver.cpp:406] Test net output #50: loss2/accuracy01 = 0.975 | |
| I0623 17:28:52.904640 10365 solver.cpp:406] Test net output #51: loss2/accuracy02 = 0.95 | |
| I0623 17:28:52.904652 10365 solver.cpp:406] Test net output #52: loss2/accuracy03 = 0.87 | |
| I0623 17:28:52.904664 10365 solver.cpp:406] Test net output #53: loss2/accuracy04 = 0.734 | |
| I0623 17:28:52.904675 10365 solver.cpp:406] Test net output #54: loss2/accuracy05 = 0.554 | |
| I0623 17:28:52.904685 10365 solver.cpp:406] Test net output #55: loss2/accuracy06 = 0.511 | |
| I0623 17:28:52.904696 10365 solver.cpp:406] Test net output #56: loss2/accuracy07 = 0.498 | |
| I0623 17:28:52.904707 10365 solver.cpp:406] Test net output #57: loss2/accuracy08 = 0.526 | |
| I0623 17:28:52.904718 10365 solver.cpp:406] Test net output #58: loss2/accuracy09 = 0.494 | |
| I0623 17:28:52.904731 10365 solver.cpp:406] Test net output #59: loss2/accuracy10 = 0.453 | |
| I0623 17:28:52.904742 10365 solver.cpp:406] Test net output #60: loss2/accuracy11 = 0.433 | |
| I0623 17:28:52.904752 10365 solver.cpp:406] Test net output #61: loss2/accuracy12 = 0.518 | |
| I0623 17:28:52.904763 10365 solver.cpp:406] Test net output #62: loss2/accuracy13 = 0.604 | |
| I0623 17:28:52.904774 10365 solver.cpp:406] Test net output #63: loss2/accuracy14 = 0.695 | |
| I0623 17:28:52.904785 10365 solver.cpp:406] Test net output #64: loss2/accuracy15 = 0.778 | |
| I0623 17:28:52.904796 10365 solver.cpp:406] Test net output #65: loss2/accuracy16 = 0.836 | |
| I0623 17:28:52.904808 10365 solver.cpp:406] Test net output #66: loss2/accuracy17 = 0.904 | |
| I0623 17:28:52.904819 10365 solver.cpp:406] Test net output #67: loss2/accuracy18 = 0.95 | |
| I0623 17:28:52.904829 10365 solver.cpp:406] Test net output #68: loss2/accuracy19 = 0.972 | |
| I0623 17:28:52.904840 10365 solver.cpp:406] Test net output #69: loss2/accuracy20 = 0.987 | |
| I0623 17:28:52.904851 10365 solver.cpp:406] Test net output #70: loss2/accuracy21 = 0.999 | |
| I0623 17:28:52.904863 10365 solver.cpp:406] Test net output #71: loss2/accuracy22 = 1 | |
| I0623 17:28:52.904875 10365 solver.cpp:406] Test net output #72: loss2/accuracy_incl_empty = 0.763 | |
| I0623 17:28:52.904886 10365 solver.cpp:406] Test net output #73: loss2/accuracy_top3 = 0.910811 | |
| I0623 17:28:52.904899 10365 solver.cpp:406] Test net output #74: loss2/cross_entropy_loss = 1.06519 (* 0.3 = 0.319558 loss) | |
| I0623 17:28:52.904912 10365 solver.cpp:406] Test net output #75: loss2/cross_entropy_loss_incl_empty = 0.669555 (* 0.3 = 0.200866 loss) | |
| I0623 17:28:52.904927 10365 solver.cpp:406] Test net output #76: loss2/loss01 = 0.186428 (* 0.0272727 = 0.0050844 loss) | |
| I0623 17:28:52.904940 10365 solver.cpp:406] Test net output #77: loss2/loss02 = 0.279208 (* 0.0272727 = 0.00761477 loss) | |
| I0623 17:28:52.904964 10365 solver.cpp:406] Test net output #78: loss2/loss03 = 0.546761 (* 0.0272727 = 0.0149117 loss) | |
| I0623 17:28:52.904983 10365 solver.cpp:406] Test net output #79: loss2/loss04 = 0.891898 (* 0.0272727 = 0.0243245 loss) | |
| I0623 17:28:52.904997 10365 solver.cpp:406] Test net output #80: loss2/loss05 = 1.19127 (* 0.0272727 = 0.0324892 loss) | |
| I0623 17:28:52.905010 10365 solver.cpp:406] Test net output #81: loss2/loss06 = 1.41573 (* 0.0272727 = 0.0386108 loss) | |
| I0623 17:28:52.905025 10365 solver.cpp:406] Test net output #82: loss2/loss07 = 1.49779 (* 0.0272727 = 0.0408489 loss) | |
| I0623 17:28:52.905037 10365 solver.cpp:406] Test net output #83: loss2/loss08 = 1.43084 (* 0.0272727 = 0.039023 loss) | |
| I0623 17:28:52.905051 10365 solver.cpp:406] Test net output #84: loss2/loss09 = 1.4965 (* 0.0272727 = 0.0408135 loss) | |
| I0623 17:28:52.905064 10365 solver.cpp:406] Test net output #85: loss2/loss10 = 1.57877 (* 0.0272727 = 0.0430574 loss) | |
| I0623 17:28:52.905077 10365 solver.cpp:406] Test net output #86: loss2/loss11 = 1.63139 (* 0.0272727 = 0.0444924 loss) | |
| I0623 17:28:52.905091 10365 solver.cpp:406] Test net output #87: loss2/loss12 = 1.35325 (* 0.0272727 = 0.0369068 loss) | |
| I0623 17:28:52.905103 10365 solver.cpp:406] Test net output #88: loss2/loss13 = 1.12253 (* 0.0272727 = 0.0306145 loss) | |
| I0623 17:28:52.905117 10365 solver.cpp:406] Test net output #89: loss2/loss14 = 0.85908 (* 0.0272727 = 0.0234294 loss) | |
| I0623 17:28:52.905129 10365 solver.cpp:406] Test net output #90: loss2/loss15 = 0.623017 (* 0.0272727 = 0.0169914 loss) | |
| I0623 17:28:52.905143 10365 solver.cpp:406] Test net output #91: loss2/loss16 = 0.460758 (* 0.0272727 = 0.0125661 loss) | |
| I0623 17:28:52.905156 10365 solver.cpp:406] Test net output #92: loss2/loss17 = 0.306844 (* 0.0272727 = 0.00836848 loss) | |
| I0623 17:28:52.905170 10365 solver.cpp:406] Test net output #93: loss2/loss18 = 0.16984 (* 0.0272727 = 0.004632 loss) | |
| I0623 17:28:52.905184 10365 solver.cpp:406] Test net output #94: loss2/loss19 = 0.110454 (* 0.0272727 = 0.00301238 loss) | |
| I0623 17:28:52.905197 10365 solver.cpp:406] Test net output #95: loss2/loss20 = 0.0623378 (* 0.0272727 = 0.00170012 loss) | |
| I0623 17:28:52.905210 10365 solver.cpp:406] Test net output #96: loss2/loss21 = 0.00793115 (* 0.0272727 = 0.000216304 loss) | |
| I0623 17:28:52.905225 10365 solver.cpp:406] Test net output #97: loss2/loss22 = 8.68589e-05 (* 0.0272727 = 2.36888e-06 loss) | |
| I0623 17:28:52.905236 10365 solver.cpp:406] Test net output #98: loss3/accuracy = 0.861131 | |
| I0623 17:28:52.905247 10365 solver.cpp:406] Test net output #99: loss3/accuracy01 = 0.982 | |
| I0623 17:28:52.905258 10365 solver.cpp:406] Test net output #100: loss3/accuracy02 = 0.973 | |
| I0623 17:28:52.905270 10365 solver.cpp:406] Test net output #101: loss3/accuracy03 = 0.962 | |
| I0623 17:28:52.905282 10365 solver.cpp:406] Test net output #102: loss3/accuracy04 = 0.946 | |
| I0623 17:28:52.905292 10365 solver.cpp:406] Test net output #103: loss3/accuracy05 = 0.938 | |
| I0623 17:28:52.905304 10365 solver.cpp:406] Test net output #104: loss3/accuracy06 = 0.91 | |
| I0623 17:28:52.905318 10365 solver.cpp:406] Test net output #105: loss3/accuracy07 = 0.905 | |
| I0623 17:28:52.905329 10365 solver.cpp:406] Test net output #106: loss3/accuracy08 = 0.874 | |
| I0623 17:28:52.905341 10365 solver.cpp:406] Test net output #107: loss3/accuracy09 = 0.794 | |
| I0623 17:28:52.905352 10365 solver.cpp:406] Test net output #108: loss3/accuracy10 = 0.711 | |
| I0623 17:28:52.905364 10365 solver.cpp:406] Test net output #109: loss3/accuracy11 = 0.611 | |
| I0623 17:28:52.905375 10365 solver.cpp:406] Test net output #110: loss3/accuracy12 = 0.652 | |
| I0623 17:28:52.905385 10365 solver.cpp:406] Test net output #111: loss3/accuracy13 = 0.689 | |
| I0623 17:28:52.905396 10365 solver.cpp:406] Test net output #112: loss3/accuracy14 = 0.744 | |
| I0623 17:28:52.905407 10365 solver.cpp:406] Test net output #113: loss3/accuracy15 = 0.813 | |
| I0623 17:28:52.905418 10365 solver.cpp:406] Test net output #114: loss3/accuracy16 = 0.871 | |
| I0623 17:28:52.905439 10365 solver.cpp:406] Test net output #115: loss3/accuracy17 = 0.923 | |
| I0623 17:28:52.905452 10365 solver.cpp:406] Test net output #116: loss3/accuracy18 = 0.966 | |
| I0623 17:28:52.905463 10365 solver.cpp:406] Test net output #117: loss3/accuracy19 = 0.977 | |
| I0623 17:28:52.905474 10365 solver.cpp:406] Test net output #118: loss3/accuracy20 = 0.987 | |
| I0623 17:28:52.905486 10365 solver.cpp:406] Test net output #119: loss3/accuracy21 = 0.999 | |
| I0623 17:28:52.905498 10365 solver.cpp:406] Test net output #120: loss3/accuracy22 = 1 | |
| I0623 17:28:52.905508 10365 solver.cpp:406] Test net output #121: loss3/accuracy_incl_empty = 0.907955 | |
| I0623 17:28:52.905519 10365 solver.cpp:406] Test net output #122: loss3/accuracy_top3 = 0.970722 | |
| I0623 17:28:52.905534 10365 solver.cpp:406] Test net output #123: loss3/cross_entropy_loss = 0.523981 (* 1 = 0.523981 loss) | |
| I0623 17:28:52.905547 10365 solver.cpp:406] Test net output #124: loss3/cross_entropy_loss_incl_empty = 0.334736 (* 1 = 0.334736 loss) | |
| I0623 17:28:52.905560 10365 solver.cpp:406] Test net output #125: loss3/loss01 = 0.144683 (* 0.0909091 = 0.013153 loss) | |
| I0623 17:28:52.905575 10365 solver.cpp:406] Test net output #126: loss3/loss02 = 0.164218 (* 0.0909091 = 0.0149289 loss) | |
| I0623 17:28:52.905587 10365 solver.cpp:406] Test net output #127: loss3/loss03 = 0.272165 (* 0.0909091 = 0.0247422 loss) | |
| I0623 17:28:52.905601 10365 solver.cpp:406] Test net output #128: loss3/loss04 = 0.33437 (* 0.0909091 = 0.0303973 loss) | |
| I0623 17:28:52.905614 10365 solver.cpp:406] Test net output #129: loss3/loss05 = 0.355665 (* 0.0909091 = 0.0323332 loss) | |
| I0623 17:28:52.905627 10365 solver.cpp:406] Test net output #130: loss3/loss06 = 0.479199 (* 0.0909091 = 0.0435635 loss) | |
| I0623 17:28:52.905640 10365 solver.cpp:406] Test net output #131: loss3/loss07 = 0.515476 (* 0.0909091 = 0.0468615 loss) | |
| I0623 17:28:52.905654 10365 solver.cpp:406] Test net output #132: loss3/loss08 = 0.554469 (* 0.0909091 = 0.0504063 loss) | |
| I0623 17:28:52.905668 10365 solver.cpp:406] Test net output #133: loss3/loss09 = 0.714712 (* 0.0909091 = 0.0649738 loss) | |
| I0623 17:28:52.905680 10365 solver.cpp:406] Test net output #134: loss3/loss10 = 0.896266 (* 0.0909091 = 0.0814787 loss) | |
| I0623 17:28:52.905694 10365 solver.cpp:406] Test net output #135: loss3/loss11 = 1.0653 (* 0.0909091 = 0.0968455 loss) | |
| I0623 17:28:52.905707 10365 solver.cpp:406] Test net output #136: loss3/loss12 = 0.913871 (* 0.0909091 = 0.0830792 loss) | |
| I0623 17:28:52.905720 10365 solver.cpp:406] Test net output #137: loss3/loss13 = 0.831192 (* 0.0909091 = 0.0755629 loss) | |
| I0623 17:28:52.905733 10365 solver.cpp:406] Test net output #138: loss3/loss14 = 0.65294 (* 0.0909091 = 0.0593582 loss) | |
| I0623 17:28:52.905747 10365 solver.cpp:406] Test net output #139: loss3/loss15 = 0.477008 (* 0.0909091 = 0.0433644 loss) | |
| I0623 17:28:52.905761 10365 solver.cpp:406] Test net output #140: loss3/loss16 = 0.358972 (* 0.0909091 = 0.0326338 loss) | |
| I0623 17:28:52.905773 10365 solver.cpp:406] Test net output #141: loss3/loss17 = 0.208151 (* 0.0909091 = 0.0189228 loss) | |
| I0623 17:28:52.905786 10365 solver.cpp:406] Test net output #142: loss3/loss18 = 0.124898 (* 0.0909091 = 0.0113543 loss) | |
| I0623 17:28:52.905800 10365 solver.cpp:406] Test net output #143: loss3/loss19 = 0.0697284 (* 0.0909091 = 0.00633895 loss) | |
| I0623 17:28:52.905814 10365 solver.cpp:406] Test net output #144: loss3/loss20 = 0.0415507 (* 0.0909091 = 0.00377734 loss) | |
| I0623 17:28:52.905827 10365 solver.cpp:406] Test net output #145: loss3/loss21 = 0.00596782 (* 0.0909091 = 0.000542529 loss) | |
| I0623 17:28:52.905841 10365 solver.cpp:406] Test net output #146: loss3/loss22 = 8.71956e-05 (* 0.0909091 = 7.92687e-06 loss) | |
| I0623 17:28:52.905854 10365 solver.cpp:406] Test net output #147: total_accuracy = 0.377 | |
| I0623 17:28:52.905866 10365 solver.cpp:406] Test net output #148: total_accuracy_not_rec = 0.237 | |
| I0623 17:28:52.905877 10365 solver.cpp:406] Test net output #149: total_confidence = 0.221642 | |
| I0623 17:28:52.905897 10365 solver.cpp:406] Test net output #150: total_confidence_not_rec = 0.141853 | |
| I0623 17:28:53.263267 10365 solver.cpp:229] Iteration 15000, loss = 4.53586 | |
| I0623 17:28:53.263317 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.38835 | |
| I0623 17:28:53.263335 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 17:28:53.263348 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 17:28:53.263362 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.625 | |
| I0623 17:28:53.263375 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.25 | |
| I0623 17:28:53.263386 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 17:28:53.263398 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 17:28:53.263411 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 17:28:53.263423 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 17:28:53.263437 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 17:28:53.263448 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 17:28:53.263460 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 17:28:53.263473 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 17:28:53.263484 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 17:28:53.263496 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 17:28:53.263507 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 17:28:53.263520 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 1 | |
| I0623 17:28:53.263531 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 17:28:53.263543 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 17:28:53.263555 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 17:28:53.263566 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 17:28:53.263577 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:28:53.263589 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:28:53.263618 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.630682 | |
| I0623 17:28:53.263633 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.776699 | |
| I0623 17:28:53.263648 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.57546 (* 0.3 = 0.472639 loss) | |
| I0623 17:28:53.263662 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.96189 (* 0.3 = 0.288567 loss) | |
| I0623 17:28:53.263677 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.583676 (* 0.0272727 = 0.0159184 loss) | |
| I0623 17:28:53.263691 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.896728 (* 0.0272727 = 0.0244562 loss) | |
| I0623 17:28:53.263705 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.28521 (* 0.0272727 = 0.0350513 loss) | |
| I0623 17:28:53.263718 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.86228 (* 0.0272727 = 0.0507895 loss) | |
| I0623 17:28:53.263732 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.24534 (* 0.0272727 = 0.0612366 loss) | |
| I0623 17:28:53.263746 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.54253 (* 0.0272727 = 0.0693417 loss) | |
| I0623 17:28:53.263761 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.76764 (* 0.0272727 = 0.0482083 loss) | |
| I0623 17:28:53.263773 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.44585 (* 0.0272727 = 0.0394322 loss) | |
| I0623 17:28:53.263787 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.51151 (* 0.0272727 = 0.0412229 loss) | |
| I0623 17:28:53.263802 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.4995 (* 0.0272727 = 0.0408955 loss) | |
| I0623 17:28:53.263814 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.44033 (* 0.0272727 = 0.0665543 loss) | |
| I0623 17:28:53.263851 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.84161 (* 0.0272727 = 0.0502257 loss) | |
| I0623 17:28:53.263866 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.18112 (* 0.0272727 = 0.0322125 loss) | |
| I0623 17:28:53.263880 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.959027 (* 0.0272727 = 0.0261553 loss) | |
| I0623 17:28:53.263895 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.22014 (* 0.0272727 = 0.0332766 loss) | |
| I0623 17:28:53.263908 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.182926 (* 0.0272727 = 0.00498889 loss) | |
| I0623 17:28:53.263922 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.012054 (* 0.0272727 = 0.000328746 loss) | |
| I0623 17:28:53.263936 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00180629 (* 0.0272727 = 4.92624e-05 loss) | |
| I0623 17:28:53.263950 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00038362 (* 0.0272727 = 1.04624e-05 loss) | |
| I0623 17:28:53.263965 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000140408 (* 0.0272727 = 3.82932e-06 loss) | |
| I0623 17:28:53.263979 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 8.50864e-06 (* 0.0272727 = 2.32054e-07 loss) | |
| I0623 17:28:53.263993 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 5.96052e-06 (* 0.0272727 = 1.6256e-07 loss) | |
| I0623 17:28:53.264005 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.592233 | |
| I0623 17:28:53.264021 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 17:28:53.264034 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 17:28:53.264045 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 17:28:53.264056 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.25 | |
| I0623 17:28:53.264068 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 17:28:53.264080 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 17:28:53.264091 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 17:28:53.264102 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.625 | |
| I0623 17:28:53.264114 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.625 | |
| I0623 17:28:53.264125 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.625 | |
| I0623 17:28:53.264139 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 17:28:53.264150 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.25 | |
| I0623 17:28:53.264163 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 17:28:53.264173 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 17:28:53.264185 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 17:28:53.264196 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 17:28:53.264209 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 17:28:53.264219 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 17:28:53.264230 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 17:28:53.264242 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 17:28:53.264253 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:28:53.264264 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:28:53.264276 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.761364 | |
| I0623 17:28:53.264286 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.893204 | |
| I0623 17:28:53.264300 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.2147 (* 0.3 = 0.364411 loss) | |
| I0623 17:28:53.264314 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.723021 (* 0.3 = 0.216906 loss) | |
| I0623 17:28:53.264339 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.215374 (* 0.0272727 = 0.00587383 loss) | |
| I0623 17:28:53.264354 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.290659 (* 0.0272727 = 0.00792707 loss) | |
| I0623 17:28:53.264369 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.19357 (* 0.0272727 = 0.032552 loss) | |
| I0623 17:28:53.264381 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.71224 (* 0.0272727 = 0.0466974 loss) | |
| I0623 17:28:53.264395 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.03056 (* 0.0272727 = 0.055379 loss) | |
| I0623 17:28:53.264410 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.27185 (* 0.0272727 = 0.0619595 loss) | |
| I0623 17:28:53.264422 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.42014 (* 0.0272727 = 0.038731 loss) | |
| I0623 17:28:53.264436 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.21442 (* 0.0272727 = 0.0331204 loss) | |
| I0623 17:28:53.264449 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.21967 (* 0.0272727 = 0.0332636 loss) | |
| I0623 17:28:53.264463 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.48561 (* 0.0272727 = 0.0405166 loss) | |
| I0623 17:28:53.264477 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.90224 (* 0.0272727 = 0.0518792 loss) | |
| I0623 17:28:53.264489 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.8501 (* 0.0272727 = 0.0504573 loss) | |
| I0623 17:28:53.264503 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.40896 (* 0.0272727 = 0.0384261 loss) | |
| I0623 17:28:53.264516 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.6928 (* 0.0272727 = 0.0188945 loss) | |
| I0623 17:28:53.264529 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.20926 (* 0.0272727 = 0.0329798 loss) | |
| I0623 17:28:53.264544 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.128833 (* 0.0272727 = 0.00351363 loss) | |
| I0623 17:28:53.264557 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.00605171 (* 0.0272727 = 0.000165047 loss) | |
| I0623 17:28:53.264571 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.000326895 (* 0.0272727 = 8.91531e-06 loss) | |
| I0623 17:28:53.264585 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 6.93183e-05 (* 0.0272727 = 1.8905e-06 loss) | |
| I0623 17:28:53.264600 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 1.63171e-05 (* 0.0272727 = 4.45013e-07 loss) | |
| I0623 17:28:53.264612 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 7.45073e-06 (* 0.0272727 = 2.03202e-07 loss) | |
| I0623 17:28:53.264626 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 7.5996e-07 (* 0.0272727 = 2.07262e-08 loss) | |
| I0623 17:28:53.264638 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.825243 | |
| I0623 17:28:53.264650 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 17:28:53.264662 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 17:28:53.264673 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 17:28:53.264684 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 17:28:53.264696 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.75 | |
| I0623 17:28:53.264708 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 17:28:53.264719 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 17:28:53.264730 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 17:28:53.264741 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 17:28:53.264753 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 17:28:53.264765 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.5 | |
| I0623 17:28:53.264776 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 17:28:53.264787 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 17:28:53.264808 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 17:28:53.264822 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 17:28:53.264833 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 17:28:53.264845 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 17:28:53.264856 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 17:28:53.264868 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 17:28:53.264879 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 17:28:53.264890 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:28:53.264901 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:28:53.264914 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.892045 | |
| I0623 17:28:53.264925 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.980583 | |
| I0623 17:28:53.264940 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.619287 (* 1 = 0.619287 loss) | |
| I0623 17:28:53.264952 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.380476 (* 1 = 0.380476 loss) | |
| I0623 17:28:53.264966 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0193719 (* 0.0909091 = 0.00176108 loss) | |
| I0623 17:28:53.264981 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0167952 (* 0.0909091 = 0.00152684 loss) | |
| I0623 17:28:53.264996 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.980324 (* 0.0909091 = 0.0891204 loss) | |
| I0623 17:28:53.265009 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.590955 (* 0.0909091 = 0.0537232 loss) | |
| I0623 17:28:53.265023 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.811132 (* 0.0909091 = 0.0737393 loss) | |
| I0623 17:28:53.265036 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.559689 (* 0.0909091 = 0.0508809 loss) | |
| I0623 17:28:53.265050 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.10312 (* 0.0909091 = 0.00937454 loss) | |
| I0623 17:28:53.265067 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.735585 (* 0.0909091 = 0.0668714 loss) | |
| I0623 17:28:53.265081 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.53615 (* 0.0909091 = 0.0487409 loss) | |
| I0623 17:28:53.265095 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.658233 (* 0.0909091 = 0.0598394 loss) | |
| I0623 17:28:53.265110 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.30224 (* 0.0909091 = 0.118385 loss) | |
| I0623 17:28:53.265122 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.08289 (* 0.0909091 = 0.0984443 loss) | |
| I0623 17:28:53.265136 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.662824 (* 0.0909091 = 0.0602567 loss) | |
| I0623 17:28:53.265149 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.772975 (* 0.0909091 = 0.0702704 loss) | |
| I0623 17:28:53.265163 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.53553 (* 0.0909091 = 0.139594 loss) | |
| I0623 17:28:53.265177 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.0418765 (* 0.0909091 = 0.00380695 loss) | |
| I0623 17:28:53.265192 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0103875 (* 0.0909091 = 0.000944315 loss) | |
| I0623 17:28:53.265207 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00059918 (* 0.0909091 = 5.44709e-05 loss) | |
| I0623 17:28:53.265220 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 8.3243e-05 (* 0.0909091 = 7.56754e-06 loss) | |
| I0623 17:28:53.265234 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 4.40448e-05 (* 0.0909091 = 4.00408e-06 loss) | |
| I0623 17:28:53.265247 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 3.26885e-05 (* 0.0909091 = 2.97168e-06 loss) | |
| I0623 17:28:53.265261 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.72371e-06 (* 0.0909091 = 4.29428e-07 loss) | |
| I0623 17:28:53.265285 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 17:28:53.265297 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.25 | |
| I0623 17:28:53.265310 10365 solver.cpp:245] Train net output #149: total_confidence = 0.176528 | |
| I0623 17:28:53.265321 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.121512 | |
| I0623 17:28:53.265334 10365 sgd_solver.cpp:106] Iteration 15000, lr = 0.001 | |
| I0623 17:29:22.751054 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.3646 > 30) by scale factor 0.743226 | |
| I0623 17:29:45.724217 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.9702 > 30) by scale factor 0.968672 | |
| I0623 17:30:04.903573 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 41.4752 > 30) by scale factor 0.723324 | |
| I0623 17:30:36.316987 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.3237 > 30) by scale factor 0.874031 | |
| I0623 17:32:41.260399 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 51.1439 > 30) by scale factor 0.58658 | |
| I0623 17:34:49.174701 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.829 > 30) by scale factor 0.70046 | |
| I0623 17:35:16.407706 10365 solver.cpp:229] Iteration 15500, loss = 4.54745 | |
| I0623 17:35:16.407775 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.466667 | |
| I0623 17:35:16.407794 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 17:35:16.407809 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 17:35:16.407822 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 17:35:16.407835 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 17:35:16.407847 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 17:35:16.407860 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 17:35:16.407872 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.125 | |
| I0623 17:35:16.407884 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 17:35:16.407896 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0 | |
| I0623 17:35:16.407908 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0 | |
| I0623 17:35:16.407920 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 17:35:16.407932 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 17:35:16.407945 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.75 | |
| I0623 17:35:16.407956 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.875 | |
| I0623 17:35:16.407968 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.875 | |
| I0623 17:35:16.407980 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 17:35:16.407992 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 17:35:16.408004 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 17:35:16.408015 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 17:35:16.408026 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 17:35:16.408037 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:35:16.408049 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:35:16.408061 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.659091 | |
| I0623 17:35:16.408072 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.655556 | |
| I0623 17:35:16.408089 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 2.28596 (* 0.3 = 0.685788 loss) | |
| I0623 17:35:16.408104 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.46457 (* 0.3 = 0.439372 loss) | |
| I0623 17:35:16.408121 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 2.93585 (* 0.0272727 = 0.0800685 loss) | |
| I0623 17:35:16.408136 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 2.22823 (* 0.0272727 = 0.06077 loss) | |
| I0623 17:35:16.408150 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.18706 (* 0.0272727 = 0.059647 loss) | |
| I0623 17:35:16.408164 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.37376 (* 0.0272727 = 0.0647389 loss) | |
| I0623 17:35:16.408177 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 3.02434 (* 0.0272727 = 0.0824819 loss) | |
| I0623 17:35:16.408191 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.53562 (* 0.0272727 = 0.0691534 loss) | |
| I0623 17:35:16.408205 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.74952 (* 0.0272727 = 0.0749869 loss) | |
| I0623 17:35:16.408220 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.62507 (* 0.0272727 = 0.0715929 loss) | |
| I0623 17:35:16.408233 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.99846 (* 0.0272727 = 0.0817762 loss) | |
| I0623 17:35:16.408246 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.89856 (* 0.0272727 = 0.0790517 loss) | |
| I0623 17:35:16.408260 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.23837 (* 0.0272727 = 0.0610465 loss) | |
| I0623 17:35:16.408273 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.73604 (* 0.0272727 = 0.0473466 loss) | |
| I0623 17:35:16.408318 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.739951 (* 0.0272727 = 0.0201805 loss) | |
| I0623 17:35:16.408334 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.57301 (* 0.0272727 = 0.0156275 loss) | |
| I0623 17:35:16.408349 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.478734 (* 0.0272727 = 0.0130564 loss) | |
| I0623 17:35:16.408362 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.511794 (* 0.0272727 = 0.013958 loss) | |
| I0623 17:35:16.408376 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0155141 (* 0.0272727 = 0.000423111 loss) | |
| I0623 17:35:16.408395 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00557572 (* 0.0272727 = 0.000152065 loss) | |
| I0623 17:35:16.408408 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00144915 (* 0.0272727 = 3.95223e-05 loss) | |
| I0623 17:35:16.408423 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000515403 (* 0.0272727 = 1.40564e-05 loss) | |
| I0623 17:35:16.408437 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000110729 (* 0.0272727 = 3.01987e-06 loss) | |
| I0623 17:35:16.408452 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000125478 (* 0.0272727 = 3.42212e-06 loss) | |
| I0623 17:35:16.408464 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.533333 | |
| I0623 17:35:16.408476 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 17:35:16.408488 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 17:35:16.408500 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 17:35:16.408512 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.375 | |
| I0623 17:35:16.408524 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 17:35:16.408535 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.25 | |
| I0623 17:35:16.408546 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 17:35:16.408558 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 17:35:16.408570 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.25 | |
| I0623 17:35:16.408581 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 17:35:16.408592 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.125 | |
| I0623 17:35:16.408604 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.75 | |
| I0623 17:35:16.408615 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 17:35:16.408627 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 17:35:16.408638 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 17:35:16.408650 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 17:35:16.408661 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 17:35:16.408674 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 17:35:16.408684 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 17:35:16.408696 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 17:35:16.408709 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:35:16.408720 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:35:16.408731 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.693182 | |
| I0623 17:35:16.408742 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.788889 | |
| I0623 17:35:16.408756 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 2.10909 (* 0.3 = 0.632726 loss) | |
| I0623 17:35:16.408771 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 1.40059 (* 0.3 = 0.420176 loss) | |
| I0623 17:35:16.408784 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 2.00901 (* 0.0272727 = 0.0547911 loss) | |
| I0623 17:35:16.408809 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 2.27652 (* 0.0272727 = 0.062087 loss) | |
| I0623 17:35:16.408824 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 2.01874 (* 0.0272727 = 0.0550567 loss) | |
| I0623 17:35:16.408838 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 2.43882 (* 0.0272727 = 0.0665133 loss) | |
| I0623 17:35:16.408852 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.39784 (* 0.0272727 = 0.0653957 loss) | |
| I0623 17:35:16.408865 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.1731 (* 0.0272727 = 0.0592663 loss) | |
| I0623 17:35:16.408879 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 2.20226 (* 0.0272727 = 0.0600615 loss) | |
| I0623 17:35:16.408893 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.74857 (* 0.0272727 = 0.0749609 loss) | |
| I0623 17:35:16.408907 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 2.96974 (* 0.0272727 = 0.0809929 loss) | |
| I0623 17:35:16.408921 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.98137 (* 0.0272727 = 0.0813101 loss) | |
| I0623 17:35:16.408934 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.54121 (* 0.0272727 = 0.0693057 loss) | |
| I0623 17:35:16.408948 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.00447 (* 0.0272727 = 0.0273945 loss) | |
| I0623 17:35:16.408962 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.09263 (* 0.0272727 = 0.0297991 loss) | |
| I0623 17:35:16.408977 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.504173 (* 0.0272727 = 0.0137502 loss) | |
| I0623 17:35:16.408989 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.300776 (* 0.0272727 = 0.00820299 loss) | |
| I0623 17:35:16.409004 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.290896 (* 0.0272727 = 0.00793354 loss) | |
| I0623 17:35:16.409018 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0182443 (* 0.0272727 = 0.000497572 loss) | |
| I0623 17:35:16.409031 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00292967 (* 0.0272727 = 7.99002e-05 loss) | |
| I0623 17:35:16.409046 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000642305 (* 0.0272727 = 1.75174e-05 loss) | |
| I0623 17:35:16.409060 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000173773 (* 0.0272727 = 4.73927e-06 loss) | |
| I0623 17:35:16.409075 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 3.52829e-05 (* 0.0272727 = 9.6226e-07 loss) | |
| I0623 17:35:16.409088 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 9.01537e-06 (* 0.0272727 = 2.45874e-07 loss) | |
| I0623 17:35:16.409101 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.777778 | |
| I0623 17:35:16.409112 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.75 | |
| I0623 17:35:16.409124 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.75 | |
| I0623 17:35:16.409135 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 17:35:16.409147 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 17:35:16.409160 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 17:35:16.409173 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 17:35:16.409184 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 17:35:16.409196 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 17:35:16.409207 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.5 | |
| I0623 17:35:16.409219 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.375 | |
| I0623 17:35:16.409230 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.25 | |
| I0623 17:35:16.409242 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 17:35:16.409255 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 17:35:16.409262 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 17:35:16.409270 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 17:35:16.409291 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 17:35:16.409304 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 17:35:16.409315 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 17:35:16.409327 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 17:35:16.409338 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 17:35:16.409349 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:35:16.409361 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:35:16.409373 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.806818 | |
| I0623 17:35:16.409384 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.888889 | |
| I0623 17:35:16.409399 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 1.31245 (* 1 = 1.31245 loss) | |
| I0623 17:35:16.409411 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 1.05854 (* 1 = 1.05854 loss) | |
| I0623 17:35:16.409425 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 1.52684 (* 0.0909091 = 0.138804 loss) | |
| I0623 17:35:16.409443 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 1.87922 (* 0.0909091 = 0.170838 loss) | |
| I0623 17:35:16.409457 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 1.31591 (* 0.0909091 = 0.119628 loss) | |
| I0623 17:35:16.409471 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 1.84772 (* 0.0909091 = 0.167974 loss) | |
| I0623 17:35:16.409484 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 1.55414 (* 0.0909091 = 0.141286 loss) | |
| I0623 17:35:16.409498 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 1.29058 (* 0.0909091 = 0.117326 loss) | |
| I0623 17:35:16.409512 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 1.30882 (* 0.0909091 = 0.118984 loss) | |
| I0623 17:35:16.409525 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 2.22921 (* 0.0909091 = 0.202655 loss) | |
| I0623 17:35:16.409539 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 2.16163 (* 0.0909091 = 0.196512 loss) | |
| I0623 17:35:16.409553 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 2.3957 (* 0.0909091 = 0.217791 loss) | |
| I0623 17:35:16.409566 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 2.71559 (* 0.0909091 = 0.246872 loss) | |
| I0623 17:35:16.409580 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.866483 (* 0.0909091 = 0.0787712 loss) | |
| I0623 17:35:16.409593 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.949237 (* 0.0909091 = 0.0862943 loss) | |
| I0623 17:35:16.409607 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.418448 (* 0.0909091 = 0.0380407 loss) | |
| I0623 17:35:16.409621 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.190765 (* 0.0909091 = 0.0173423 loss) | |
| I0623 17:35:16.409634 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.0514438 (* 0.0909091 = 0.00467671 loss) | |
| I0623 17:35:16.409648 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.111881 (* 0.0909091 = 0.010171 loss) | |
| I0623 17:35:16.409662 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0150806 (* 0.0909091 = 0.00137097 loss) | |
| I0623 17:35:16.409677 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00185079 (* 0.0909091 = 0.000168254 loss) | |
| I0623 17:35:16.409690 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000328922 (* 0.0909091 = 2.9902e-05 loss) | |
| I0623 17:35:16.409703 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000100428 (* 0.0909091 = 9.12977e-06 loss) | |
| I0623 17:35:16.409718 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.11276e-06 (* 0.0909091 = 3.73887e-07 loss) | |
| I0623 17:35:16.409729 10365 solver.cpp:245] Train net output #147: total_accuracy = 0 | |
| I0623 17:35:16.409740 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 17:35:16.409762 10365 solver.cpp:245] Train net output #149: total_confidence = 0.052811 | |
| I0623 17:35:16.409775 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0359814 | |
| I0623 17:35:16.409788 10365 sgd_solver.cpp:106] Iteration 15500, lr = 0.001 | |
| I0623 17:36:30.299654 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 43.4475 > 30) by scale factor 0.690488 | |
| I0623 17:36:38.723073 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.4381 > 30) by scale factor 0.985608 | |
| I0623 17:40:40.784248 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.9591 > 30) by scale factor 0.811708 | |
| I0623 17:41:09.893095 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.8781 > 30) by scale factor 0.971563 | |
| I0623 17:41:39.432080 10365 solver.cpp:229] Iteration 16000, loss = 4.51 | |
| I0623 17:41:39.432183 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.455556 | |
| I0623 17:41:39.432201 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.5 | |
| I0623 17:41:39.432214 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 17:41:39.432229 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.625 | |
| I0623 17:41:39.432241 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 17:41:39.432255 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0 | |
| I0623 17:41:39.432267 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.125 | |
| I0623 17:41:39.432279 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 17:41:39.432292 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 17:41:39.432304 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.875 | |
| I0623 17:41:39.432322 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 17:41:39.432334 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 17:41:39.432348 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.75 | |
| I0623 17:41:39.432359 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 17:41:39.432370 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 17:41:39.432382 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 17:41:39.432394 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 17:41:39.432405 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 17:41:39.432416 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 17:41:39.432428 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 17:41:39.432440 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 17:41:39.432451 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:41:39.432462 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:41:39.432474 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.693182 | |
| I0623 17:41:39.432487 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.811111 | |
| I0623 17:41:39.432503 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.58504 (* 0.3 = 0.475513 loss) | |
| I0623 17:41:39.432518 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.898963 (* 0.3 = 0.269689 loss) | |
| I0623 17:41:39.432533 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 1.62194 (* 0.0272727 = 0.0442348 loss) | |
| I0623 17:41:39.432546 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.08082 (* 0.0272727 = 0.029477 loss) | |
| I0623 17:41:39.432559 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.18115 (* 0.0272727 = 0.0322132 loss) | |
| I0623 17:41:39.432574 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.89612 (* 0.0272727 = 0.0517125 loss) | |
| I0623 17:41:39.432587 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.64415 (* 0.0272727 = 0.0721132 loss) | |
| I0623 17:41:39.432601 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.90363 (* 0.0272727 = 0.0519171 loss) | |
| I0623 17:41:39.432615 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.17232 (* 0.0272727 = 0.0319725 loss) | |
| I0623 17:41:39.432628 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.34754 (* 0.0272727 = 0.0367511 loss) | |
| I0623 17:41:39.432641 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 0.895944 (* 0.0272727 = 0.0244348 loss) | |
| I0623 17:41:39.432656 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.28084 (* 0.0272727 = 0.0622047 loss) | |
| I0623 17:41:39.432669 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.36303 (* 0.0272727 = 0.0371735 loss) | |
| I0623 17:41:39.432682 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 0.875732 (* 0.0272727 = 0.0238836 loss) | |
| I0623 17:41:39.432713 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.19028 (* 0.0272727 = 0.0324621 loss) | |
| I0623 17:41:39.432729 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.625492 (* 0.0272727 = 0.0170589 loss) | |
| I0623 17:41:39.432742 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.37941 (* 0.0272727 = 0.0376204 loss) | |
| I0623 17:41:39.432755 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 1.14794 (* 0.0272727 = 0.0313074 loss) | |
| I0623 17:41:39.432770 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00526318 (* 0.0272727 = 0.000143541 loss) | |
| I0623 17:41:39.432783 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.000656226 (* 0.0272727 = 1.78971e-05 loss) | |
| I0623 17:41:39.432798 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.000106272 (* 0.0272727 = 2.89833e-06 loss) | |
| I0623 17:41:39.432812 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 2.33959e-05 (* 0.0272727 = 6.3807e-07 loss) | |
| I0623 17:41:39.432826 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 1.9805e-05 (* 0.0272727 = 5.40136e-07 loss) | |
| I0623 17:41:39.432840 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 1.66894e-06 (* 0.0272727 = 4.55165e-08 loss) | |
| I0623 17:41:39.432852 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.566667 | |
| I0623 17:41:39.432864 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.75 | |
| I0623 17:41:39.432878 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 17:41:39.432888 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 17:41:39.432900 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 17:41:39.432911 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 17:41:39.432924 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 17:41:39.432934 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.625 | |
| I0623 17:41:39.432946 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 17:41:39.432958 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.75 | |
| I0623 17:41:39.432970 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 17:41:39.432981 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 17:41:39.432992 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 17:41:39.433004 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 17:41:39.433015 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 17:41:39.433027 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 17:41:39.433038 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 17:41:39.433049 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 17:41:39.433060 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 17:41:39.433071 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 17:41:39.433084 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 17:41:39.433094 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:41:39.433105 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:41:39.433116 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.772727 | |
| I0623 17:41:39.433128 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.844444 | |
| I0623 17:41:39.433141 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.24788 (* 0.3 = 0.374364 loss) | |
| I0623 17:41:39.433156 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.675333 (* 0.3 = 0.2026 loss) | |
| I0623 17:41:39.433171 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 1.11968 (* 0.0272727 = 0.0305368 loss) | |
| I0623 17:41:39.433185 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.524048 (* 0.0272727 = 0.0142922 loss) | |
| I0623 17:41:39.433212 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.191705 (* 0.0272727 = 0.00522833 loss) | |
| I0623 17:41:39.433226 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.775424 (* 0.0272727 = 0.0211479 loss) | |
| I0623 17:41:39.433240 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.91093 (* 0.0272727 = 0.0521163 loss) | |
| I0623 17:41:39.433254 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.23119 (* 0.0272727 = 0.0335778 loss) | |
| I0623 17:41:39.433267 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.0429 (* 0.0272727 = 0.0284428 loss) | |
| I0623 17:41:39.433280 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.11421 (* 0.0272727 = 0.0303874 loss) | |
| I0623 17:41:39.433295 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 0.756442 (* 0.0272727 = 0.0206302 loss) | |
| I0623 17:41:39.433307 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.92475 (* 0.0272727 = 0.0524932 loss) | |
| I0623 17:41:39.433321 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.87471 (* 0.0272727 = 0.0511285 loss) | |
| I0623 17:41:39.433334 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.10935 (* 0.0272727 = 0.030255 loss) | |
| I0623 17:41:39.433347 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.10159 (* 0.0272727 = 0.0300434 loss) | |
| I0623 17:41:39.433362 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.542901 (* 0.0272727 = 0.0148064 loss) | |
| I0623 17:41:39.433380 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.73165 (* 0.0272727 = 0.0472268 loss) | |
| I0623 17:41:39.433394 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.867831 (* 0.0272727 = 0.0236681 loss) | |
| I0623 17:41:39.433408 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0123861 (* 0.0272727 = 0.000337803 loss) | |
| I0623 17:41:39.433421 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00129497 (* 0.0272727 = 3.53173e-05 loss) | |
| I0623 17:41:39.433435 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000553851 (* 0.0272727 = 1.5105e-05 loss) | |
| I0623 17:41:39.433449 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 9.33429e-05 (* 0.0272727 = 2.54572e-06 loss) | |
| I0623 17:41:39.433464 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 1.34117e-05 (* 0.0272727 = 3.65772e-07 loss) | |
| I0623 17:41:39.433477 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 3.4571e-06 (* 0.0272727 = 9.42847e-08 loss) | |
| I0623 17:41:39.433490 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.866667 | |
| I0623 17:41:39.433501 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.875 | |
| I0623 17:41:39.433513 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 17:41:39.433524 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 17:41:39.433537 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 17:41:39.433547 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 17:41:39.433559 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 17:41:39.433570 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 17:41:39.433583 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 17:41:39.433593 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 17:41:39.433605 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 17:41:39.433616 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.875 | |
| I0623 17:41:39.433629 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.875 | |
| I0623 17:41:39.433640 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 17:41:39.433650 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 17:41:39.433661 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 17:41:39.433673 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 17:41:39.433696 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 17:41:39.433709 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 17:41:39.433720 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 17:41:39.433732 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 17:41:39.433743 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:41:39.433754 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:41:39.433765 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.920455 | |
| I0623 17:41:39.433778 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.988889 | |
| I0623 17:41:39.433790 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.492194 (* 1 = 0.492194 loss) | |
| I0623 17:41:39.433804 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.267416 (* 1 = 0.267416 loss) | |
| I0623 17:41:39.433818 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 1.20024 (* 0.0909091 = 0.109112 loss) | |
| I0623 17:41:39.433832 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0261308 (* 0.0909091 = 0.00237552 loss) | |
| I0623 17:41:39.433842 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0209459 (* 0.0909091 = 0.00190417 loss) | |
| I0623 17:41:39.433852 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0243902 (* 0.0909091 = 0.0022173 loss) | |
| I0623 17:41:39.433866 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.113577 (* 0.0909091 = 0.0103252 loss) | |
| I0623 17:41:39.433881 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.180774 (* 0.0909091 = 0.016434 loss) | |
| I0623 17:41:39.433894 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.140943 (* 0.0909091 = 0.012813 loss) | |
| I0623 17:41:39.433907 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.888942 (* 0.0909091 = 0.0808129 loss) | |
| I0623 17:41:39.433922 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.468491 (* 0.0909091 = 0.0425901 loss) | |
| I0623 17:41:39.433934 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.91449 (* 0.0909091 = 0.0831355 loss) | |
| I0623 17:41:39.433948 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.789303 (* 0.0909091 = 0.0717548 loss) | |
| I0623 17:41:39.433961 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.520681 (* 0.0909091 = 0.0473346 loss) | |
| I0623 17:41:39.433975 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.07106 (* 0.0909091 = 0.0973688 loss) | |
| I0623 17:41:39.433989 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.504619 (* 0.0909091 = 0.0458744 loss) | |
| I0623 17:41:39.434002 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.778453 (* 0.0909091 = 0.0707685 loss) | |
| I0623 17:41:39.434015 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.604533 (* 0.0909091 = 0.0549575 loss) | |
| I0623 17:41:39.434029 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.007127 (* 0.0909091 = 0.000647909 loss) | |
| I0623 17:41:39.434043 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0012936 (* 0.0909091 = 0.0001176 loss) | |
| I0623 17:41:39.434057 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000213831 (* 0.0909091 = 1.94392e-05 loss) | |
| I0623 17:41:39.434072 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000118517 (* 0.0909091 = 1.07743e-05 loss) | |
| I0623 17:41:39.434085 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 3.82927e-05 (* 0.0909091 = 3.48115e-06 loss) | |
| I0623 17:41:39.434099 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.09785e-06 (* 0.0909091 = 3.72532e-07 loss) | |
| I0623 17:41:39.434110 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 17:41:39.434123 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 17:41:39.434134 10365 solver.cpp:245] Train net output #149: total_confidence = 0.285479 | |
| I0623 17:41:39.434155 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.212331 | |
| I0623 17:41:39.434170 10365 sgd_solver.cpp:106] Iteration 16000, lr = 0.001 | |
| I0623 17:46:37.756687 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.2567 > 30) by scale factor 0.709946 | |
| I0623 17:46:43.880172 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.0532 > 30) by scale factor 0.880975 | |
| I0623 17:48:02.473757 10365 solver.cpp:229] Iteration 16500, loss = 4.59404 | |
| I0623 17:48:02.473951 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.342105 | |
| I0623 17:48:02.473975 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 17:48:02.473989 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.375 | |
| I0623 17:48:02.474002 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 17:48:02.474015 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.125 | |
| I0623 17:48:02.474028 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 17:48:02.474041 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.125 | |
| I0623 17:48:02.474056 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 17:48:02.474067 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.25 | |
| I0623 17:48:02.474081 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.125 | |
| I0623 17:48:02.474094 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 17:48:02.474107 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0 | |
| I0623 17:48:02.474119 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.125 | |
| I0623 17:48:02.474131 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.25 | |
| I0623 17:48:02.474144 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.25 | |
| I0623 17:48:02.474156 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 17:48:02.474169 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 1 | |
| I0623 17:48:02.474180 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 17:48:02.474192 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 17:48:02.474205 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 17:48:02.474216 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 17:48:02.474228 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:48:02.474241 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:48:02.474251 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.556818 | |
| I0623 17:48:02.474267 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.72807 | |
| I0623 17:48:02.474284 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.90602 (* 0.3 = 0.571805 loss) | |
| I0623 17:48:02.474300 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.29113 (* 0.3 = 0.387339 loss) | |
| I0623 17:48:02.474314 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.631756 (* 0.0272727 = 0.0172297 loss) | |
| I0623 17:48:02.474328 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.86203 (* 0.0272727 = 0.0507826 loss) | |
| I0623 17:48:02.474342 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.3251 (* 0.0272727 = 0.0634117 loss) | |
| I0623 17:48:02.474357 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.43253 (* 0.0272727 = 0.0663417 loss) | |
| I0623 17:48:02.474371 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.62407 (* 0.0272727 = 0.0715655 loss) | |
| I0623 17:48:02.474385 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.7407 (* 0.0272727 = 0.0747462 loss) | |
| I0623 17:48:02.474400 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.26611 (* 0.0272727 = 0.061803 loss) | |
| I0623 17:48:02.474413 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.92698 (* 0.0272727 = 0.052554 loss) | |
| I0623 17:48:02.474427 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.30695 (* 0.0272727 = 0.0629167 loss) | |
| I0623 17:48:02.474442 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.37578 (* 0.0272727 = 0.0647941 loss) | |
| I0623 17:48:02.474455 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.69803 (* 0.0272727 = 0.0735827 loss) | |
| I0623 17:48:02.474469 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.16032 (* 0.0272727 = 0.0589179 loss) | |
| I0623 17:48:02.474509 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 2.11578 (* 0.0272727 = 0.057703 loss) | |
| I0623 17:48:02.474524 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 2.78621 (* 0.0272727 = 0.0759874 loss) | |
| I0623 17:48:02.474537 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.35998 (* 0.0272727 = 0.0370903 loss) | |
| I0623 17:48:02.474552 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.24712 (* 0.0272727 = 0.00673964 loss) | |
| I0623 17:48:02.474566 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0854436 (* 0.0272727 = 0.00233028 loss) | |
| I0623 17:48:02.474581 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0164669 (* 0.0272727 = 0.000449097 loss) | |
| I0623 17:48:02.474596 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00318793 (* 0.0272727 = 8.69435e-05 loss) | |
| I0623 17:48:02.474609 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000429785 (* 0.0272727 = 1.17214e-05 loss) | |
| I0623 17:48:02.474624 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000129577 (* 0.0272727 = 3.53392e-06 loss) | |
| I0623 17:48:02.474638 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.65095e-05 (* 0.0272727 = 9.95714e-07 loss) | |
| I0623 17:48:02.474650 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.464912 | |
| I0623 17:48:02.474663 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 17:48:02.474675 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.5 | |
| I0623 17:48:02.474687 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.625 | |
| I0623 17:48:02.474699 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.375 | |
| I0623 17:48:02.474711 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 17:48:02.474723 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 17:48:02.474736 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 17:48:02.474748 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.25 | |
| I0623 17:48:02.474761 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 17:48:02.474772 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.125 | |
| I0623 17:48:02.474784 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.125 | |
| I0623 17:48:02.474797 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.375 | |
| I0623 17:48:02.474808 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 17:48:02.474819 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.25 | |
| I0623 17:48:02.474831 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 17:48:02.474844 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 17:48:02.474855 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 17:48:02.474867 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 17:48:02.474879 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 17:48:02.474891 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 17:48:02.474905 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:48:02.474915 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:48:02.474927 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.647727 | |
| I0623 17:48:02.474939 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.754386 | |
| I0623 17:48:02.474958 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.56781 (* 0.3 = 0.470344 loss) | |
| I0623 17:48:02.474972 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 1.05364 (* 0.3 = 0.316091 loss) | |
| I0623 17:48:02.474987 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.326965 (* 0.0272727 = 0.00891723 loss) | |
| I0623 17:48:02.475003 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 1.21812 (* 0.0272727 = 0.0332214 loss) | |
| I0623 17:48:02.475028 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.812755 (* 0.0272727 = 0.022166 loss) | |
| I0623 17:48:02.475044 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.57434 (* 0.0272727 = 0.0429367 loss) | |
| I0623 17:48:02.475057 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.73693 (* 0.0272727 = 0.0473707 loss) | |
| I0623 17:48:02.475071 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.91344 (* 0.0272727 = 0.0521848 loss) | |
| I0623 17:48:02.475085 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.95928 (* 0.0272727 = 0.0534348 loss) | |
| I0623 17:48:02.475098 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.9801 (* 0.0272727 = 0.0540027 loss) | |
| I0623 17:48:02.475112 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.98284 (* 0.0272727 = 0.0540775 loss) | |
| I0623 17:48:02.475126 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.07466 (* 0.0272727 = 0.0565817 loss) | |
| I0623 17:48:02.475141 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.67868 (* 0.0272727 = 0.073055 loss) | |
| I0623 17:48:02.475154 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 2.38791 (* 0.0272727 = 0.0651247 loss) | |
| I0623 17:48:02.475167 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 2.83277 (* 0.0272727 = 0.0772574 loss) | |
| I0623 17:48:02.475183 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 2.49633 (* 0.0272727 = 0.0680819 loss) | |
| I0623 17:48:02.475196 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.984696 (* 0.0272727 = 0.0268554 loss) | |
| I0623 17:48:02.475211 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.16739 (* 0.0272727 = 0.00456519 loss) | |
| I0623 17:48:02.475225 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0501484 (* 0.0272727 = 0.00136768 loss) | |
| I0623 17:48:02.475239 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0136613 (* 0.0272727 = 0.000372582 loss) | |
| I0623 17:48:02.475255 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0052132 (* 0.0272727 = 0.000142178 loss) | |
| I0623 17:48:02.475268 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000709621 (* 0.0272727 = 1.93533e-05 loss) | |
| I0623 17:48:02.475282 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000236055 (* 0.0272727 = 6.43785e-06 loss) | |
| I0623 17:48:02.475296 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 2.00429e-05 (* 0.0272727 = 5.46625e-07 loss) | |
| I0623 17:48:02.475308 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.710526 | |
| I0623 17:48:02.475323 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 17:48:02.475335 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 17:48:02.475348 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 17:48:02.475359 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 17:48:02.475371 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 17:48:02.475383 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 17:48:02.475394 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 17:48:02.475406 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 17:48:02.475419 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 17:48:02.475430 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 17:48:02.475441 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.25 | |
| I0623 17:48:02.475453 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 17:48:02.475466 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.375 | |
| I0623 17:48:02.475476 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.125 | |
| I0623 17:48:02.475488 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 17:48:02.475500 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 17:48:02.475522 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 17:48:02.475535 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 17:48:02.475548 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 17:48:02.475559 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 17:48:02.475571 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:48:02.475582 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:48:02.475594 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.806818 | |
| I0623 17:48:02.475622 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.964912 | |
| I0623 17:48:02.475637 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.826579 (* 1 = 0.826579 loss) | |
| I0623 17:48:02.475651 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.54792 (* 1 = 0.54792 loss) | |
| I0623 17:48:02.475666 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.26911 (* 0.0909091 = 0.0244645 loss) | |
| I0623 17:48:02.475679 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.716964 (* 0.0909091 = 0.0651786 loss) | |
| I0623 17:48:02.475693 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.289093 (* 0.0909091 = 0.0262811 loss) | |
| I0623 17:48:02.475708 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.361016 (* 0.0909091 = 0.0328196 loss) | |
| I0623 17:48:02.475723 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.330225 (* 0.0909091 = 0.0300204 loss) | |
| I0623 17:48:02.475736 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.368041 (* 0.0909091 = 0.0334582 loss) | |
| I0623 17:48:02.475750 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 1.12028 (* 0.0909091 = 0.101843 loss) | |
| I0623 17:48:02.475761 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.07226 (* 0.0909091 = 0.0974782 loss) | |
| I0623 17:48:02.475770 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.70184 (* 0.0909091 = 0.154712 loss) | |
| I0623 17:48:02.475785 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.59125 (* 0.0909091 = 0.144659 loss) | |
| I0623 17:48:02.475800 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 2.02936 (* 0.0909091 = 0.184487 loss) | |
| I0623 17:48:02.475813 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.70936 (* 0.0909091 = 0.155396 loss) | |
| I0623 17:48:02.475827 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.67451 (* 0.0909091 = 0.152228 loss) | |
| I0623 17:48:02.475841 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 2.08177 (* 0.0909091 = 0.189252 loss) | |
| I0623 17:48:02.475854 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.79383 (* 0.0909091 = 0.0721664 loss) | |
| I0623 17:48:02.475868 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.080064 (* 0.0909091 = 0.00727855 loss) | |
| I0623 17:48:02.475883 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0162906 (* 0.0909091 = 0.00148097 loss) | |
| I0623 17:48:02.475898 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000939162 (* 0.0909091 = 8.53784e-05 loss) | |
| I0623 17:48:02.475911 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000151144 (* 0.0909091 = 1.37404e-05 loss) | |
| I0623 17:48:02.475925 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 7.54958e-05 (* 0.0909091 = 6.86326e-06 loss) | |
| I0623 17:48:02.475939 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 3.83572e-05 (* 0.0909091 = 3.48701e-06 loss) | |
| I0623 17:48:02.475955 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 5.78169e-06 (* 0.0909091 = 5.25608e-07 loss) | |
| I0623 17:48:02.475966 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 17:48:02.475978 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 17:48:02.475991 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0338437 | |
| I0623 17:48:02.476019 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0150467 | |
| I0623 17:48:02.476034 10365 sgd_solver.cpp:106] Iteration 16500, lr = 0.001 | |
| I0623 17:48:49.566346 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.2378 > 30) by scale factor 0.902588 | |
| I0623 17:49:20.967878 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 37.4694 > 30) by scale factor 0.800653 | |
| I0623 17:50:59.829407 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 50.5313 > 30) by scale factor 0.593691 | |
| I0623 17:53:19.283123 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.4745 > 30) by scale factor 0.98443 | |
| I0623 17:53:29.233710 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.8065 > 30) by scale factor 0.837838 | |
| I0623 17:54:25.578884 10365 solver.cpp:229] Iteration 17000, loss = 4.45992 | |
| I0623 17:54:25.578979 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.421569 | |
| I0623 17:54:25.578999 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 17:54:25.579011 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 17:54:25.579025 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 17:54:25.579037 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 17:54:25.579051 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 17:54:25.579062 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.625 | |
| I0623 17:54:25.579076 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.25 | |
| I0623 17:54:25.579087 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.25 | |
| I0623 17:54:25.579102 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 17:54:25.579115 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 17:54:25.579128 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 17:54:25.579140 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 17:54:25.579154 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 17:54:25.579165 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 17:54:25.579177 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 17:54:25.579190 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 17:54:25.579201 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 17:54:25.579212 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 17:54:25.579224 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 17:54:25.579236 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 17:54:25.579247 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 17:54:25.579259 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 17:54:25.579272 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.653409 | |
| I0623 17:54:25.579283 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.803922 | |
| I0623 17:54:25.579299 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.6809 (* 0.3 = 0.504269 loss) | |
| I0623 17:54:25.579313 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.01045 (* 0.3 = 0.303135 loss) | |
| I0623 17:54:25.579329 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.259098 (* 0.0272727 = 0.0070663 loss) | |
| I0623 17:54:25.579342 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.727537 (* 0.0272727 = 0.0198419 loss) | |
| I0623 17:54:25.579356 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.39442 (* 0.0272727 = 0.0653024 loss) | |
| I0623 17:54:25.579370 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.92569 (* 0.0272727 = 0.0525187 loss) | |
| I0623 17:54:25.579383 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.48099 (* 0.0272727 = 0.0676633 loss) | |
| I0623 17:54:25.579397 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.43087 (* 0.0272727 = 0.0390237 loss) | |
| I0623 17:54:25.579411 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.15126 (* 0.0272727 = 0.0586708 loss) | |
| I0623 17:54:25.579424 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.874 (* 0.0272727 = 0.051109 loss) | |
| I0623 17:54:25.579438 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.81765 (* 0.0272727 = 0.0495724 loss) | |
| I0623 17:54:25.579452 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.86741 (* 0.0272727 = 0.0509293 loss) | |
| I0623 17:54:25.579465 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.10394 (* 0.0272727 = 0.0573802 loss) | |
| I0623 17:54:25.579479 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.19646 (* 0.0272727 = 0.0326307 loss) | |
| I0623 17:54:25.579511 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.15134 (* 0.0272727 = 0.0314001 loss) | |
| I0623 17:54:25.579526 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.19377 (* 0.0272727 = 0.0325573 loss) | |
| I0623 17:54:25.579540 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.449862 (* 0.0272727 = 0.012269 loss) | |
| I0623 17:54:25.579553 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.41403 (* 0.0272727 = 0.0112917 loss) | |
| I0623 17:54:25.579567 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.287194 (* 0.0272727 = 0.00783256 loss) | |
| I0623 17:54:25.579582 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.596937 (* 0.0272727 = 0.0162801 loss) | |
| I0623 17:54:25.579610 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0142284 (* 0.0272727 = 0.000388048 loss) | |
| I0623 17:54:25.579629 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00121278 (* 0.0272727 = 3.30759e-05 loss) | |
| I0623 17:54:25.579643 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 1.84185e-05 (* 0.0272727 = 5.02322e-07 loss) | |
| I0623 17:54:25.579658 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 9.77545e-06 (* 0.0272727 = 2.66603e-07 loss) | |
| I0623 17:54:25.579670 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.490196 | |
| I0623 17:54:25.579682 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 17:54:25.579694 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 17:54:25.579705 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 17:54:25.579717 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.875 | |
| I0623 17:54:25.579728 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 17:54:25.579741 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 17:54:25.579752 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 17:54:25.579764 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 17:54:25.579776 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 17:54:25.579787 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.125 | |
| I0623 17:54:25.579797 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 17:54:25.579808 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.375 | |
| I0623 17:54:25.579820 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 17:54:25.579831 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 17:54:25.579843 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 17:54:25.579854 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 17:54:25.579865 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 17:54:25.579876 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 17:54:25.579888 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 17:54:25.579900 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 17:54:25.579911 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 17:54:25.579922 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 17:54:25.579933 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.704545 | |
| I0623 17:54:25.579944 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.892157 | |
| I0623 17:54:25.579958 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.38124 (* 0.3 = 0.414372 loss) | |
| I0623 17:54:25.579972 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.819376 (* 0.3 = 0.245813 loss) | |
| I0623 17:54:25.579987 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.15543 (* 0.0272727 = 0.004239 loss) | |
| I0623 17:54:25.579999 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.570801 (* 0.0272727 = 0.0155673 loss) | |
| I0623 17:54:25.580026 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.21869 (* 0.0272727 = 0.0332369 loss) | |
| I0623 17:54:25.580041 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.30106 (* 0.0272727 = 0.0354836 loss) | |
| I0623 17:54:25.580054 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 2.38203 (* 0.0272727 = 0.0649643 loss) | |
| I0623 17:54:25.580068 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.37355 (* 0.0272727 = 0.0374605 loss) | |
| I0623 17:54:25.580082 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.92756 (* 0.0272727 = 0.0525698 loss) | |
| I0623 17:54:25.580096 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.50637 (* 0.0272727 = 0.0410828 loss) | |
| I0623 17:54:25.580109 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.63654 (* 0.0272727 = 0.044633 loss) | |
| I0623 17:54:25.580122 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.91653 (* 0.0272727 = 0.0522689 loss) | |
| I0623 17:54:25.580137 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.13678 (* 0.0272727 = 0.0582757 loss) | |
| I0623 17:54:25.580152 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.46417 (* 0.0272727 = 0.0399318 loss) | |
| I0623 17:54:25.580166 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.0184 (* 0.0272727 = 0.0277745 loss) | |
| I0623 17:54:25.580180 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.700636 (* 0.0272727 = 0.0191083 loss) | |
| I0623 17:54:25.580193 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.27372 (* 0.0272727 = 0.00746509 loss) | |
| I0623 17:54:25.580207 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.813301 (* 0.0272727 = 0.0221809 loss) | |
| I0623 17:54:25.580221 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.25356 (* 0.0272727 = 0.00691526 loss) | |
| I0623 17:54:25.580235 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.391748 (* 0.0272727 = 0.010684 loss) | |
| I0623 17:54:25.580250 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00194399 (* 0.0272727 = 5.30179e-05 loss) | |
| I0623 17:54:25.580263 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 7.53776e-05 (* 0.0272727 = 2.05575e-06 loss) | |
| I0623 17:54:25.580277 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 5.51355e-06 (* 0.0272727 = 1.5037e-07 loss) | |
| I0623 17:54:25.580291 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.49012e-07 (* 0.0272727 = 4.06395e-09 loss) | |
| I0623 17:54:25.580303 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.833333 | |
| I0623 17:54:25.580315 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 17:54:25.580327 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 17:54:25.580338 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 17:54:25.580350 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 17:54:25.580361 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 17:54:25.580373 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 17:54:25.580384 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 17:54:25.580395 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 17:54:25.580407 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 17:54:25.580418 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.375 | |
| I0623 17:54:25.580430 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.375 | |
| I0623 17:54:25.580441 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 17:54:25.580452 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 17:54:25.580464 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 17:54:25.580476 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 1 | |
| I0623 17:54:25.580488 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 17:54:25.580509 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 17:54:25.580523 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.875 | |
| I0623 17:54:25.580533 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 17:54:25.580545 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 17:54:25.580556 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 17:54:25.580567 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 17:54:25.580579 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.897727 | |
| I0623 17:54:25.580591 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.960784 | |
| I0623 17:54:25.580605 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.725996 (* 1 = 0.725996 loss) | |
| I0623 17:54:25.580618 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.444748 (* 1 = 0.444748 loss) | |
| I0623 17:54:25.580631 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0148518 (* 0.0909091 = 0.00135016 loss) | |
| I0623 17:54:25.580646 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0378752 (* 0.0909091 = 0.0034432 loss) | |
| I0623 17:54:25.580665 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.987357 (* 0.0909091 = 0.0897597 loss) | |
| I0623 17:54:25.580680 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 1.29385 (* 0.0909091 = 0.117623 loss) | |
| I0623 17:54:25.580693 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 1.57983 (* 0.0909091 = 0.14362 loss) | |
| I0623 17:54:25.580708 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.222734 (* 0.0909091 = 0.0202486 loss) | |
| I0623 17:54:25.580721 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 1.26086 (* 0.0909091 = 0.114624 loss) | |
| I0623 17:54:25.580734 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.63508 (* 0.0909091 = 0.0577345 loss) | |
| I0623 17:54:25.580749 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.589944 (* 0.0909091 = 0.0536313 loss) | |
| I0623 17:54:25.580761 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.16374 (* 0.0909091 = 0.105795 loss) | |
| I0623 17:54:25.580775 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.3742 (* 0.0909091 = 0.124927 loss) | |
| I0623 17:54:25.580790 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.10868 (* 0.0909091 = 0.100789 loss) | |
| I0623 17:54:25.580802 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.914521 (* 0.0909091 = 0.0831383 loss) | |
| I0623 17:54:25.580816 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.469686 (* 0.0909091 = 0.0426987 loss) | |
| I0623 17:54:25.580829 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.1412 (* 0.0909091 = 0.0128364 loss) | |
| I0623 17:54:25.580843 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.108474 (* 0.0909091 = 0.00986127 loss) | |
| I0623 17:54:25.580857 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.182879 (* 0.0909091 = 0.0166254 loss) | |
| I0623 17:54:25.580870 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.154912 (* 0.0909091 = 0.0140829 loss) | |
| I0623 17:54:25.580885 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0641493 (* 0.0909091 = 0.00583175 loss) | |
| I0623 17:54:25.580899 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.00333933 (* 0.0909091 = 0.000303575 loss) | |
| I0623 17:54:25.580914 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000857602 (* 0.0909091 = 7.79639e-05 loss) | |
| I0623 17:54:25.580926 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 6.60131e-05 (* 0.0909091 = 6.00119e-06 loss) | |
| I0623 17:54:25.580938 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 17:54:25.580950 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 17:54:25.580961 10365 solver.cpp:245] Train net output #149: total_confidence = 0.12991 | |
| I0623 17:54:25.580983 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.134456 | |
| I0623 17:54:25.580998 10365 sgd_solver.cpp:106] Iteration 17000, lr = 0.001 | |
| I0623 17:57:20.605973 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.2129 > 30) by scale factor 0.785074 | |
| I0623 17:59:16.346112 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.058 > 30) by scale factor 0.788271 | |
| I0623 18:00:48.733796 10365 solver.cpp:229] Iteration 17500, loss = 4.55397 | |
| I0623 18:00:48.733893 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.465909 | |
| I0623 18:00:48.733913 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.5 | |
| I0623 18:00:48.733927 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.875 | |
| I0623 18:00:48.733939 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 18:00:48.733952 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.75 | |
| I0623 18:00:48.733964 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 18:00:48.733978 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 18:00:48.733990 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 18:00:48.734002 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.75 | |
| I0623 18:00:48.734015 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 18:00:48.734027 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.75 | |
| I0623 18:00:48.734041 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 18:00:48.734053 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 18:00:48.734066 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 18:00:48.734081 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 18:00:48.734093 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 18:00:48.734105 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 18:00:48.734117 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 18:00:48.734128 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:00:48.734140 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:00:48.734153 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:00:48.734166 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:00:48.734179 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:00:48.734190 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.698864 | |
| I0623 18:00:48.734202 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.784091 | |
| I0623 18:00:48.734218 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.58247 (* 0.3 = 0.474741 loss) | |
| I0623 18:00:48.734233 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.908175 (* 0.3 = 0.272453 loss) | |
| I0623 18:00:48.734247 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 1.24395 (* 0.0272727 = 0.0339258 loss) | |
| I0623 18:00:48.734261 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.671046 (* 0.0272727 = 0.0183012 loss) | |
| I0623 18:00:48.734275 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.90179 (* 0.0272727 = 0.0518671 loss) | |
| I0623 18:00:48.734289 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.02668 (* 0.0272727 = 0.0280005 loss) | |
| I0623 18:00:48.734303 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.67903 (* 0.0272727 = 0.0457918 loss) | |
| I0623 18:00:48.734318 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.69868 (* 0.0272727 = 0.0463276 loss) | |
| I0623 18:00:48.734330 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.27664 (* 0.0272727 = 0.0348174 loss) | |
| I0623 18:00:48.734344 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 0.736902 (* 0.0272727 = 0.0200973 loss) | |
| I0623 18:00:48.734359 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.00499 (* 0.0272727 = 0.0546815 loss) | |
| I0623 18:00:48.734372 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.21174 (* 0.0272727 = 0.0330473 loss) | |
| I0623 18:00:48.734385 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.11601 (* 0.0272727 = 0.0577094 loss) | |
| I0623 18:00:48.734400 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.60824 (* 0.0272727 = 0.043861 loss) | |
| I0623 18:00:48.734431 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.5044 (* 0.0272727 = 0.0410291 loss) | |
| I0623 18:00:48.734447 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.95008 (* 0.0272727 = 0.053184 loss) | |
| I0623 18:00:48.734459 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.629642 (* 0.0272727 = 0.017172 loss) | |
| I0623 18:00:48.734474 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.325212 (* 0.0272727 = 0.00886941 loss) | |
| I0623 18:00:48.734488 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.463415 (* 0.0272727 = 0.0126386 loss) | |
| I0623 18:00:48.734503 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0163424 (* 0.0272727 = 0.000445703 loss) | |
| I0623 18:00:48.734516 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00247051 (* 0.0272727 = 6.73775e-05 loss) | |
| I0623 18:00:48.734530 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000204931 (* 0.0272727 = 5.58902e-06 loss) | |
| I0623 18:00:48.734545 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 8.4477e-05 (* 0.0272727 = 2.30392e-06 loss) | |
| I0623 18:00:48.734560 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.22575e-05 (* 0.0272727 = 8.79749e-07 loss) | |
| I0623 18:00:48.734571 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.659091 | |
| I0623 18:00:48.734583 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 18:00:48.734596 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 18:00:48.734606 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 18:00:48.734618 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 18:00:48.734629 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.375 | |
| I0623 18:00:48.734642 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 18:00:48.734652 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.625 | |
| I0623 18:00:48.734663 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.875 | |
| I0623 18:00:48.734675 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 18:00:48.734686 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.625 | |
| I0623 18:00:48.734697 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 18:00:48.734709 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 18:00:48.734720 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 18:00:48.734731 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 18:00:48.734742 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 18:00:48.734755 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 18:00:48.734766 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 18:00:48.734777 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:00:48.734788 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:00:48.734799 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:00:48.734812 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:00:48.734822 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:00:48.734833 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.795455 | |
| I0623 18:00:48.734845 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.852273 | |
| I0623 18:00:48.734858 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.14436 (* 0.3 = 0.343308 loss) | |
| I0623 18:00:48.734872 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.738139 (* 0.3 = 0.221442 loss) | |
| I0623 18:00:48.734886 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.360458 (* 0.0272727 = 0.00983068 loss) | |
| I0623 18:00:48.734901 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.257223 (* 0.0272727 = 0.00701518 loss) | |
| I0623 18:00:48.734926 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.490091 (* 0.0272727 = 0.0133661 loss) | |
| I0623 18:00:48.734941 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.955415 (* 0.0272727 = 0.0260568 loss) | |
| I0623 18:00:48.734956 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.26903 (* 0.0272727 = 0.0346099 loss) | |
| I0623 18:00:48.734968 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.15263 (* 0.0272727 = 0.0314353 loss) | |
| I0623 18:00:48.734982 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.05207 (* 0.0272727 = 0.0286927 loss) | |
| I0623 18:00:48.734995 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 0.524293 (* 0.0272727 = 0.0142989 loss) | |
| I0623 18:00:48.735009 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.79985 (* 0.0272727 = 0.0490867 loss) | |
| I0623 18:00:48.735023 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.13331 (* 0.0272727 = 0.0581812 loss) | |
| I0623 18:00:48.735036 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.45902 (* 0.0272727 = 0.0397915 loss) | |
| I0623 18:00:48.735049 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.20312 (* 0.0272727 = 0.0328124 loss) | |
| I0623 18:00:48.735064 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.60556 (* 0.0272727 = 0.043788 loss) | |
| I0623 18:00:48.735076 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.22588 (* 0.0272727 = 0.0334331 loss) | |
| I0623 18:00:48.735090 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.793331 (* 0.0272727 = 0.0216363 loss) | |
| I0623 18:00:48.735103 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.421267 (* 0.0272727 = 0.0114891 loss) | |
| I0623 18:00:48.735117 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.31259 (* 0.0272727 = 0.00852517 loss) | |
| I0623 18:00:48.735136 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00955757 (* 0.0272727 = 0.000260661 loss) | |
| I0623 18:00:48.735149 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000887116 (* 0.0272727 = 2.41941e-05 loss) | |
| I0623 18:00:48.735164 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000288889 (* 0.0272727 = 7.87878e-06 loss) | |
| I0623 18:00:48.735178 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 1.04907e-05 (* 0.0272727 = 2.86109e-07 loss) | |
| I0623 18:00:48.735193 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.1385e-05 (* 0.0272727 = 3.105e-07 loss) | |
| I0623 18:00:48.735204 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.875 | |
| I0623 18:00:48.735219 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 18:00:48.735231 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 18:00:48.735244 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 18:00:48.735255 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 18:00:48.735265 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 18:00:48.735277 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 18:00:48.735288 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 18:00:48.735301 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 18:00:48.735311 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 18:00:48.735323 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 18:00:48.735334 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.375 | |
| I0623 18:00:48.735347 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 18:00:48.735357 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 18:00:48.735369 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 18:00:48.735380 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 18:00:48.735393 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 18:00:48.735414 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:00:48.735426 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:00:48.735438 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:00:48.735450 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:00:48.735461 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:00:48.735472 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:00:48.735484 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.897727 | |
| I0623 18:00:48.735496 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.931818 | |
| I0623 18:00:48.735509 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.594083 (* 1 = 0.594083 loss) | |
| I0623 18:00:48.735522 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.460154 (* 1 = 0.460154 loss) | |
| I0623 18:00:48.735538 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0143969 (* 0.0909091 = 0.00130881 loss) | |
| I0623 18:00:48.735551 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.039081 (* 0.0909091 = 0.00355281 loss) | |
| I0623 18:00:48.735564 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.102585 (* 0.0909091 = 0.00932595 loss) | |
| I0623 18:00:48.735579 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.137611 (* 0.0909091 = 0.0125101 loss) | |
| I0623 18:00:48.735592 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.0949161 (* 0.0909091 = 0.00862874 loss) | |
| I0623 18:00:48.735623 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.0379471 (* 0.0909091 = 0.00344974 loss) | |
| I0623 18:00:48.735637 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.0520084 (* 0.0909091 = 0.00472804 loss) | |
| I0623 18:00:48.735652 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.158343 (* 0.0909091 = 0.0143948 loss) | |
| I0623 18:00:48.735666 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.61817 (* 0.0909091 = 0.147106 loss) | |
| I0623 18:00:48.735679 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.9113 (* 0.0909091 = 0.173754 loss) | |
| I0623 18:00:48.735693 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 2.5526 (* 0.0909091 = 0.232055 loss) | |
| I0623 18:00:48.735707 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.47336 (* 0.0909091 = 0.133942 loss) | |
| I0623 18:00:48.735720 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.39707 (* 0.0909091 = 0.127007 loss) | |
| I0623 18:00:48.735733 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.6421 (* 0.0909091 = 0.149282 loss) | |
| I0623 18:00:48.735748 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 1.27587 (* 0.0909091 = 0.115988 loss) | |
| I0623 18:00:48.735760 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 1.03158 (* 0.0909091 = 0.0937803 loss) | |
| I0623 18:00:48.735774 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.210846 (* 0.0909091 = 0.0191678 loss) | |
| I0623 18:00:48.735787 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0376946 (* 0.0909091 = 0.00342678 loss) | |
| I0623 18:00:48.735801 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00202212 (* 0.0909091 = 0.000183829 loss) | |
| I0623 18:00:48.735816 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000335729 (* 0.0909091 = 3.05208e-05 loss) | |
| I0623 18:00:48.735829 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000167639 (* 0.0909091 = 1.52399e-05 loss) | |
| I0623 18:00:48.735843 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 7.37623e-06 (* 0.0909091 = 6.70566e-07 loss) | |
| I0623 18:00:48.735855 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 18:00:48.735864 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 18:00:48.735872 10365 solver.cpp:245] Train net output #149: total_confidence = 0.271907 | |
| I0623 18:00:48.735895 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.239194 | |
| I0623 18:00:48.735910 10365 sgd_solver.cpp:106] Iteration 17500, lr = 0.001 | |
| I0623 18:00:56.762245 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.7845 > 30) by scale factor 0.974518 | |
| I0623 18:02:21.863108 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.2111 > 30) by scale factor 0.746063 | |
| I0623 18:04:29.864533 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 46.4021 > 30) by scale factor 0.646522 | |
| I0623 18:06:12.531642 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.693 > 30) by scale factor 0.94658 | |
| I0623 18:07:11.908936 10365 solver.cpp:229] Iteration 18000, loss = 4.48089 | |
| I0623 18:07:11.909078 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.544304 | |
| I0623 18:07:11.909098 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 18:07:11.909112 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 18:07:11.909126 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.625 | |
| I0623 18:07:11.909137 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 18:07:11.909152 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.625 | |
| I0623 18:07:11.909163 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 18:07:11.909176 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 18:07:11.909188 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 18:07:11.909200 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 18:07:11.909212 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 18:07:11.909224 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 18:07:11.909236 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 18:07:11.909248 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.75 | |
| I0623 18:07:11.909262 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 18:07:11.909276 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 1 | |
| I0623 18:07:11.909287 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 1 | |
| I0623 18:07:11.909299 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 18:07:11.909310 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:07:11.909323 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:07:11.909335 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:07:11.909346 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:07:11.909358 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:07:11.909369 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.789773 | |
| I0623 18:07:11.909381 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.810127 | |
| I0623 18:07:11.909397 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.39967 (* 0.3 = 0.419902 loss) | |
| I0623 18:07:11.909412 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.648673 (* 0.3 = 0.194602 loss) | |
| I0623 18:07:11.909427 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.343728 (* 0.0272727 = 0.00937439 loss) | |
| I0623 18:07:11.909441 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.984038 (* 0.0272727 = 0.0268374 loss) | |
| I0623 18:07:11.909456 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.24276 (* 0.0272727 = 0.0338934 loss) | |
| I0623 18:07:11.909469 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.0776 (* 0.0272727 = 0.0293891 loss) | |
| I0623 18:07:11.909482 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.20325 (* 0.0272727 = 0.032816 loss) | |
| I0623 18:07:11.909497 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.67704 (* 0.0272727 = 0.0457374 loss) | |
| I0623 18:07:11.909510 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.21468 (* 0.0272727 = 0.0331276 loss) | |
| I0623 18:07:11.909524 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.46313 (* 0.0272727 = 0.0399037 loss) | |
| I0623 18:07:11.909538 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.53533 (* 0.0272727 = 0.0418726 loss) | |
| I0623 18:07:11.909553 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.22985 (* 0.0272727 = 0.0335414 loss) | |
| I0623 18:07:11.909565 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.64547 (* 0.0272727 = 0.0448764 loss) | |
| I0623 18:07:11.909579 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.38899 (* 0.0272727 = 0.0378817 loss) | |
| I0623 18:07:11.909610 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 0.952298 (* 0.0272727 = 0.0259717 loss) | |
| I0623 18:07:11.909626 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.759527 (* 0.0272727 = 0.0207144 loss) | |
| I0623 18:07:11.909639 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.0810432 (* 0.0272727 = 0.00221027 loss) | |
| I0623 18:07:11.909653 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.00878042 (* 0.0272727 = 0.000239466 loss) | |
| I0623 18:07:11.909667 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.000714553 (* 0.0272727 = 1.94878e-05 loss) | |
| I0623 18:07:11.909682 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 5.26989e-05 (* 0.0272727 = 1.43724e-06 loss) | |
| I0623 18:07:11.909696 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 4.06806e-06 (* 0.0272727 = 1.10947e-07 loss) | |
| I0623 18:07:11.909710 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 1.04308e-06 (* 0.0272727 = 2.84477e-08 loss) | |
| I0623 18:07:11.909724 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 5.36442e-07 (* 0.0272727 = 1.46302e-08 loss) | |
| I0623 18:07:11.909739 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 2.98023e-08 (* 0.0272727 = 8.12791e-10 loss) | |
| I0623 18:07:11.909750 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.696203 | |
| I0623 18:07:11.909764 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 18:07:11.909775 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 18:07:11.909786 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 18:07:11.909798 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 18:07:11.909809 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 18:07:11.909821 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 18:07:11.909832 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 18:07:11.909844 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.625 | |
| I0623 18:07:11.909855 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.75 | |
| I0623 18:07:11.909868 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.75 | |
| I0623 18:07:11.909878 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 18:07:11.909889 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 18:07:11.909901 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 18:07:11.909912 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 18:07:11.909924 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 1 | |
| I0623 18:07:11.909935 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 18:07:11.909947 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 18:07:11.909958 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:07:11.909970 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:07:11.909981 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:07:11.909992 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:07:11.910004 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:07:11.910015 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.857955 | |
| I0623 18:07:11.910027 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.848101 | |
| I0623 18:07:11.910042 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 0.989374 (* 0.3 = 0.296812 loss) | |
| I0623 18:07:11.910054 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.478526 (* 0.3 = 0.143558 loss) | |
| I0623 18:07:11.910068 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.534094 (* 0.0272727 = 0.0145662 loss) | |
| I0623 18:07:11.910082 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.266665 (* 0.0272727 = 0.00727268 loss) | |
| I0623 18:07:11.910111 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.186941 (* 0.0272727 = 0.00509838 loss) | |
| I0623 18:07:11.910126 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.743793 (* 0.0272727 = 0.0202853 loss) | |
| I0623 18:07:11.910140 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.03393 (* 0.0272727 = 0.028198 loss) | |
| I0623 18:07:11.910154 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.09667 (* 0.0272727 = 0.0299091 loss) | |
| I0623 18:07:11.910168 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.40128 (* 0.0272727 = 0.0382166 loss) | |
| I0623 18:07:11.910181 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.04816 (* 0.0272727 = 0.0285862 loss) | |
| I0623 18:07:11.910195 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 0.962212 (* 0.0272727 = 0.0262422 loss) | |
| I0623 18:07:11.910208 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.00578 (* 0.0272727 = 0.0274305 loss) | |
| I0623 18:07:11.910223 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.60848 (* 0.0272727 = 0.0438677 loss) | |
| I0623 18:07:11.910235 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.64047 (* 0.0272727 = 0.0447401 loss) | |
| I0623 18:07:11.910250 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.777915 (* 0.0272727 = 0.0212159 loss) | |
| I0623 18:07:11.910264 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.798775 (* 0.0272727 = 0.0217848 loss) | |
| I0623 18:07:11.910279 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.114016 (* 0.0272727 = 0.00310954 loss) | |
| I0623 18:07:11.910291 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.013448 (* 0.0272727 = 0.000366763 loss) | |
| I0623 18:07:11.910305 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.00176165 (* 0.0272727 = 4.80449e-05 loss) | |
| I0623 18:07:11.910322 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.000454417 (* 0.0272727 = 1.23932e-05 loss) | |
| I0623 18:07:11.910336 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 3.9081e-05 (* 0.0272727 = 1.06585e-06 loss) | |
| I0623 18:07:11.910351 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 1.76138e-05 (* 0.0272727 = 4.80377e-07 loss) | |
| I0623 18:07:11.910364 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 9.67099e-06 (* 0.0272727 = 2.63754e-07 loss) | |
| I0623 18:07:11.910379 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 3.24847e-06 (* 0.0272727 = 8.85947e-08 loss) | |
| I0623 18:07:11.910392 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.848101 | |
| I0623 18:07:11.910403 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.875 | |
| I0623 18:07:11.910415 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 18:07:11.910426 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 18:07:11.910439 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 18:07:11.910449 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 18:07:11.910461 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 18:07:11.910472 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 18:07:11.910483 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 18:07:11.910495 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 18:07:11.910506 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 18:07:11.910517 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 18:07:11.910529 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.5 | |
| I0623 18:07:11.910540 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 18:07:11.910552 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 18:07:11.910563 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 1 | |
| I0623 18:07:11.910575 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 18:07:11.910596 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:07:11.910609 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:07:11.910620 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:07:11.910631 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:07:11.910643 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:07:11.910655 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:07:11.910665 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.920455 | |
| I0623 18:07:11.910677 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.949367 | |
| I0623 18:07:11.910691 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.494675 (* 1 = 0.494675 loss) | |
| I0623 18:07:11.910704 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.267532 (* 1 = 0.267532 loss) | |
| I0623 18:07:11.910718 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.485924 (* 0.0909091 = 0.0441749 loss) | |
| I0623 18:07:11.910733 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.064071 (* 0.0909091 = 0.00582463 loss) | |
| I0623 18:07:11.910743 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0258527 (* 0.0909091 = 0.00235024 loss) | |
| I0623 18:07:11.910753 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.046736 (* 0.0909091 = 0.00424873 loss) | |
| I0623 18:07:11.910768 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.0220015 (* 0.0909091 = 0.00200013 loss) | |
| I0623 18:07:11.910781 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.0637698 (* 0.0909091 = 0.00579725 loss) | |
| I0623 18:07:11.910794 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.848043 (* 0.0909091 = 0.0770948 loss) | |
| I0623 18:07:11.910809 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.203326 (* 0.0909091 = 0.0184842 loss) | |
| I0623 18:07:11.910822 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.789733 (* 0.0909091 = 0.071794 loss) | |
| I0623 18:07:11.910835 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.755641 (* 0.0909091 = 0.0686947 loss) | |
| I0623 18:07:11.910850 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.19038 (* 0.0909091 = 0.108216 loss) | |
| I0623 18:07:11.910862 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.1577 (* 0.0909091 = 0.105245 loss) | |
| I0623 18:07:11.910876 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.969188 (* 0.0909091 = 0.088108 loss) | |
| I0623 18:07:11.910889 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.640839 (* 0.0909091 = 0.0582581 loss) | |
| I0623 18:07:11.910903 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.109741 (* 0.0909091 = 0.00997641 loss) | |
| I0623 18:07:11.910917 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.00900497 (* 0.0909091 = 0.000818633 loss) | |
| I0623 18:07:11.910931 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.000681151 (* 0.0909091 = 6.19228e-05 loss) | |
| I0623 18:07:11.910945 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000106506 (* 0.0909091 = 9.68236e-06 loss) | |
| I0623 18:07:11.910959 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 3.03633e-05 (* 0.0909091 = 2.7603e-06 loss) | |
| I0623 18:07:11.910972 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 2.43351e-05 (* 0.0909091 = 2.21228e-06 loss) | |
| I0623 18:07:11.910986 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 1.29793e-05 (* 0.0909091 = 1.17994e-06 loss) | |
| I0623 18:07:11.911000 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 9.83478e-07 (* 0.0909091 = 8.94071e-08 loss) | |
| I0623 18:07:11.911012 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.5 | |
| I0623 18:07:11.911023 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 18:07:11.911044 10365 solver.cpp:245] Train net output #149: total_confidence = 0.329733 | |
| I0623 18:07:11.911057 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.27058 | |
| I0623 18:07:11.911070 10365 sgd_solver.cpp:106] Iteration 18000, lr = 0.001 | |
| I0623 18:09:32.573993 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 43.7649 > 30) by scale factor 0.685481 | |
| I0623 18:09:55.588659 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.9557 > 30) by scale factor 0.858228 | |
| I0623 18:10:36.978855 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.427 > 30) by scale factor 0.7807 | |
| I0623 18:12:38.825834 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.026 > 30) by scale factor 0.881678 | |
| I0623 18:13:27.126118 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 67.6652 > 30) by scale factor 0.443359 | |
| I0623 18:13:35.215574 10365 solver.cpp:229] Iteration 18500, loss = 4.46008 | |
| I0623 18:13:35.215682 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.519231 | |
| I0623 18:13:35.215701 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 18:13:35.215715 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.875 | |
| I0623 18:13:35.215728 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.875 | |
| I0623 18:13:35.215741 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 18:13:35.215754 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.625 | |
| I0623 18:13:35.215767 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 18:13:35.215780 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 18:13:35.215793 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 18:13:35.215806 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 18:13:35.215818 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 18:13:35.215831 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 18:13:35.215844 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.875 | |
| I0623 18:13:35.215857 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.375 | |
| I0623 18:13:35.215868 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 18:13:35.215881 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 18:13:35.215893 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 18:13:35.215905 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 18:13:35.215916 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:13:35.215929 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:13:35.215940 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:13:35.215952 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:13:35.215965 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:13:35.215976 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.698864 | |
| I0623 18:13:35.215988 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.798077 | |
| I0623 18:13:35.216006 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.28909 (* 0.3 = 0.386728 loss) | |
| I0623 18:13:35.216020 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.817224 (* 0.3 = 0.245167 loss) | |
| I0623 18:13:35.216035 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.104238 (* 0.0272727 = 0.00284285 loss) | |
| I0623 18:13:35.216050 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.257711 (* 0.0272727 = 0.00702847 loss) | |
| I0623 18:13:35.216064 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 0.833951 (* 0.0272727 = 0.0227441 loss) | |
| I0623 18:13:35.216078 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.7165 (* 0.0272727 = 0.0468136 loss) | |
| I0623 18:13:35.216092 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.26669 (* 0.0272727 = 0.0345461 loss) | |
| I0623 18:13:35.216109 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.70862 (* 0.0272727 = 0.0465987 loss) | |
| I0623 18:13:35.216125 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.46257 (* 0.0272727 = 0.0398883 loss) | |
| I0623 18:13:35.216140 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.76732 (* 0.0272727 = 0.0481997 loss) | |
| I0623 18:13:35.216155 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.84279 (* 0.0272727 = 0.0502579 loss) | |
| I0623 18:13:35.216168 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.03947 (* 0.0272727 = 0.055622 loss) | |
| I0623 18:13:35.216182 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.70849 (* 0.0272727 = 0.0465953 loss) | |
| I0623 18:13:35.216197 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.2775 (* 0.0272727 = 0.034841 loss) | |
| I0623 18:13:35.216251 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.47958 (* 0.0272727 = 0.0403522 loss) | |
| I0623 18:13:35.216267 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.32747 (* 0.0272727 = 0.0362037 loss) | |
| I0623 18:13:35.216281 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.15374 (* 0.0272727 = 0.0314655 loss) | |
| I0623 18:13:35.216295 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.849994 (* 0.0272727 = 0.0231817 loss) | |
| I0623 18:13:35.216310 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0825569 (* 0.0272727 = 0.00225155 loss) | |
| I0623 18:13:35.216325 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0173534 (* 0.0272727 = 0.000473273 loss) | |
| I0623 18:13:35.216339 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00211542 (* 0.0272727 = 5.76932e-05 loss) | |
| I0623 18:13:35.216354 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000403842 (* 0.0272727 = 1.10139e-05 loss) | |
| I0623 18:13:35.216368 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000206977 (* 0.0272727 = 5.64482e-06 loss) | |
| I0623 18:13:35.216387 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.85462e-05 (* 0.0272727 = 1.05126e-06 loss) | |
| I0623 18:13:35.216399 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.605769 | |
| I0623 18:13:35.216413 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 18:13:35.216424 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 18:13:35.216437 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 18:13:35.216449 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 18:13:35.216461 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.875 | |
| I0623 18:13:35.216472 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.875 | |
| I0623 18:13:35.216485 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 18:13:35.216496 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.25 | |
| I0623 18:13:35.216508 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 18:13:35.216521 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 18:13:35.216532 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 18:13:35.216544 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 18:13:35.216557 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 18:13:35.216567 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.375 | |
| I0623 18:13:35.216579 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.5 | |
| I0623 18:13:35.216591 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 18:13:35.216603 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 18:13:35.216614 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:13:35.216626 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:13:35.216639 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:13:35.216650 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:13:35.216661 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:13:35.216672 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.761364 | |
| I0623 18:13:35.216684 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.855769 | |
| I0623 18:13:35.216699 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.1533 (* 0.3 = 0.345991 loss) | |
| I0623 18:13:35.216712 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.708994 (* 0.3 = 0.212698 loss) | |
| I0623 18:13:35.216727 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.366384 (* 0.0272727 = 0.00999229 loss) | |
| I0623 18:13:35.216753 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.178486 (* 0.0272727 = 0.00486779 loss) | |
| I0623 18:13:35.216768 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.349324 (* 0.0272727 = 0.00952701 loss) | |
| I0623 18:13:35.216783 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.01338 (* 0.0272727 = 0.0276376 loss) | |
| I0623 18:13:35.216797 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 0.746746 (* 0.0272727 = 0.0203658 loss) | |
| I0623 18:13:35.216811 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.09575 (* 0.0272727 = 0.0298841 loss) | |
| I0623 18:13:35.216825 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.5587 (* 0.0272727 = 0.0425101 loss) | |
| I0623 18:13:35.216840 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.46646 (* 0.0272727 = 0.0399944 loss) | |
| I0623 18:13:35.216852 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.43902 (* 0.0272727 = 0.0392459 loss) | |
| I0623 18:13:35.216866 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.54755 (* 0.0272727 = 0.0422059 loss) | |
| I0623 18:13:35.216881 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.35941 (* 0.0272727 = 0.0370749 loss) | |
| I0623 18:13:35.216893 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.43977 (* 0.0272727 = 0.0392664 loss) | |
| I0623 18:13:35.216907 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.05149 (* 0.0272727 = 0.0286771 loss) | |
| I0623 18:13:35.216922 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.36058 (* 0.0272727 = 0.0371066 loss) | |
| I0623 18:13:35.216935 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.09794 (* 0.0272727 = 0.0299438 loss) | |
| I0623 18:13:35.216949 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.609367 (* 0.0272727 = 0.0166191 loss) | |
| I0623 18:13:35.216964 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0531756 (* 0.0272727 = 0.00145024 loss) | |
| I0623 18:13:35.216977 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0212325 (* 0.0272727 = 0.000579067 loss) | |
| I0623 18:13:35.216991 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00342409 (* 0.0272727 = 9.33842e-05 loss) | |
| I0623 18:13:35.217005 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00260366 (* 0.0272727 = 7.10089e-05 loss) | |
| I0623 18:13:35.217020 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00139613 (* 0.0272727 = 3.80763e-05 loss) | |
| I0623 18:13:35.217033 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000998447 (* 0.0272727 = 2.72304e-05 loss) | |
| I0623 18:13:35.217046 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.894231 | |
| I0623 18:13:35.217058 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 18:13:35.217069 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 18:13:35.217082 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 18:13:35.217093 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 18:13:35.217104 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 18:13:35.217116 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 18:13:35.217128 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 18:13:35.217139 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 18:13:35.217151 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 1 | |
| I0623 18:13:35.217164 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 18:13:35.217178 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 18:13:35.217190 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 18:13:35.217202 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 18:13:35.217214 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 18:13:35.217226 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.5 | |
| I0623 18:13:35.217248 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 18:13:35.217262 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:13:35.217273 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:13:35.217286 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:13:35.217298 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:13:35.217309 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:13:35.217321 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:13:35.217330 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.926136 | |
| I0623 18:13:35.217339 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 1 | |
| I0623 18:13:35.217347 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.349244 (* 1 = 0.349244 loss) | |
| I0623 18:13:35.217357 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.229314 (* 1 = 0.229314 loss) | |
| I0623 18:13:35.217372 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0177366 (* 0.0909091 = 0.00161242 loss) | |
| I0623 18:13:35.217386 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0323148 (* 0.0909091 = 0.00293771 loss) | |
| I0623 18:13:35.217401 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.044865 (* 0.0909091 = 0.00407864 loss) | |
| I0623 18:13:35.217414 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0768077 (* 0.0909091 = 0.00698252 loss) | |
| I0623 18:13:35.217432 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.0523965 (* 0.0909091 = 0.00476332 loss) | |
| I0623 18:13:35.217447 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.0895117 (* 0.0909091 = 0.00813743 loss) | |
| I0623 18:13:35.217461 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.476593 (* 0.0909091 = 0.0433266 loss) | |
| I0623 18:13:35.217475 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.151705 (* 0.0909091 = 0.0137914 loss) | |
| I0623 18:13:35.217489 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.258067 (* 0.0909091 = 0.0234606 loss) | |
| I0623 18:13:35.217504 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.611302 (* 0.0909091 = 0.0555729 loss) | |
| I0623 18:13:35.217516 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.412978 (* 0.0909091 = 0.0375434 loss) | |
| I0623 18:13:35.217530 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.815917 (* 0.0909091 = 0.0741743 loss) | |
| I0623 18:13:35.217545 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.756472 (* 0.0909091 = 0.0687702 loss) | |
| I0623 18:13:35.217557 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.927886 (* 0.0909091 = 0.0843533 loss) | |
| I0623 18:13:35.217571 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.898386 (* 0.0909091 = 0.0816715 loss) | |
| I0623 18:13:35.217586 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.186753 (* 0.0909091 = 0.0169775 loss) | |
| I0623 18:13:35.217599 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0504796 (* 0.0909091 = 0.00458906 loss) | |
| I0623 18:13:35.217613 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00229147 (* 0.0909091 = 0.000208316 loss) | |
| I0623 18:13:35.217628 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000524091 (* 0.0909091 = 4.76446e-05 loss) | |
| I0623 18:13:35.217641 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000194689 (* 0.0909091 = 1.7699e-05 loss) | |
| I0623 18:13:35.217655 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000149974 (* 0.0909091 = 1.3634e-05 loss) | |
| I0623 18:13:35.217669 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 1.01776e-05 (* 0.0909091 = 9.25235e-07 loss) | |
| I0623 18:13:35.217682 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 18:13:35.217694 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 18:13:35.217715 10365 solver.cpp:245] Train net output #149: total_confidence = 0.146849 | |
| I0623 18:13:35.217730 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.14585 | |
| I0623 18:13:35.217742 10365 sgd_solver.cpp:106] Iteration 18500, lr = 0.001 | |
| I0623 18:14:13.173852 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.0437 > 30) by scale factor 0.998544 | |
| I0623 18:15:16.778084 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 43.9493 > 30) by scale factor 0.682604 | |
| I0623 18:19:12.026722 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 49.7948 > 30) by scale factor 0.602472 | |
| I0623 18:19:31.183141 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.9614 > 30) by scale factor 0.811658 | |
| I0623 18:19:58.394333 10365 solver.cpp:229] Iteration 19000, loss = 4.4366 | |
| I0623 18:19:58.394439 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.561798 | |
| I0623 18:19:58.394459 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 18:19:58.394474 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.875 | |
| I0623 18:19:58.394486 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 18:19:58.394500 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 18:19:58.394513 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.125 | |
| I0623 18:19:58.394526 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 18:19:58.394538 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 18:19:58.394551 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 18:19:58.394562 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 18:19:58.394575 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 18:19:58.394587 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.625 | |
| I0623 18:19:58.394600 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.875 | |
| I0623 18:19:58.394613 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 18:19:58.394623 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 18:19:58.394635 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 18:19:58.394647 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 1 | |
| I0623 18:19:58.394659 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 18:19:58.394670 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:19:58.394682 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:19:58.394693 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:19:58.394706 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:19:58.394716 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:19:58.394733 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.761364 | |
| I0623 18:19:58.394747 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.808989 | |
| I0623 18:19:58.394763 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.4483 (* 0.3 = 0.434491 loss) | |
| I0623 18:19:58.394778 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.800616 (* 0.3 = 0.240185 loss) | |
| I0623 18:19:58.394793 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.496638 (* 0.0272727 = 0.0135447 loss) | |
| I0623 18:19:58.394807 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.569521 (* 0.0272727 = 0.0155324 loss) | |
| I0623 18:19:58.394820 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.53689 (* 0.0272727 = 0.0419152 loss) | |
| I0623 18:19:58.394834 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.1904 (* 0.0272727 = 0.0597382 loss) | |
| I0623 18:19:58.394848 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.28905 (* 0.0272727 = 0.0624286 loss) | |
| I0623 18:19:58.394862 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.38138 (* 0.0272727 = 0.0376739 loss) | |
| I0623 18:19:58.394876 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.29429 (* 0.0272727 = 0.0352989 loss) | |
| I0623 18:19:58.394889 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.80908 (* 0.0272727 = 0.0493385 loss) | |
| I0623 18:19:58.394904 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.52983 (* 0.0272727 = 0.0417227 loss) | |
| I0623 18:19:58.394917 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.32082 (* 0.0272727 = 0.0360224 loss) | |
| I0623 18:19:58.394932 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.26478 (* 0.0272727 = 0.0344939 loss) | |
| I0623 18:19:58.394945 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 0.689919 (* 0.0272727 = 0.018816 loss) | |
| I0623 18:19:58.394978 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.37442 (* 0.0272727 = 0.0374841 loss) | |
| I0623 18:19:58.394992 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.828773 (* 0.0272727 = 0.0226029 loss) | |
| I0623 18:19:58.395005 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.685733 (* 0.0272727 = 0.0187018 loss) | |
| I0623 18:19:58.395020 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.402437 (* 0.0272727 = 0.0109756 loss) | |
| I0623 18:19:58.395035 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0100637 (* 0.0272727 = 0.000274465 loss) | |
| I0623 18:19:58.395048 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0014355 (* 0.0272727 = 3.915e-05 loss) | |
| I0623 18:19:58.395062 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.000408243 (* 0.0272727 = 1.11339e-05 loss) | |
| I0623 18:19:58.395076 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 8.91003e-05 (* 0.0272727 = 2.43001e-06 loss) | |
| I0623 18:19:58.395090 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 1.92235e-05 (* 0.0272727 = 5.24278e-07 loss) | |
| I0623 18:19:58.395105 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 4.72374e-06 (* 0.0272727 = 1.28829e-07 loss) | |
| I0623 18:19:58.395117 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.573034 | |
| I0623 18:19:58.395131 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 18:19:58.395144 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 18:19:58.395156 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 18:19:58.395167 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 18:19:58.395179 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 18:19:58.395191 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 18:19:58.395202 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 18:19:58.395215 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.625 | |
| I0623 18:19:58.395226 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 18:19:58.395236 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 18:19:58.395248 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.75 | |
| I0623 18:19:58.395259 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 18:19:58.395270 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 18:19:58.395282 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 18:19:58.395293 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 18:19:58.395305 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 18:19:58.395316 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 18:19:58.395328 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:19:58.395339 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:19:58.395350 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:19:58.395362 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:19:58.395373 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:19:58.395385 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.75 | |
| I0623 18:19:58.395395 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.853933 | |
| I0623 18:19:58.395409 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.14068 (* 0.3 = 0.342205 loss) | |
| I0623 18:19:58.395423 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.647838 (* 0.3 = 0.194351 loss) | |
| I0623 18:19:58.395437 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.24778 (* 0.0272727 = 0.00675764 loss) | |
| I0623 18:19:58.395450 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.215671 (* 0.0272727 = 0.00588194 loss) | |
| I0623 18:19:58.395476 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.569533 (* 0.0272727 = 0.0155327 loss) | |
| I0623 18:19:58.395490 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.5444 (* 0.0272727 = 0.0421201 loss) | |
| I0623 18:19:58.395504 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.1354 (* 0.0272727 = 0.0309655 loss) | |
| I0623 18:19:58.395519 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.09117 (* 0.0272727 = 0.0297592 loss) | |
| I0623 18:19:58.395532 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.16903 (* 0.0272727 = 0.0318827 loss) | |
| I0623 18:19:58.395545 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.28953 (* 0.0272727 = 0.0351689 loss) | |
| I0623 18:19:58.395560 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.7634 (* 0.0272727 = 0.0480926 loss) | |
| I0623 18:19:58.395572 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.72798 (* 0.0272727 = 0.0471268 loss) | |
| I0623 18:19:58.395586 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 0.840667 (* 0.0272727 = 0.0229273 loss) | |
| I0623 18:19:58.395613 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 0.907671 (* 0.0272727 = 0.0247547 loss) | |
| I0623 18:19:58.395630 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.855826 (* 0.0272727 = 0.0233407 loss) | |
| I0623 18:19:58.395644 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.664141 (* 0.0272727 = 0.0181129 loss) | |
| I0623 18:19:58.395658 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.740842 (* 0.0272727 = 0.0202048 loss) | |
| I0623 18:19:58.395671 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.736317 (* 0.0272727 = 0.0200814 loss) | |
| I0623 18:19:58.395685 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.147411 (* 0.0272727 = 0.00402031 loss) | |
| I0623 18:19:58.395699 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0705079 (* 0.0272727 = 0.00192294 loss) | |
| I0623 18:19:58.395712 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0111062 (* 0.0272727 = 0.000302897 loss) | |
| I0623 18:19:58.395726 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00144027 (* 0.0272727 = 3.92801e-05 loss) | |
| I0623 18:19:58.395740 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000361488 (* 0.0272727 = 9.85877e-06 loss) | |
| I0623 18:19:58.395755 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.27409e-05 (* 0.0272727 = 3.4748e-07 loss) | |
| I0623 18:19:58.395766 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.831461 | |
| I0623 18:19:58.395782 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 18:19:58.395795 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 18:19:58.395807 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 18:19:58.395818 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 18:19:58.395829 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 18:19:58.395840 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 18:19:58.395853 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 18:19:58.395864 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 18:19:58.395875 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 18:19:58.395886 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 18:19:58.395897 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 18:19:58.395910 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 18:19:58.395920 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 18:19:58.395931 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 18:19:58.395943 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 18:19:58.395954 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 18:19:58.395977 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:19:58.395990 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:19:58.396001 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:19:58.396013 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:19:58.396024 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:19:58.396035 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:19:58.396047 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.892045 | |
| I0623 18:19:58.396059 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.988764 | |
| I0623 18:19:58.396072 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.522394 (* 1 = 0.522394 loss) | |
| I0623 18:19:58.396086 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.337006 (* 1 = 0.337006 loss) | |
| I0623 18:19:58.396100 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0563238 (* 0.0909091 = 0.00512034 loss) | |
| I0623 18:19:58.396114 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0881131 (* 0.0909091 = 0.00801028 loss) | |
| I0623 18:19:58.396128 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.106982 (* 0.0909091 = 0.00972562 loss) | |
| I0623 18:19:58.396142 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.160097 (* 0.0909091 = 0.0145543 loss) | |
| I0623 18:19:58.396155 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.172627 (* 0.0909091 = 0.0156934 loss) | |
| I0623 18:19:58.396169 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.21184 (* 0.0909091 = 0.0192582 loss) | |
| I0623 18:19:58.396185 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.229515 (* 0.0909091 = 0.020865 loss) | |
| I0623 18:19:58.396199 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.723382 (* 0.0909091 = 0.065762 loss) | |
| I0623 18:19:58.396214 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.830792 (* 0.0909091 = 0.0755265 loss) | |
| I0623 18:19:58.396226 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.06503 (* 0.0909091 = 0.0968209 loss) | |
| I0623 18:19:58.396240 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.535471 (* 0.0909091 = 0.0486792 loss) | |
| I0623 18:19:58.396255 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.822923 (* 0.0909091 = 0.0748112 loss) | |
| I0623 18:19:58.396267 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.634303 (* 0.0909091 = 0.0576639 loss) | |
| I0623 18:19:58.396281 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.87515 (* 0.0909091 = 0.0795591 loss) | |
| I0623 18:19:58.396294 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.348024 (* 0.0909091 = 0.0316385 loss) | |
| I0623 18:19:58.396308 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.681751 (* 0.0909091 = 0.0619773 loss) | |
| I0623 18:19:58.396322 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.00344075 (* 0.0909091 = 0.000312795 loss) | |
| I0623 18:19:58.396335 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000837303 (* 0.0909091 = 7.61184e-05 loss) | |
| I0623 18:19:58.396349 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000116505 (* 0.0909091 = 1.05914e-05 loss) | |
| I0623 18:19:58.396363 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 6.32032e-05 (* 0.0909091 = 5.74574e-06 loss) | |
| I0623 18:19:58.396378 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 3.89692e-05 (* 0.0909091 = 3.54265e-06 loss) | |
| I0623 18:19:58.396391 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 7.28675e-06 (* 0.0909091 = 6.62432e-07 loss) | |
| I0623 18:19:58.396402 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 18:19:58.396414 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 18:19:58.396437 10365 solver.cpp:245] Train net output #149: total_confidence = 0.247096 | |
| I0623 18:19:58.396450 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.212557 | |
| I0623 18:19:58.396463 10365 sgd_solver.cpp:106] Iteration 19000, lr = 0.001 | |
| I0623 18:21:19.252107 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 43.0154 > 30) by scale factor 0.697425 | |
| I0623 18:24:14.751392 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.1141 > 30) by scale factor 0.830702 | |
| I0623 18:26:21.609807 10365 solver.cpp:229] Iteration 19500, loss = 4.52639 | |
| I0623 18:26:21.609897 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.422222 | |
| I0623 18:26:21.609916 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 18:26:21.609930 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 18:26:21.609942 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 18:26:21.609954 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 18:26:21.609967 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.5 | |
| I0623 18:26:21.609979 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.625 | |
| I0623 18:26:21.609992 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 18:26:21.610005 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 18:26:21.610018 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 18:26:21.610030 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 18:26:21.610043 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 18:26:21.610055 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 18:26:21.610067 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 18:26:21.610081 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 18:26:21.610095 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 18:26:21.610106 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 18:26:21.610117 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 18:26:21.610129 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:26:21.610141 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:26:21.610153 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:26:21.610165 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:26:21.610177 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:26:21.610188 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.693182 | |
| I0623 18:26:21.610200 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.788889 | |
| I0623 18:26:21.610216 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.62005 (* 0.3 = 0.486016 loss) | |
| I0623 18:26:21.610230 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.877283 (* 0.3 = 0.263185 loss) | |
| I0623 18:26:21.610245 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.634987 (* 0.0272727 = 0.0173178 loss) | |
| I0623 18:26:21.610260 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.925613 (* 0.0272727 = 0.025244 loss) | |
| I0623 18:26:21.610273 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.58459 (* 0.0272727 = 0.0704889 loss) | |
| I0623 18:26:21.610287 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.34681 (* 0.0272727 = 0.0367311 loss) | |
| I0623 18:26:21.610301 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.53748 (* 0.0272727 = 0.0419312 loss) | |
| I0623 18:26:21.610314 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 0.93002 (* 0.0272727 = 0.0253642 loss) | |
| I0623 18:26:21.610328 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.68787 (* 0.0272727 = 0.0460329 loss) | |
| I0623 18:26:21.610343 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.0293 (* 0.0272727 = 0.0280718 loss) | |
| I0623 18:26:21.610357 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.74106 (* 0.0272727 = 0.0474836 loss) | |
| I0623 18:26:21.610371 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.08348 (* 0.0272727 = 0.0295494 loss) | |
| I0623 18:26:21.610385 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.74345 (* 0.0272727 = 0.0475485 loss) | |
| I0623 18:26:21.610399 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.25896 (* 0.0272727 = 0.0343353 loss) | |
| I0623 18:26:21.610430 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.72376 (* 0.0272727 = 0.0470117 loss) | |
| I0623 18:26:21.610445 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.43609 (* 0.0272727 = 0.039166 loss) | |
| I0623 18:26:21.610460 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.729805 (* 0.0272727 = 0.0199038 loss) | |
| I0623 18:26:21.610473 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.603387 (* 0.0272727 = 0.016456 loss) | |
| I0623 18:26:21.610487 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0353391 (* 0.0272727 = 0.000963793 loss) | |
| I0623 18:26:21.610502 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00481085 (* 0.0272727 = 0.000131205 loss) | |
| I0623 18:26:21.610517 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.000311905 (* 0.0272727 = 8.5065e-06 loss) | |
| I0623 18:26:21.610535 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 8.59085e-05 (* 0.0272727 = 2.34296e-06 loss) | |
| I0623 18:26:21.610550 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 1.95068e-05 (* 0.0272727 = 5.32005e-07 loss) | |
| I0623 18:26:21.610564 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 6.70553e-07 (* 0.0272727 = 1.82878e-08 loss) | |
| I0623 18:26:21.610576 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.622222 | |
| I0623 18:26:21.610589 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 18:26:21.610600 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 18:26:21.610611 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 18:26:21.610623 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 1 | |
| I0623 18:26:21.610635 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 18:26:21.610646 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 18:26:21.610657 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 18:26:21.610669 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.5 | |
| I0623 18:26:21.610680 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 18:26:21.610692 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.75 | |
| I0623 18:26:21.610703 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.25 | |
| I0623 18:26:21.610715 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.375 | |
| I0623 18:26:21.610726 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.5 | |
| I0623 18:26:21.610738 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 18:26:21.610749 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 18:26:21.610760 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 18:26:21.610772 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 18:26:21.610783 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:26:21.610795 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:26:21.610806 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:26:21.610817 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:26:21.610829 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:26:21.610841 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.789773 | |
| I0623 18:26:21.610852 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.8 | |
| I0623 18:26:21.610865 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.34022 (* 0.3 = 0.402065 loss) | |
| I0623 18:26:21.610879 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.762504 (* 0.3 = 0.228751 loss) | |
| I0623 18:26:21.610893 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.194981 (* 0.0272727 = 0.00531766 loss) | |
| I0623 18:26:21.610908 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.713218 (* 0.0272727 = 0.0194514 loss) | |
| I0623 18:26:21.610932 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 1.43522 (* 0.0272727 = 0.0391425 loss) | |
| I0623 18:26:21.610947 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.503766 (* 0.0272727 = 0.0137391 loss) | |
| I0623 18:26:21.610962 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.71524 (* 0.0272727 = 0.0467794 loss) | |
| I0623 18:26:21.610976 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.18356 (* 0.0272727 = 0.0322789 loss) | |
| I0623 18:26:21.610990 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.7121 (* 0.0272727 = 0.0466936 loss) | |
| I0623 18:26:21.611003 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.38314 (* 0.0272727 = 0.0377219 loss) | |
| I0623 18:26:21.611016 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.38457 (* 0.0272727 = 0.0377611 loss) | |
| I0623 18:26:21.611030 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.27449 (* 0.0272727 = 0.0347587 loss) | |
| I0623 18:26:21.611044 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.4925 (* 0.0272727 = 0.0407046 loss) | |
| I0623 18:26:21.611057 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.46826 (* 0.0272727 = 0.0400434 loss) | |
| I0623 18:26:21.611070 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.33142 (* 0.0272727 = 0.0363115 loss) | |
| I0623 18:26:21.611084 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.32296 (* 0.0272727 = 0.0360807 loss) | |
| I0623 18:26:21.611098 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.07046 (* 0.0272727 = 0.0291944 loss) | |
| I0623 18:26:21.611111 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.863502 (* 0.0272727 = 0.0235501 loss) | |
| I0623 18:26:21.611127 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0639322 (* 0.0272727 = 0.00174361 loss) | |
| I0623 18:26:21.611142 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00939828 (* 0.0272727 = 0.000256317 loss) | |
| I0623 18:26:21.611156 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.000671022 (* 0.0272727 = 1.83006e-05 loss) | |
| I0623 18:26:21.611171 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000168089 (* 0.0272727 = 4.58425e-06 loss) | |
| I0623 18:26:21.611186 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 3.72058e-05 (* 0.0272727 = 1.0147e-06 loss) | |
| I0623 18:26:21.611199 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.98187e-06 (* 0.0272727 = 5.40509e-08 loss) | |
| I0623 18:26:21.611212 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.822222 | |
| I0623 18:26:21.611223 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 18:26:21.611234 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 18:26:21.611246 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.75 | |
| I0623 18:26:21.611258 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 18:26:21.611269 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 18:26:21.611280 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 18:26:21.611292 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 18:26:21.611304 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 1 | |
| I0623 18:26:21.611315 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 18:26:21.611327 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 18:26:21.611338 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 18:26:21.611351 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.75 | |
| I0623 18:26:21.611361 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.5 | |
| I0623 18:26:21.611373 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 18:26:21.611384 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 18:26:21.611397 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 18:26:21.611418 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:26:21.611430 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:26:21.611443 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:26:21.611454 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:26:21.611464 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:26:21.611476 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:26:21.611487 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.892045 | |
| I0623 18:26:21.611500 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.966667 | |
| I0623 18:26:21.611513 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.559292 (* 1 = 0.559292 loss) | |
| I0623 18:26:21.611527 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.373348 (* 1 = 0.373348 loss) | |
| I0623 18:26:21.611541 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0253321 (* 0.0909091 = 0.00230292 loss) | |
| I0623 18:26:21.611555 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.575523 (* 0.0909091 = 0.0523203 loss) | |
| I0623 18:26:21.611584 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.285894 (* 0.0909091 = 0.0259903 loss) | |
| I0623 18:26:21.611603 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0659277 (* 0.0909091 = 0.00599342 loss) | |
| I0623 18:26:21.611616 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.721681 (* 0.0909091 = 0.0656074 loss) | |
| I0623 18:26:21.611630 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.31757 (* 0.0909091 = 0.02887 loss) | |
| I0623 18:26:21.611644 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.994358 (* 0.0909091 = 0.0903962 loss) | |
| I0623 18:26:21.611659 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.126681 (* 0.0909091 = 0.0115164 loss) | |
| I0623 18:26:21.611672 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.424916 (* 0.0909091 = 0.0386287 loss) | |
| I0623 18:26:21.611685 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.521495 (* 0.0909091 = 0.0474087 loss) | |
| I0623 18:26:21.611699 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.03026 (* 0.0909091 = 0.0936601 loss) | |
| I0623 18:26:21.611712 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.891414 (* 0.0909091 = 0.0810376 loss) | |
| I0623 18:26:21.611726 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.00409 (* 0.0909091 = 0.0912812 loss) | |
| I0623 18:26:21.611739 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.756903 (* 0.0909091 = 0.0688094 loss) | |
| I0623 18:26:21.611753 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.865219 (* 0.0909091 = 0.0786563 loss) | |
| I0623 18:26:21.611768 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.116198 (* 0.0909091 = 0.0105635 loss) | |
| I0623 18:26:21.611780 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.127184 (* 0.0909091 = 0.0115622 loss) | |
| I0623 18:26:21.611794 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0161266 (* 0.0909091 = 0.00146605 loss) | |
| I0623 18:26:21.611809 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00180237 (* 0.0909091 = 0.000163851 loss) | |
| I0623 18:26:21.611822 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000322933 (* 0.0909091 = 2.93575e-05 loss) | |
| I0623 18:26:21.611836 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000242784 (* 0.0909091 = 2.20713e-05 loss) | |
| I0623 18:26:21.611850 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 6.37517e-05 (* 0.0909091 = 5.79561e-06 loss) | |
| I0623 18:26:21.611862 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 18:26:21.611873 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 18:26:21.611886 10365 solver.cpp:245] Train net output #149: total_confidence = 0.19066 | |
| I0623 18:26:21.611908 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.144479 | |
| I0623 18:26:21.611922 10365 sgd_solver.cpp:106] Iteration 19500, lr = 0.001 | |
| I0623 18:27:30.953132 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 37.5821 > 30) by scale factor 0.798251 | |
| I0623 18:30:36.458030 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.5687 > 30) by scale factor 0.981395 | |
| I0623 18:31:06.356283 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.4035 > 30) by scale factor 0.872005 | |
| I0623 18:32:44.449863 10365 solver.cpp:456] Snapshotting to binary proto file /mnt/snapshots/mixed_lstm22_iter_20000.caffemodel | |
| I0623 18:32:45.049746 10365 sgd_solver.cpp:273] Snapshotting solver state to binary proto file /mnt/snapshots/mixed_lstm22_iter_20000.solverstate | |
| I0623 18:32:45.334161 10365 solver.cpp:338] Iteration 20000, Testing net (#0) | |
| I0623 18:33:42.489037 10365 solver.cpp:393] Test loss: 3.77954 | |
| I0623 18:33:42.489163 10365 solver.cpp:406] Test net output #0: loss1/accuracy = 0.537847 | |
| I0623 18:33:42.489182 10365 solver.cpp:406] Test net output #1: loss1/accuracy01 = 0.937 | |
| I0623 18:33:42.489197 10365 solver.cpp:406] Test net output #2: loss1/accuracy02 = 0.801 | |
| I0623 18:33:42.489209 10365 solver.cpp:406] Test net output #3: loss1/accuracy03 = 0.581 | |
| I0623 18:33:42.489222 10365 solver.cpp:406] Test net output #4: loss1/accuracy04 = 0.502 | |
| I0623 18:33:42.489234 10365 solver.cpp:406] Test net output #5: loss1/accuracy05 = 0.421 | |
| I0623 18:33:42.489246 10365 solver.cpp:406] Test net output #6: loss1/accuracy06 = 0.449 | |
| I0623 18:33:42.489261 10365 solver.cpp:406] Test net output #7: loss1/accuracy07 = 0.427 | |
| I0623 18:33:42.489274 10365 solver.cpp:406] Test net output #8: loss1/accuracy08 = 0.497 | |
| I0623 18:33:42.489286 10365 solver.cpp:406] Test net output #9: loss1/accuracy09 = 0.452 | |
| I0623 18:33:42.489298 10365 solver.cpp:406] Test net output #10: loss1/accuracy10 = 0.427 | |
| I0623 18:33:42.489310 10365 solver.cpp:406] Test net output #11: loss1/accuracy11 = 0.415 | |
| I0623 18:33:42.489322 10365 solver.cpp:406] Test net output #12: loss1/accuracy12 = 0.493 | |
| I0623 18:33:42.489334 10365 solver.cpp:406] Test net output #13: loss1/accuracy13 = 0.593 | |
| I0623 18:33:42.489346 10365 solver.cpp:406] Test net output #14: loss1/accuracy14 = 0.684 | |
| I0623 18:33:42.489358 10365 solver.cpp:406] Test net output #15: loss1/accuracy15 = 0.783 | |
| I0623 18:33:42.489369 10365 solver.cpp:406] Test net output #16: loss1/accuracy16 = 0.833 | |
| I0623 18:33:42.489382 10365 solver.cpp:406] Test net output #17: loss1/accuracy17 = 0.909 | |
| I0623 18:33:42.489392 10365 solver.cpp:406] Test net output #18: loss1/accuracy18 = 0.952 | |
| I0623 18:33:42.489404 10365 solver.cpp:406] Test net output #19: loss1/accuracy19 = 0.972 | |
| I0623 18:33:42.489415 10365 solver.cpp:406] Test net output #20: loss1/accuracy20 = 0.987 | |
| I0623 18:33:42.489428 10365 solver.cpp:406] Test net output #21: loss1/accuracy21 = 0.999 | |
| I0623 18:33:42.489439 10365 solver.cpp:406] Test net output #22: loss1/accuracy22 = 1 | |
| I0623 18:33:42.489450 10365 solver.cpp:406] Test net output #23: loss1/accuracy_incl_empty = 0.714772 | |
| I0623 18:33:42.489462 10365 solver.cpp:406] Test net output #24: loss1/accuracy_top3 = 0.858661 | |
| I0623 18:33:42.489480 10365 solver.cpp:406] Test net output #25: loss1/cross_entropy_loss = 1.33372 (* 0.3 = 0.400115 loss) | |
| I0623 18:33:42.489493 10365 solver.cpp:406] Test net output #26: loss1/cross_entropy_loss_incl_empty = 0.822943 (* 0.3 = 0.246883 loss) | |
| I0623 18:33:42.489507 10365 solver.cpp:406] Test net output #27: loss1/loss01 = 0.300707 (* 0.0272727 = 0.00820111 loss) | |
| I0623 18:33:42.489521 10365 solver.cpp:406] Test net output #28: loss1/loss02 = 0.676593 (* 0.0272727 = 0.0184525 loss) | |
| I0623 18:33:42.489536 10365 solver.cpp:406] Test net output #29: loss1/loss03 = 1.30714 (* 0.0272727 = 0.0356493 loss) | |
| I0623 18:33:42.489549 10365 solver.cpp:406] Test net output #30: loss1/loss04 = 1.47333 (* 0.0272727 = 0.0401817 loss) | |
| I0623 18:33:42.489562 10365 solver.cpp:406] Test net output #31: loss1/loss05 = 1.62145 (* 0.0272727 = 0.0442214 loss) | |
| I0623 18:33:42.489576 10365 solver.cpp:406] Test net output #32: loss1/loss06 = 1.70078 (* 0.0272727 = 0.0463848 loss) | |
| I0623 18:33:42.489589 10365 solver.cpp:406] Test net output #33: loss1/loss07 = 1.73264 (* 0.0272727 = 0.0472538 loss) | |
| I0623 18:33:42.489603 10365 solver.cpp:406] Test net output #34: loss1/loss08 = 1.57808 (* 0.0272727 = 0.0430387 loss) | |
| I0623 18:33:42.489617 10365 solver.cpp:406] Test net output #35: loss1/loss09 = 1.66549 (* 0.0272727 = 0.0454224 loss) | |
| I0623 18:33:42.489630 10365 solver.cpp:406] Test net output #36: loss1/loss10 = 1.70699 (* 0.0272727 = 0.0465543 loss) | |
| I0623 18:33:42.489645 10365 solver.cpp:406] Test net output #37: loss1/loss11 = 1.77548 (* 0.0272727 = 0.0484222 loss) | |
| I0623 18:33:42.489657 10365 solver.cpp:406] Test net output #38: loss1/loss12 = 1.48579 (* 0.0272727 = 0.0405216 loss) | |
| I0623 18:33:42.489689 10365 solver.cpp:406] Test net output #39: loss1/loss13 = 1.19254 (* 0.0272727 = 0.0325238 loss) | |
| I0623 18:33:42.489706 10365 solver.cpp:406] Test net output #40: loss1/loss14 = 0.916804 (* 0.0272727 = 0.0250037 loss) | |
| I0623 18:33:42.489718 10365 solver.cpp:406] Test net output #41: loss1/loss15 = 0.665252 (* 0.0272727 = 0.0181432 loss) | |
| I0623 18:33:42.489732 10365 solver.cpp:406] Test net output #42: loss1/loss16 = 0.503395 (* 0.0272727 = 0.013729 loss) | |
| I0623 18:33:42.489747 10365 solver.cpp:406] Test net output #43: loss1/loss17 = 0.310623 (* 0.0272727 = 0.00847155 loss) | |
| I0623 18:33:42.489759 10365 solver.cpp:406] Test net output #44: loss1/loss18 = 0.184884 (* 0.0272727 = 0.00504229 loss) | |
| I0623 18:33:42.489773 10365 solver.cpp:406] Test net output #45: loss1/loss19 = 0.115352 (* 0.0272727 = 0.00314597 loss) | |
| I0623 18:33:42.489786 10365 solver.cpp:406] Test net output #46: loss1/loss20 = 0.0686231 (* 0.0272727 = 0.00187154 loss) | |
| I0623 18:33:42.489800 10365 solver.cpp:406] Test net output #47: loss1/loss21 = 0.00701989 (* 0.0272727 = 0.000191451 loss) | |
| I0623 18:33:42.489814 10365 solver.cpp:406] Test net output #48: loss1/loss22 = 7.3252e-05 (* 0.0272727 = 1.99778e-06 loss) | |
| I0623 18:33:42.489825 10365 solver.cpp:406] Test net output #49: loss2/accuracy = 0.638041 | |
| I0623 18:33:42.489837 10365 solver.cpp:406] Test net output #50: loss2/accuracy01 = 0.971 | |
| I0623 18:33:42.489848 10365 solver.cpp:406] Test net output #51: loss2/accuracy02 = 0.948 | |
| I0623 18:33:42.489859 10365 solver.cpp:406] Test net output #52: loss2/accuracy03 = 0.866 | |
| I0623 18:33:42.489871 10365 solver.cpp:406] Test net output #53: loss2/accuracy04 = 0.746 | |
| I0623 18:33:42.489882 10365 solver.cpp:406] Test net output #54: loss2/accuracy05 = 0.585 | |
| I0623 18:33:42.489893 10365 solver.cpp:406] Test net output #55: loss2/accuracy06 = 0.539 | |
| I0623 18:33:42.489904 10365 solver.cpp:406] Test net output #56: loss2/accuracy07 = 0.534 | |
| I0623 18:33:42.489915 10365 solver.cpp:406] Test net output #57: loss2/accuracy08 = 0.549 | |
| I0623 18:33:42.489926 10365 solver.cpp:406] Test net output #58: loss2/accuracy09 = 0.486 | |
| I0623 18:33:42.489938 10365 solver.cpp:406] Test net output #59: loss2/accuracy10 = 0.437 | |
| I0623 18:33:42.489949 10365 solver.cpp:406] Test net output #60: loss2/accuracy11 = 0.454 | |
| I0623 18:33:42.489960 10365 solver.cpp:406] Test net output #61: loss2/accuracy12 = 0.539 | |
| I0623 18:33:42.489971 10365 solver.cpp:406] Test net output #62: loss2/accuracy13 = 0.608 | |
| I0623 18:33:42.489982 10365 solver.cpp:406] Test net output #63: loss2/accuracy14 = 0.708 | |
| I0623 18:33:42.489995 10365 solver.cpp:406] Test net output #64: loss2/accuracy15 = 0.791 | |
| I0623 18:33:42.490005 10365 solver.cpp:406] Test net output #65: loss2/accuracy16 = 0.85 | |
| I0623 18:33:42.490016 10365 solver.cpp:406] Test net output #66: loss2/accuracy17 = 0.908 | |
| I0623 18:33:42.490027 10365 solver.cpp:406] Test net output #67: loss2/accuracy18 = 0.953 | |
| I0623 18:33:42.490038 10365 solver.cpp:406] Test net output #68: loss2/accuracy19 = 0.972 | |
| I0623 18:33:42.490049 10365 solver.cpp:406] Test net output #69: loss2/accuracy20 = 0.987 | |
| I0623 18:33:42.490061 10365 solver.cpp:406] Test net output #70: loss2/accuracy21 = 0.999 | |
| I0623 18:33:42.490072 10365 solver.cpp:406] Test net output #71: loss2/accuracy22 = 1 | |
| I0623 18:33:42.490083 10365 solver.cpp:406] Test net output #72: loss2/accuracy_incl_empty = 0.771409 | |
| I0623 18:33:42.490094 10365 solver.cpp:406] Test net output #73: loss2/accuracy_top3 = 0.9189 | |
| I0623 18:33:42.490108 10365 solver.cpp:406] Test net output #74: loss2/cross_entropy_loss = 1.02947 (* 0.3 = 0.308842 loss) | |
| I0623 18:33:42.490121 10365 solver.cpp:406] Test net output #75: loss2/cross_entropy_loss_incl_empty = 0.644754 (* 0.3 = 0.193426 loss) | |
| I0623 18:33:42.490135 10365 solver.cpp:406] Test net output #76: loss2/loss01 = 0.19625 (* 0.0272727 = 0.00535226 loss) | |
| I0623 18:33:42.490149 10365 solver.cpp:406] Test net output #77: loss2/loss02 = 0.268124 (* 0.0272727 = 0.00731247 loss) | |
| I0623 18:33:42.490177 10365 solver.cpp:406] Test net output #78: loss2/loss03 = 0.54132 (* 0.0272727 = 0.0147633 loss) | |
| I0623 18:33:42.490193 10365 solver.cpp:406] Test net output #79: loss2/loss04 = 0.854885 (* 0.0272727 = 0.0233151 loss) | |
| I0623 18:33:42.490207 10365 solver.cpp:406] Test net output #80: loss2/loss05 = 1.12717 (* 0.0272727 = 0.030741 loss) | |
| I0623 18:33:42.490221 10365 solver.cpp:406] Test net output #81: loss2/loss06 = 1.35126 (* 0.0272727 = 0.0368524 loss) | |
| I0623 18:33:42.490234 10365 solver.cpp:406] Test net output #82: loss2/loss07 = 1.41486 (* 0.0272727 = 0.0385871 loss) | |
| I0623 18:33:42.490248 10365 solver.cpp:406] Test net output #83: loss2/loss08 = 1.37768 (* 0.0272727 = 0.0375731 loss) | |
| I0623 18:33:42.490262 10365 solver.cpp:406] Test net output #84: loss2/loss09 = 1.46953 (* 0.0272727 = 0.040078 loss) | |
| I0623 18:33:42.490277 10365 solver.cpp:406] Test net output #85: loss2/loss10 = 1.55917 (* 0.0272727 = 0.0425228 loss) | |
| I0623 18:33:42.490289 10365 solver.cpp:406] Test net output #86: loss2/loss11 = 1.59495 (* 0.0272727 = 0.0434987 loss) | |
| I0623 18:33:42.490303 10365 solver.cpp:406] Test net output #87: loss2/loss12 = 1.33204 (* 0.0272727 = 0.0363284 loss) | |
| I0623 18:33:42.490319 10365 solver.cpp:406] Test net output #88: loss2/loss13 = 1.08575 (* 0.0272727 = 0.0296115 loss) | |
| I0623 18:33:42.490332 10365 solver.cpp:406] Test net output #89: loss2/loss14 = 0.83723 (* 0.0272727 = 0.0228336 loss) | |
| I0623 18:33:42.490346 10365 solver.cpp:406] Test net output #90: loss2/loss15 = 0.60647 (* 0.0272727 = 0.0165401 loss) | |
| I0623 18:33:42.490360 10365 solver.cpp:406] Test net output #91: loss2/loss16 = 0.446274 (* 0.0272727 = 0.0121711 loss) | |
| I0623 18:33:42.490373 10365 solver.cpp:406] Test net output #92: loss2/loss17 = 0.29328 (* 0.0272727 = 0.00799854 loss) | |
| I0623 18:33:42.490387 10365 solver.cpp:406] Test net output #93: loss2/loss18 = 0.162741 (* 0.0272727 = 0.00443839 loss) | |
| I0623 18:33:42.490401 10365 solver.cpp:406] Test net output #94: loss2/loss19 = 0.104279 (* 0.0272727 = 0.00284398 loss) | |
| I0623 18:33:42.490414 10365 solver.cpp:406] Test net output #95: loss2/loss20 = 0.0613253 (* 0.0272727 = 0.00167251 loss) | |
| I0623 18:33:42.490428 10365 solver.cpp:406] Test net output #96: loss2/loss21 = 0.00781317 (* 0.0272727 = 0.000213086 loss) | |
| I0623 18:33:42.490442 10365 solver.cpp:406] Test net output #97: loss2/loss22 = 9.60817e-05 (* 0.0272727 = 2.62041e-06 loss) | |
| I0623 18:33:42.490454 10365 solver.cpp:406] Test net output #98: loss3/accuracy = 0.880357 | |
| I0623 18:33:42.490465 10365 solver.cpp:406] Test net output #99: loss3/accuracy01 = 0.978 | |
| I0623 18:33:42.490478 10365 solver.cpp:406] Test net output #100: loss3/accuracy02 = 0.978 | |
| I0623 18:33:42.490489 10365 solver.cpp:406] Test net output #101: loss3/accuracy03 = 0.958 | |
| I0623 18:33:42.490499 10365 solver.cpp:406] Test net output #102: loss3/accuracy04 = 0.948 | |
| I0623 18:33:42.490511 10365 solver.cpp:406] Test net output #103: loss3/accuracy05 = 0.94 | |
| I0623 18:33:42.490522 10365 solver.cpp:406] Test net output #104: loss3/accuracy06 = 0.916 | |
| I0623 18:33:42.490533 10365 solver.cpp:406] Test net output #105: loss3/accuracy07 = 0.911 | |
| I0623 18:33:42.490545 10365 solver.cpp:406] Test net output #106: loss3/accuracy08 = 0.881 | |
| I0623 18:33:42.490556 10365 solver.cpp:406] Test net output #107: loss3/accuracy09 = 0.823 | |
| I0623 18:33:42.490566 10365 solver.cpp:406] Test net output #108: loss3/accuracy10 = 0.742 | |
| I0623 18:33:42.490577 10365 solver.cpp:406] Test net output #109: loss3/accuracy11 = 0.648 | |
| I0623 18:33:42.490588 10365 solver.cpp:406] Test net output #110: loss3/accuracy12 = 0.672 | |
| I0623 18:33:42.490599 10365 solver.cpp:406] Test net output #111: loss3/accuracy13 = 0.706 | |
| I0623 18:33:42.490610 10365 solver.cpp:406] Test net output #112: loss3/accuracy14 = 0.769 | |
| I0623 18:33:42.490622 10365 solver.cpp:406] Test net output #113: loss3/accuracy15 = 0.837 | |
| I0623 18:33:42.490633 10365 solver.cpp:406] Test net output #114: loss3/accuracy16 = 0.882 | |
| I0623 18:33:42.490654 10365 solver.cpp:406] Test net output #115: loss3/accuracy17 = 0.934 | |
| I0623 18:33:42.490667 10365 solver.cpp:406] Test net output #116: loss3/accuracy18 = 0.965 | |
| I0623 18:33:42.490679 10365 solver.cpp:406] Test net output #117: loss3/accuracy19 = 0.977 | |
| I0623 18:33:42.490690 10365 solver.cpp:406] Test net output #118: loss3/accuracy20 = 0.987 | |
| I0623 18:33:42.490701 10365 solver.cpp:406] Test net output #119: loss3/accuracy21 = 0.999 | |
| I0623 18:33:42.490712 10365 solver.cpp:406] Test net output #120: loss3/accuracy22 = 1 | |
| I0623 18:33:42.490723 10365 solver.cpp:406] Test net output #121: loss3/accuracy_incl_empty = 0.918637 | |
| I0623 18:33:42.490734 10365 solver.cpp:406] Test net output #122: loss3/accuracy_top3 = 0.972377 | |
| I0623 18:33:42.490747 10365 solver.cpp:406] Test net output #123: loss3/cross_entropy_loss = 0.492678 (* 1 = 0.492678 loss) | |
| I0623 18:33:42.490761 10365 solver.cpp:406] Test net output #124: loss3/cross_entropy_loss_incl_empty = 0.320652 (* 1 = 0.320652 loss) | |
| I0623 18:33:42.490774 10365 solver.cpp:406] Test net output #125: loss3/loss01 = 0.149093 (* 0.0909091 = 0.013554 loss) | |
| I0623 18:33:42.490788 10365 solver.cpp:406] Test net output #126: loss3/loss02 = 0.167129 (* 0.0909091 = 0.0151935 loss) | |
| I0623 18:33:42.490802 10365 solver.cpp:406] Test net output #127: loss3/loss03 = 0.277119 (* 0.0909091 = 0.0251927 loss) | |
| I0623 18:33:42.490815 10365 solver.cpp:406] Test net output #128: loss3/loss04 = 0.327528 (* 0.0909091 = 0.0297752 loss) | |
| I0623 18:33:42.490828 10365 solver.cpp:406] Test net output #129: loss3/loss05 = 0.347202 (* 0.0909091 = 0.0315638 loss) | |
| I0623 18:33:42.490841 10365 solver.cpp:406] Test net output #130: loss3/loss06 = 0.439179 (* 0.0909091 = 0.0399254 loss) | |
| I0623 18:33:42.490855 10365 solver.cpp:406] Test net output #131: loss3/loss07 = 0.479171 (* 0.0909091 = 0.043561 loss) | |
| I0623 18:33:42.490869 10365 solver.cpp:406] Test net output #132: loss3/loss08 = 0.51575 (* 0.0909091 = 0.0468864 loss) | |
| I0623 18:33:42.490881 10365 solver.cpp:406] Test net output #133: loss3/loss09 = 0.644845 (* 0.0909091 = 0.0586223 loss) | |
| I0623 18:33:42.490895 10365 solver.cpp:406] Test net output #134: loss3/loss10 = 0.811556 (* 0.0909091 = 0.0737778 loss) | |
| I0623 18:33:42.490908 10365 solver.cpp:406] Test net output #135: loss3/loss11 = 0.982552 (* 0.0909091 = 0.0893229 loss) | |
| I0623 18:33:42.490921 10365 solver.cpp:406] Test net output #136: loss3/loss12 = 0.874559 (* 0.0909091 = 0.0795054 loss) | |
| I0623 18:33:42.490936 10365 solver.cpp:406] Test net output #137: loss3/loss13 = 0.793505 (* 0.0909091 = 0.0721369 loss) | |
| I0623 18:33:42.490948 10365 solver.cpp:406] Test net output #138: loss3/loss14 = 0.611483 (* 0.0909091 = 0.0555894 loss) | |
| I0623 18:33:42.490962 10365 solver.cpp:406] Test net output #139: loss3/loss15 = 0.467804 (* 0.0909091 = 0.0425277 loss) | |
| I0623 18:33:42.490974 10365 solver.cpp:406] Test net output #140: loss3/loss16 = 0.35022 (* 0.0909091 = 0.0318382 loss) | |
| I0623 18:33:42.490988 10365 solver.cpp:406] Test net output #141: loss3/loss17 = 0.206004 (* 0.0909091 = 0.0187277 loss) | |
| I0623 18:33:42.491001 10365 solver.cpp:406] Test net output #142: loss3/loss18 = 0.120077 (* 0.0909091 = 0.0109161 loss) | |
| I0623 18:33:42.491015 10365 solver.cpp:406] Test net output #143: loss3/loss19 = 0.0700516 (* 0.0909091 = 0.00636833 loss) | |
| I0623 18:33:42.491029 10365 solver.cpp:406] Test net output #144: loss3/loss20 = 0.0408868 (* 0.0909091 = 0.00371699 loss) | |
| I0623 18:33:42.491042 10365 solver.cpp:406] Test net output #145: loss3/loss21 = 0.00614451 (* 0.0909091 = 0.000558592 loss) | |
| I0623 18:33:42.491055 10365 solver.cpp:406] Test net output #146: loss3/loss22 = 9.89341e-05 (* 0.0909091 = 8.99401e-06 loss) | |
| I0623 18:33:42.491067 10365 solver.cpp:406] Test net output #147: total_accuracy = 0.444 | |
| I0623 18:33:42.491078 10365 solver.cpp:406] Test net output #148: total_accuracy_not_rec = 0.242 | |
| I0623 18:33:42.491089 10365 solver.cpp:406] Test net output #149: total_confidence = 0.235038 | |
| I0623 18:33:42.491109 10365 solver.cpp:406] Test net output #150: total_confidence_not_rec = 0.145771 | |
| I0623 18:33:42.849488 10365 solver.cpp:229] Iteration 20000, loss = 4.45801 | |
| I0623 18:33:42.849548 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.407767 | |
| I0623 18:33:42.849565 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.875 | |
| I0623 18:33:42.849580 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 18:33:42.849592 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 18:33:42.849606 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.75 | |
| I0623 18:33:42.849618 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 18:33:42.849630 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 18:33:42.849642 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 18:33:42.849654 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 18:33:42.849666 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 18:33:42.849679 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 18:33:42.849691 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.125 | |
| I0623 18:33:42.849704 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 18:33:42.849715 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 18:33:42.849726 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 18:33:42.849738 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 18:33:42.849750 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 18:33:42.849761 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 18:33:42.849771 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 18:33:42.849783 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:33:42.849795 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:33:42.849807 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:33:42.849818 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:33:42.849829 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.647727 | |
| I0623 18:33:42.849841 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.699029 | |
| I0623 18:33:42.849858 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.71817 (* 0.3 = 0.515452 loss) | |
| I0623 18:33:42.849871 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.02843 (* 0.3 = 0.308529 loss) | |
| I0623 18:33:42.849886 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.439219 (* 0.0272727 = 0.0119787 loss) | |
| I0623 18:33:42.849900 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.63014 (* 0.0272727 = 0.0444582 loss) | |
| I0623 18:33:42.849913 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.62636 (* 0.0272727 = 0.0443553 loss) | |
| I0623 18:33:42.849927 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.69922 (* 0.0272727 = 0.0463424 loss) | |
| I0623 18:33:42.849941 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.06833 (* 0.0272727 = 0.0564091 loss) | |
| I0623 18:33:42.849956 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.12039 (* 0.0272727 = 0.0578289 loss) | |
| I0623 18:33:42.849968 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.31819 (* 0.0272727 = 0.0632233 loss) | |
| I0623 18:33:42.849982 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.21549 (* 0.0272727 = 0.0604225 loss) | |
| I0623 18:33:42.849997 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.65716 (* 0.0272727 = 0.0451953 loss) | |
| I0623 18:33:42.850010 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.13015 (* 0.0272727 = 0.058095 loss) | |
| I0623 18:33:42.850023 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.5891 (* 0.0272727 = 0.0706118 loss) | |
| I0623 18:33:42.850062 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.3594 (* 0.0272727 = 0.0643472 loss) | |
| I0623 18:33:42.850080 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.44259 (* 0.0272727 = 0.0393435 loss) | |
| I0623 18:33:42.850095 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.17335 (* 0.0272727 = 0.0320004 loss) | |
| I0623 18:33:42.850109 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.632072 (* 0.0272727 = 0.0172383 loss) | |
| I0623 18:33:42.850123 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.387776 (* 0.0272727 = 0.0105757 loss) | |
| I0623 18:33:42.850137 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.428292 (* 0.0272727 = 0.0116807 loss) | |
| I0623 18:33:42.850152 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.225877 (* 0.0272727 = 0.00616027 loss) | |
| I0623 18:33:42.850165 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0569277 (* 0.0272727 = 0.00155257 loss) | |
| I0623 18:33:42.850179 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00928081 (* 0.0272727 = 0.000253113 loss) | |
| I0623 18:33:42.850194 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000628595 (* 0.0272727 = 1.71435e-05 loss) | |
| I0623 18:33:42.850208 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 4.62299e-05 (* 0.0272727 = 1.26082e-06 loss) | |
| I0623 18:33:42.850220 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.504854 | |
| I0623 18:33:42.850232 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 18:33:42.850244 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 18:33:42.850256 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 18:33:42.850267 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 18:33:42.850280 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 18:33:42.850291 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 18:33:42.850302 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 18:33:42.850313 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.375 | |
| I0623 18:33:42.850325 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 18:33:42.850337 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 18:33:42.850348 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 18:33:42.850359 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.375 | |
| I0623 18:33:42.850370 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 18:33:42.850381 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 18:33:42.850394 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 18:33:42.850405 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 18:33:42.850417 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 18:33:42.850428 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 18:33:42.850440 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:33:42.850451 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:33:42.850463 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:33:42.850474 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:33:42.850491 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.693182 | |
| I0623 18:33:42.850502 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.805825 | |
| I0623 18:33:42.850517 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.45114 (* 0.3 = 0.435343 loss) | |
| I0623 18:33:42.850530 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.895353 (* 0.3 = 0.268606 loss) | |
| I0623 18:33:42.850555 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.679559 (* 0.0272727 = 0.0185334 loss) | |
| I0623 18:33:42.850570 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 1.02676 (* 0.0272727 = 0.0280026 loss) | |
| I0623 18:33:42.850584 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.211755 (* 0.0272727 = 0.00577514 loss) | |
| I0623 18:33:42.850600 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.99202 (* 0.0272727 = 0.0543277 loss) | |
| I0623 18:33:42.850613 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.41297 (* 0.0272727 = 0.0385356 loss) | |
| I0623 18:33:42.850626 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.57525 (* 0.0272727 = 0.0429613 loss) | |
| I0623 18:33:42.850641 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 2.04138 (* 0.0272727 = 0.0556739 loss) | |
| I0623 18:33:42.850653 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.81936 (* 0.0272727 = 0.049619 loss) | |
| I0623 18:33:42.850667 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 2.03306 (* 0.0272727 = 0.0554472 loss) | |
| I0623 18:33:42.850682 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.97962 (* 0.0272727 = 0.0539896 loss) | |
| I0623 18:33:42.850694 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.85288 (* 0.0272727 = 0.0505331 loss) | |
| I0623 18:33:42.850708 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.64006 (* 0.0272727 = 0.044729 loss) | |
| I0623 18:33:42.850721 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.33047 (* 0.0272727 = 0.0362854 loss) | |
| I0623 18:33:42.850735 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.15627 (* 0.0272727 = 0.0315345 loss) | |
| I0623 18:33:42.850749 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.659573 (* 0.0272727 = 0.0179884 loss) | |
| I0623 18:33:42.850764 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.222237 (* 0.0272727 = 0.006061 loss) | |
| I0623 18:33:42.850777 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.382793 (* 0.0272727 = 0.0104398 loss) | |
| I0623 18:33:42.850790 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.42583 (* 0.0272727 = 0.0116135 loss) | |
| I0623 18:33:42.850805 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0226779 (* 0.0272727 = 0.000618489 loss) | |
| I0623 18:33:42.850819 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.0018586 (* 0.0272727 = 5.06891e-05 loss) | |
| I0623 18:33:42.850833 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000177707 (* 0.0272727 = 4.84656e-06 loss) | |
| I0623 18:33:42.850847 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.85228e-05 (* 0.0272727 = 5.05166e-07 loss) | |
| I0623 18:33:42.850859 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.68932 | |
| I0623 18:33:42.850872 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.875 | |
| I0623 18:33:42.850883 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 18:33:42.850895 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 18:33:42.850906 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.625 | |
| I0623 18:33:42.850919 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 18:33:42.850930 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.75 | |
| I0623 18:33:42.850941 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 18:33:42.850953 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 18:33:42.850965 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.5 | |
| I0623 18:33:42.850976 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 18:33:42.850987 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.25 | |
| I0623 18:33:42.850999 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.375 | |
| I0623 18:33:42.851011 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 18:33:42.851032 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 18:33:42.851044 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 18:33:42.851057 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 18:33:42.851068 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:33:42.851079 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 0.875 | |
| I0623 18:33:42.851091 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:33:42.851102 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:33:42.851114 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:33:42.851125 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:33:42.851140 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.8125 | |
| I0623 18:33:42.851152 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.883495 | |
| I0623 18:33:42.851166 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 1.21909 (* 1 = 1.21909 loss) | |
| I0623 18:33:42.851181 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.723777 (* 1 = 0.723777 loss) | |
| I0623 18:33:42.851194 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.489799 (* 0.0909091 = 0.0445272 loss) | |
| I0623 18:33:42.851208 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.462655 (* 0.0909091 = 0.0420596 loss) | |
| I0623 18:33:42.851222 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.474679 (* 0.0909091 = 0.0431526 loss) | |
| I0623 18:33:42.851235 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 1.81521 (* 0.0909091 = 0.165019 loss) | |
| I0623 18:33:42.851249 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.940518 (* 0.0909091 = 0.0855017 loss) | |
| I0623 18:33:42.851263 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 1.36871 (* 0.0909091 = 0.124428 loss) | |
| I0623 18:33:42.851277 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 1.25111 (* 0.0909091 = 0.113737 loss) | |
| I0623 18:33:42.851290 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.950444 (* 0.0909091 = 0.086404 loss) | |
| I0623 18:33:42.851305 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.58747 (* 0.0909091 = 0.144315 loss) | |
| I0623 18:33:42.851318 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.2655 (* 0.0909091 = 0.115045 loss) | |
| I0623 18:33:42.851331 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.76707 (* 0.0909091 = 0.160643 loss) | |
| I0623 18:33:42.851346 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 2.06346 (* 0.0909091 = 0.187588 loss) | |
| I0623 18:33:42.851358 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.838388 (* 0.0909091 = 0.0762171 loss) | |
| I0623 18:33:42.851372 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.488011 (* 0.0909091 = 0.0443647 loss) | |
| I0623 18:33:42.851385 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.774001 (* 0.0909091 = 0.0703637 loss) | |
| I0623 18:33:42.851399 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.166547 (* 0.0909091 = 0.0151406 loss) | |
| I0623 18:33:42.851413 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.13781 (* 0.0909091 = 0.0125282 loss) | |
| I0623 18:33:42.851426 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.507906 (* 0.0909091 = 0.0461733 loss) | |
| I0623 18:33:42.851440 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.00722897 (* 0.0909091 = 0.000657179 loss) | |
| I0623 18:33:42.851454 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000507185 (* 0.0909091 = 4.61077e-05 loss) | |
| I0623 18:33:42.851469 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 6.69967e-05 (* 0.0909091 = 6.09061e-06 loss) | |
| I0623 18:33:42.851482 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 2.47361e-06 (* 0.0909091 = 2.24874e-07 loss) | |
| I0623 18:33:42.851503 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 18:33:42.851516 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.25 | |
| I0623 18:33:42.851532 10365 solver.cpp:245] Train net output #149: total_confidence = 0.124154 | |
| I0623 18:33:42.851546 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.105737 | |
| I0623 18:33:42.851559 10365 sgd_solver.cpp:106] Iteration 20000, lr = 0.001 | |
| I0623 18:35:12.915496 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.0829 > 30) by scale factor 0.997245 | |
| I0623 18:36:04.232640 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.4525 > 30) by scale factor 0.706671 | |
| I0623 18:37:32.313570 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.2527 > 30) by scale factor 0.930156 | |
| I0623 18:38:45.077514 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.7755 > 30) by scale factor 0.974801 | |
| I0623 18:40:05.955054 10365 solver.cpp:229] Iteration 20500, loss = 4.34772 | |
| I0623 18:40:05.955139 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.505747 | |
| I0623 18:40:05.955157 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.625 | |
| I0623 18:40:05.955170 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.375 | |
| I0623 18:40:05.955183 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.25 | |
| I0623 18:40:05.955195 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 18:40:05.955209 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 18:40:05.955221 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 18:40:05.955235 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 18:40:05.955246 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 18:40:05.955258 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.75 | |
| I0623 18:40:05.955271 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.5 | |
| I0623 18:40:05.955283 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.625 | |
| I0623 18:40:05.955296 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.5 | |
| I0623 18:40:05.955307 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 18:40:05.955318 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 18:40:05.955330 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.875 | |
| I0623 18:40:05.955343 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 18:40:05.955353 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 18:40:05.955365 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:40:05.955377 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:40:05.955389 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:40:05.955400 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:40:05.955412 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:40:05.955423 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.744318 | |
| I0623 18:40:05.955435 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.793103 | |
| I0623 18:40:05.955451 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.59723 (* 0.3 = 0.47917 loss) | |
| I0623 18:40:05.955466 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.809651 (* 0.3 = 0.242895 loss) | |
| I0623 18:40:05.955481 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 1.5724 (* 0.0272727 = 0.0428837 loss) | |
| I0623 18:40:05.955493 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.10122 (* 0.0272727 = 0.0300333 loss) | |
| I0623 18:40:05.955507 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.68326 (* 0.0272727 = 0.0459071 loss) | |
| I0623 18:40:05.955521 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.2731 (* 0.0272727 = 0.034721 loss) | |
| I0623 18:40:05.955535 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.77366 (* 0.0272727 = 0.0483725 loss) | |
| I0623 18:40:05.955549 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.93828 (* 0.0272727 = 0.0528622 loss) | |
| I0623 18:40:05.955562 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.24603 (* 0.0272727 = 0.0339826 loss) | |
| I0623 18:40:05.955576 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.24144 (* 0.0272727 = 0.0338576 loss) | |
| I0623 18:40:05.955590 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 0.676724 (* 0.0272727 = 0.0184561 loss) | |
| I0623 18:40:05.955623 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.74247 (* 0.0272727 = 0.0475218 loss) | |
| I0623 18:40:05.955638 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.3072 (* 0.0272727 = 0.0356508 loss) | |
| I0623 18:40:05.955652 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.27399 (* 0.0272727 = 0.0347451 loss) | |
| I0623 18:40:05.955685 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.36896 (* 0.0272727 = 0.0373353 loss) | |
| I0623 18:40:05.955700 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.34129 (* 0.0272727 = 0.0365807 loss) | |
| I0623 18:40:05.955713 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.477847 (* 0.0272727 = 0.0130322 loss) | |
| I0623 18:40:05.955727 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.512905 (* 0.0272727 = 0.0139883 loss) | |
| I0623 18:40:05.955742 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.00297387 (* 0.0272727 = 8.11056e-05 loss) | |
| I0623 18:40:05.955756 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.000134223 (* 0.0272727 = 3.66062e-06 loss) | |
| I0623 18:40:05.955770 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 9.67097e-06 (* 0.0272727 = 2.63754e-07 loss) | |
| I0623 18:40:05.955785 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 1.49012e-06 (* 0.0272727 = 4.06396e-08 loss) | |
| I0623 18:40:05.955799 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 4.32134e-07 (* 0.0272727 = 1.17855e-08 loss) | |
| I0623 18:40:05.955813 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 2.98023e-08 (* 0.0272727 = 8.12791e-10 loss) | |
| I0623 18:40:05.955826 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.666667 | |
| I0623 18:40:05.955837 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.75 | |
| I0623 18:40:05.955849 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 18:40:05.955862 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.75 | |
| I0623 18:40:05.955873 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 18:40:05.955884 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.875 | |
| I0623 18:40:05.955895 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 18:40:05.955907 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.625 | |
| I0623 18:40:05.955919 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.5 | |
| I0623 18:40:05.955931 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.875 | |
| I0623 18:40:05.955942 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 18:40:05.955955 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 18:40:05.955965 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 18:40:05.955977 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 18:40:05.955988 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 18:40:05.956001 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 18:40:05.956012 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 18:40:05.956022 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 18:40:05.956033 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:40:05.956045 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:40:05.956056 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:40:05.956068 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:40:05.956079 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:40:05.956090 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.818182 | |
| I0623 18:40:05.956102 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.908046 | |
| I0623 18:40:05.956116 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.04185 (* 0.3 = 0.312555 loss) | |
| I0623 18:40:05.956130 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.559516 (* 0.3 = 0.167855 loss) | |
| I0623 18:40:05.956143 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 1.03171 (* 0.0272727 = 0.0281374 loss) | |
| I0623 18:40:05.956162 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.475708 (* 0.0272727 = 0.0129738 loss) | |
| I0623 18:40:05.956187 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.648719 (* 0.0272727 = 0.0176923 loss) | |
| I0623 18:40:05.956202 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.21668 (* 0.0272727 = 0.0331822 loss) | |
| I0623 18:40:05.956217 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 0.810416 (* 0.0272727 = 0.0221023 loss) | |
| I0623 18:40:05.956229 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.49563 (* 0.0272727 = 0.04079 loss) | |
| I0623 18:40:05.956243 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 0.908799 (* 0.0272727 = 0.0247854 loss) | |
| I0623 18:40:05.956257 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.17785 (* 0.0272727 = 0.0321231 loss) | |
| I0623 18:40:05.956270 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 0.578638 (* 0.0272727 = 0.015781 loss) | |
| I0623 18:40:05.956284 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.54733 (* 0.0272727 = 0.0422 loss) | |
| I0623 18:40:05.956298 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 0.854463 (* 0.0272727 = 0.0233035 loss) | |
| I0623 18:40:05.956312 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.14849 (* 0.0272727 = 0.0313223 loss) | |
| I0623 18:40:05.956326 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.14533 (* 0.0272727 = 0.0312364 loss) | |
| I0623 18:40:05.956339 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.01977 (* 0.0272727 = 0.0278119 loss) | |
| I0623 18:40:05.956353 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.742505 (* 0.0272727 = 0.0202501 loss) | |
| I0623 18:40:05.956367 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.615304 (* 0.0272727 = 0.016781 loss) | |
| I0623 18:40:05.956380 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0469668 (* 0.0272727 = 0.00128091 loss) | |
| I0623 18:40:05.956394 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0071033 (* 0.0272727 = 0.000193726 loss) | |
| I0623 18:40:05.956408 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00130424 (* 0.0272727 = 3.55701e-05 loss) | |
| I0623 18:40:05.956423 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00030964 (* 0.0272727 = 8.44474e-06 loss) | |
| I0623 18:40:05.956436 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000214458 (* 0.0272727 = 5.84886e-06 loss) | |
| I0623 18:40:05.956450 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 3.01106e-05 (* 0.0272727 = 8.21198e-07 loss) | |
| I0623 18:40:05.956462 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.908046 | |
| I0623 18:40:05.956475 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.875 | |
| I0623 18:40:05.956486 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 18:40:05.956497 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 18:40:05.956509 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 18:40:05.956521 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 18:40:05.956532 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 18:40:05.956543 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 18:40:05.956555 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 18:40:05.956567 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.75 | |
| I0623 18:40:05.956578 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 18:40:05.956589 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 18:40:05.956601 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.75 | |
| I0623 18:40:05.956612 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 18:40:05.956624 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 18:40:05.956635 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 18:40:05.956646 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 18:40:05.956668 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:40:05.956681 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:40:05.956692 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:40:05.956704 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:40:05.956715 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:40:05.956727 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:40:05.956737 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.954545 | |
| I0623 18:40:05.956749 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.977012 | |
| I0623 18:40:05.956763 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.424117 (* 1 = 0.424117 loss) | |
| I0623 18:40:05.956778 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.214997 (* 1 = 0.214997 loss) | |
| I0623 18:40:05.956792 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.968974 (* 0.0909091 = 0.0880886 loss) | |
| I0623 18:40:05.956806 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0952706 (* 0.0909091 = 0.00866096 loss) | |
| I0623 18:40:05.956820 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.540184 (* 0.0909091 = 0.0491077 loss) | |
| I0623 18:40:05.956835 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.126084 (* 0.0909091 = 0.0114622 loss) | |
| I0623 18:40:05.956847 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.103917 (* 0.0909091 = 0.00944704 loss) | |
| I0623 18:40:05.956861 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.403409 (* 0.0909091 = 0.0366735 loss) | |
| I0623 18:40:05.956876 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.0749303 (* 0.0909091 = 0.00681185 loss) | |
| I0623 18:40:05.956889 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.539251 (* 0.0909091 = 0.0490228 loss) | |
| I0623 18:40:05.956902 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.352132 (* 0.0909091 = 0.032012 loss) | |
| I0623 18:40:05.956917 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.466425 (* 0.0909091 = 0.0424023 loss) | |
| I0623 18:40:05.956929 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.747306 (* 0.0909091 = 0.0679369 loss) | |
| I0623 18:40:05.956943 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.650043 (* 0.0909091 = 0.0590949 loss) | |
| I0623 18:40:05.956957 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.619518 (* 0.0909091 = 0.0563198 loss) | |
| I0623 18:40:05.956970 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.587671 (* 0.0909091 = 0.0534247 loss) | |
| I0623 18:40:05.956984 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.448063 (* 0.0909091 = 0.040733 loss) | |
| I0623 18:40:05.956997 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.04482 (* 0.0909091 = 0.00407454 loss) | |
| I0623 18:40:05.957010 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0130088 (* 0.0909091 = 0.00118262 loss) | |
| I0623 18:40:05.957025 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00111991 (* 0.0909091 = 0.00010181 loss) | |
| I0623 18:40:05.957038 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000118997 (* 0.0909091 = 1.08179e-05 loss) | |
| I0623 18:40:05.957052 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 3.76876e-05 (* 0.0909091 = 3.42614e-06 loss) | |
| I0623 18:40:05.957065 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 5.91817e-05 (* 0.0909091 = 5.38016e-06 loss) | |
| I0623 18:40:05.957079 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.21707e-06 (* 0.0909091 = 3.8337e-07 loss) | |
| I0623 18:40:05.957092 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 18:40:05.957103 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.25 | |
| I0623 18:40:05.957124 10365 solver.cpp:245] Train net output #149: total_confidence = 0.326521 | |
| I0623 18:40:05.957137 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.303762 | |
| I0623 18:40:05.957150 10365 sgd_solver.cpp:106] Iteration 20500, lr = 0.001 | |
| I0623 18:41:44.376720 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 84.0854 > 30) by scale factor 0.35678 | |
| I0623 18:44:37.604583 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.5507 > 30) by scale factor 0.739814 | |
| I0623 18:45:15.137598 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.1349 > 30) by scale factor 0.963549 | |
| I0623 18:46:20.244386 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.8568 > 30) by scale factor 0.913054 | |
| I0623 18:46:29.080834 10365 solver.cpp:229] Iteration 21000, loss = 4.35115 | |
| I0623 18:46:29.080929 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.419355 | |
| I0623 18:46:29.080948 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 18:46:29.080962 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 18:46:29.080976 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 18:46:29.080989 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.375 | |
| I0623 18:46:29.081007 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 18:46:29.081022 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 18:46:29.081034 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 18:46:29.081046 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 18:46:29.081059 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.25 | |
| I0623 18:46:29.081073 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 18:46:29.081084 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 18:46:29.081097 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.5 | |
| I0623 18:46:29.081110 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 18:46:29.081122 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.875 | |
| I0623 18:46:29.081135 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 18:46:29.081146 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 18:46:29.081161 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 18:46:29.081173 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:46:29.081185 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:46:29.081197 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:46:29.081209 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:46:29.081221 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:46:29.081233 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.653409 | |
| I0623 18:46:29.081245 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.827957 | |
| I0623 18:46:29.081265 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.62704 (* 0.3 = 0.488113 loss) | |
| I0623 18:46:29.081280 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.00935 (* 0.3 = 0.302806 loss) | |
| I0623 18:46:29.081295 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.900881 (* 0.0272727 = 0.0245695 loss) | |
| I0623 18:46:29.081310 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.988716 (* 0.0272727 = 0.026965 loss) | |
| I0623 18:46:29.081323 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 2.27145 (* 0.0272727 = 0.0619485 loss) | |
| I0623 18:46:29.081337 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.81759 (* 0.0272727 = 0.0495706 loss) | |
| I0623 18:46:29.081352 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.75978 (* 0.0272727 = 0.0479939 loss) | |
| I0623 18:46:29.081367 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.48915 (* 0.0272727 = 0.0406132 loss) | |
| I0623 18:46:29.081380 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.65661 (* 0.0272727 = 0.0451804 loss) | |
| I0623 18:46:29.081394 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 2.12187 (* 0.0272727 = 0.0578693 loss) | |
| I0623 18:46:29.081408 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.07006 (* 0.0272727 = 0.0564561 loss) | |
| I0623 18:46:29.081424 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.83674 (* 0.0272727 = 0.0500929 loss) | |
| I0623 18:46:29.081439 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.40835 (* 0.0272727 = 0.0384096 loss) | |
| I0623 18:46:29.081452 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.27651 (* 0.0272727 = 0.0348139 loss) | |
| I0623 18:46:29.081511 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.65403 (* 0.0272727 = 0.0451099 loss) | |
| I0623 18:46:29.081527 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.878779 (* 0.0272727 = 0.0239667 loss) | |
| I0623 18:46:29.081542 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.598133 (* 0.0272727 = 0.0163127 loss) | |
| I0623 18:46:29.081557 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.393832 (* 0.0272727 = 0.0107409 loss) | |
| I0623 18:46:29.081571 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0595389 (* 0.0272727 = 0.00162379 loss) | |
| I0623 18:46:29.081585 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.014832 (* 0.0272727 = 0.00040451 loss) | |
| I0623 18:46:29.081600 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00259167 (* 0.0272727 = 7.06818e-05 loss) | |
| I0623 18:46:29.081614 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00123616 (* 0.0272727 = 3.37135e-05 loss) | |
| I0623 18:46:29.081629 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.00066412 (* 0.0272727 = 1.81124e-05 loss) | |
| I0623 18:46:29.081643 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000107422 (* 0.0272727 = 2.9297e-06 loss) | |
| I0623 18:46:29.081656 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.569892 | |
| I0623 18:46:29.081670 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 18:46:29.081681 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 18:46:29.081693 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.5 | |
| I0623 18:46:29.081707 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.5 | |
| I0623 18:46:29.081718 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 18:46:29.081730 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.5 | |
| I0623 18:46:29.081743 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 18:46:29.081754 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.5 | |
| I0623 18:46:29.081766 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.625 | |
| I0623 18:46:29.081779 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 18:46:29.081790 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 18:46:29.081804 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.625 | |
| I0623 18:46:29.081815 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 18:46:29.081826 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 18:46:29.081838 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 18:46:29.081851 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 18:46:29.081862 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 18:46:29.081874 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:46:29.081887 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:46:29.081898 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:46:29.081910 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:46:29.081923 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:46:29.081934 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.767045 | |
| I0623 18:46:29.081946 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.88172 | |
| I0623 18:46:29.081961 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.24459 (* 0.3 = 0.373376 loss) | |
| I0623 18:46:29.081975 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.700821 (* 0.3 = 0.210246 loss) | |
| I0623 18:46:29.081990 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.378138 (* 0.0272727 = 0.0103129 loss) | |
| I0623 18:46:29.082016 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.336842 (* 0.0272727 = 0.00918661 loss) | |
| I0623 18:46:29.082033 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.87718 (* 0.0272727 = 0.0239231 loss) | |
| I0623 18:46:29.082052 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.47582 (* 0.0272727 = 0.0402497 loss) | |
| I0623 18:46:29.082067 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.278 (* 0.0272727 = 0.0348546 loss) | |
| I0623 18:46:29.082082 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.97315 (* 0.0272727 = 0.0538133 loss) | |
| I0623 18:46:29.082096 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.17821 (* 0.0272727 = 0.0321329 loss) | |
| I0623 18:46:29.082110 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.76387 (* 0.0272727 = 0.0481054 loss) | |
| I0623 18:46:29.082124 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.54568 (* 0.0272727 = 0.0421548 loss) | |
| I0623 18:46:29.082139 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.17737 (* 0.0272727 = 0.0593828 loss) | |
| I0623 18:46:29.082152 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.69711 (* 0.0272727 = 0.0462849 loss) | |
| I0623 18:46:29.082166 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.60439 (* 0.0272727 = 0.0437562 loss) | |
| I0623 18:46:29.082180 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.36812 (* 0.0272727 = 0.0373125 loss) | |
| I0623 18:46:29.082195 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.523086 (* 0.0272727 = 0.014266 loss) | |
| I0623 18:46:29.082211 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.351122 (* 0.0272727 = 0.00957607 loss) | |
| I0623 18:46:29.082226 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.467681 (* 0.0272727 = 0.0127549 loss) | |
| I0623 18:46:29.082242 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.000963693 (* 0.0272727 = 2.62825e-05 loss) | |
| I0623 18:46:29.082255 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 2.37983e-05 (* 0.0272727 = 6.49044e-07 loss) | |
| I0623 18:46:29.082270 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 2.86105e-06 (* 0.0272727 = 7.80285e-08 loss) | |
| I0623 18:46:29.082284 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 1.80305e-06 (* 0.0272727 = 4.91741e-08 loss) | |
| I0623 18:46:29.082298 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 1.01328e-06 (* 0.0272727 = 2.7635e-08 loss) | |
| I0623 18:46:29.082314 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.35601e-06 (* 0.0272727 = 3.69822e-08 loss) | |
| I0623 18:46:29.082325 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.817204 | |
| I0623 18:46:29.082339 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 18:46:29.082350 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 18:46:29.082362 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 18:46:29.082375 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 18:46:29.082386 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 18:46:29.082399 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 18:46:29.082412 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 18:46:29.082423 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.625 | |
| I0623 18:46:29.082435 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.5 | |
| I0623 18:46:29.082448 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 18:46:29.082459 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 18:46:29.082471 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 18:46:29.082484 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 18:46:29.082495 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 18:46:29.082507 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 18:46:29.082530 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 1 | |
| I0623 18:46:29.082543 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:46:29.082556 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:46:29.082567 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:46:29.082579 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:46:29.082592 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:46:29.082602 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:46:29.082614 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.886364 | |
| I0623 18:46:29.082626 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.946237 | |
| I0623 18:46:29.082641 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.627185 (* 1 = 0.627185 loss) | |
| I0623 18:46:29.082655 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.378526 (* 1 = 0.378526 loss) | |
| I0623 18:46:29.082669 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.103517 (* 0.0909091 = 0.00941066 loss) | |
| I0623 18:46:29.082684 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0578726 (* 0.0909091 = 0.00526114 loss) | |
| I0623 18:46:29.082698 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.101121 (* 0.0909091 = 0.00919278 loss) | |
| I0623 18:46:29.082713 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.189669 (* 0.0909091 = 0.0172426 loss) | |
| I0623 18:46:29.082727 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.598522 (* 0.0909091 = 0.054411 loss) | |
| I0623 18:46:29.082741 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.999934 (* 0.0909091 = 0.0909031 loss) | |
| I0623 18:46:29.082756 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.400318 (* 0.0909091 = 0.0363926 loss) | |
| I0623 18:46:29.082769 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.810528 (* 0.0909091 = 0.0736844 loss) | |
| I0623 18:46:29.082783 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 1.17409 (* 0.0909091 = 0.106735 loss) | |
| I0623 18:46:29.082798 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.994666 (* 0.0909091 = 0.0904242 loss) | |
| I0623 18:46:29.082811 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.62533 (* 0.0909091 = 0.147758 loss) | |
| I0623 18:46:29.082825 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.14226 (* 0.0909091 = 0.103842 loss) | |
| I0623 18:46:29.082839 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.35097 (* 0.0909091 = 0.122816 loss) | |
| I0623 18:46:29.082852 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.485667 (* 0.0909091 = 0.0441515 loss) | |
| I0623 18:46:29.082867 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.202726 (* 0.0909091 = 0.0184297 loss) | |
| I0623 18:46:29.082881 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.0893894 (* 0.0909091 = 0.00812631 loss) | |
| I0623 18:46:29.082896 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.0018163 (* 0.0909091 = 0.000165118 loss) | |
| I0623 18:46:29.082911 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000217696 (* 0.0909091 = 1.97906e-05 loss) | |
| I0623 18:46:29.082926 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 5.88131e-05 (* 0.0909091 = 5.34664e-06 loss) | |
| I0623 18:46:29.082952 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 1.46034e-05 (* 0.0909091 = 1.32758e-06 loss) | |
| I0623 18:46:29.082974 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 1.01925e-05 (* 0.0909091 = 9.26592e-07 loss) | |
| I0623 18:46:29.082998 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 1.99676e-06 (* 0.0909091 = 1.81524e-07 loss) | |
| I0623 18:46:29.083019 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 18:46:29.083039 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.125 | |
| I0623 18:46:29.083073 10365 solver.cpp:245] Train net output #149: total_confidence = 0.170768 | |
| I0623 18:46:29.083094 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.084089 | |
| I0623 18:46:29.083124 10365 sgd_solver.cpp:106] Iteration 21000, lr = 0.001 | |
| I0623 18:50:36.154938 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.9349 > 30) by scale factor 0.732871 | |
| I0623 18:52:01.931349 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.1235 > 30) by scale factor 0.933895 | |
| I0623 18:52:52.137483 10365 solver.cpp:229] Iteration 21500, loss = 4.37579 | |
| I0623 18:52:52.137574 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.468468 | |
| I0623 18:52:52.137593 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 18:52:52.137606 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.875 | |
| I0623 18:52:52.137619 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 18:52:52.137632 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 18:52:52.137645 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 18:52:52.137657 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 18:52:52.137670 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.375 | |
| I0623 18:52:52.137682 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 18:52:52.137694 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 18:52:52.137707 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 18:52:52.137719 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.125 | |
| I0623 18:52:52.137732 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 18:52:52.137744 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 18:52:52.137755 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.375 | |
| I0623 18:52:52.137768 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 18:52:52.137779 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 18:52:52.137791 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 18:52:52.137802 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:52:52.137814 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:52:52.137826 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:52:52.137838 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:52:52.137850 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:52:52.137861 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.642045 | |
| I0623 18:52:52.137873 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.801802 | |
| I0623 18:52:52.137889 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.6121 (* 0.3 = 0.483629 loss) | |
| I0623 18:52:52.137903 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 1.10296 (* 0.3 = 0.330889 loss) | |
| I0623 18:52:52.137918 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.335903 (* 0.0272727 = 0.00916099 loss) | |
| I0623 18:52:52.137933 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.556325 (* 0.0272727 = 0.0151725 loss) | |
| I0623 18:52:52.137946 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.44995 (* 0.0272727 = 0.039544 loss) | |
| I0623 18:52:52.137959 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.66052 (* 0.0272727 = 0.0452868 loss) | |
| I0623 18:52:52.137974 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.93922 (* 0.0272727 = 0.0528877 loss) | |
| I0623 18:52:52.137987 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.08874 (* 0.0272727 = 0.0569657 loss) | |
| I0623 18:52:52.138001 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.86586 (* 0.0272727 = 0.050887 loss) | |
| I0623 18:52:52.138015 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.68916 (* 0.0272727 = 0.0460681 loss) | |
| I0623 18:52:52.138028 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.70717 (* 0.0272727 = 0.0465593 loss) | |
| I0623 18:52:52.138041 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.23993 (* 0.0272727 = 0.0610889 loss) | |
| I0623 18:52:52.138056 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.74778 (* 0.0272727 = 0.0749394 loss) | |
| I0623 18:52:52.138069 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.03384 (* 0.0272727 = 0.0554684 loss) | |
| I0623 18:52:52.138106 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.13205 (* 0.0272727 = 0.0308742 loss) | |
| I0623 18:52:52.138121 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.69858 (* 0.0272727 = 0.046325 loss) | |
| I0623 18:52:52.138135 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.44593 (* 0.0272727 = 0.0394345 loss) | |
| I0623 18:52:52.138149 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.865909 (* 0.0272727 = 0.0236157 loss) | |
| I0623 18:52:52.138162 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.0706981 (* 0.0272727 = 0.00192813 loss) | |
| I0623 18:52:52.138176 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0125749 (* 0.0272727 = 0.000342952 loss) | |
| I0623 18:52:52.138190 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.00462721 (* 0.0272727 = 0.000126197 loss) | |
| I0623 18:52:52.138206 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.000585772 (* 0.0272727 = 1.59756e-05 loss) | |
| I0623 18:52:52.138219 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 9.94201e-05 (* 0.0272727 = 2.71146e-06 loss) | |
| I0623 18:52:52.138233 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.57732e-05 (* 0.0272727 = 9.75632e-07 loss) | |
| I0623 18:52:52.138245 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.504505 | |
| I0623 18:52:52.138258 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.75 | |
| I0623 18:52:52.138270 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 18:52:52.138281 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.5 | |
| I0623 18:52:52.138293 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 18:52:52.138305 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.625 | |
| I0623 18:52:52.138316 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.25 | |
| I0623 18:52:52.138327 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 18:52:52.138339 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.5 | |
| I0623 18:52:52.138350 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 18:52:52.138362 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 18:52:52.138373 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.125 | |
| I0623 18:52:52.138386 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 18:52:52.138396 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 18:52:52.138408 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.375 | |
| I0623 18:52:52.138419 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.625 | |
| I0623 18:52:52.138432 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.625 | |
| I0623 18:52:52.138442 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 18:52:52.138453 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:52:52.138465 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:52:52.138476 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:52:52.138489 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:52:52.138499 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:52:52.138510 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.676136 | |
| I0623 18:52:52.138522 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.828829 | |
| I0623 18:52:52.138535 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.53526 (* 0.3 = 0.460578 loss) | |
| I0623 18:52:52.138550 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 1.02702 (* 0.3 = 0.308105 loss) | |
| I0623 18:52:52.138563 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.919409 (* 0.0272727 = 0.0250748 loss) | |
| I0623 18:52:52.138577 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.516843 (* 0.0272727 = 0.0140957 loss) | |
| I0623 18:52:52.138602 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.995548 (* 0.0272727 = 0.0271513 loss) | |
| I0623 18:52:52.138617 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.30772 (* 0.0272727 = 0.035665 loss) | |
| I0623 18:52:52.138630 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.38678 (* 0.0272727 = 0.0378214 loss) | |
| I0623 18:52:52.138644 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.23572 (* 0.0272727 = 0.0609742 loss) | |
| I0623 18:52:52.138659 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.38863 (* 0.0272727 = 0.0378718 loss) | |
| I0623 18:52:52.138671 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.81865 (* 0.0272727 = 0.0495997 loss) | |
| I0623 18:52:52.138685 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.59603 (* 0.0272727 = 0.0435282 loss) | |
| I0623 18:52:52.138700 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.80088 (* 0.0272727 = 0.049115 loss) | |
| I0623 18:52:52.138712 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 2.94058 (* 0.0272727 = 0.0801976 loss) | |
| I0623 18:52:52.138726 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.26093 (* 0.0272727 = 0.0343889 loss) | |
| I0623 18:52:52.138739 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.04098 (* 0.0272727 = 0.0283905 loss) | |
| I0623 18:52:52.138754 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.82803 (* 0.0272727 = 0.0498554 loss) | |
| I0623 18:52:52.138768 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 1.46489 (* 0.0272727 = 0.0399516 loss) | |
| I0623 18:52:52.138782 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.970589 (* 0.0272727 = 0.0264706 loss) | |
| I0623 18:52:52.138795 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.116799 (* 0.0272727 = 0.00318544 loss) | |
| I0623 18:52:52.138809 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0220512 (* 0.0272727 = 0.000601397 loss) | |
| I0623 18:52:52.138823 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00764431 (* 0.0272727 = 0.000208481 loss) | |
| I0623 18:52:52.138838 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00201862 (* 0.0272727 = 5.50532e-05 loss) | |
| I0623 18:52:52.138851 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000600691 (* 0.0272727 = 1.63825e-05 loss) | |
| I0623 18:52:52.138865 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000139477 (* 0.0272727 = 3.80392e-06 loss) | |
| I0623 18:52:52.138877 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.666667 | |
| I0623 18:52:52.138890 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 18:52:52.138900 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 18:52:52.138912 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 0.875 | |
| I0623 18:52:52.138924 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.75 | |
| I0623 18:52:52.138936 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.75 | |
| I0623 18:52:52.138947 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.625 | |
| I0623 18:52:52.138959 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 18:52:52.138972 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.5 | |
| I0623 18:52:52.138983 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.5 | |
| I0623 18:52:52.138994 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.5 | |
| I0623 18:52:52.139006 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.375 | |
| I0623 18:52:52.139019 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 18:52:52.139029 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 18:52:52.139041 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.5 | |
| I0623 18:52:52.139053 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.625 | |
| I0623 18:52:52.139065 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 18:52:52.139086 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 18:52:52.139099 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:52:52.139111 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:52:52.139122 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:52:52.139138 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:52:52.139152 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:52:52.139163 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.784091 | |
| I0623 18:52:52.139174 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.954955 | |
| I0623 18:52:52.139189 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.910466 (* 1 = 0.910466 loss) | |
| I0623 18:52:52.139202 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.604112 (* 1 = 0.604112 loss) | |
| I0623 18:52:52.139217 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0494104 (* 0.0909091 = 0.00449185 loss) | |
| I0623 18:52:52.139230 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.983915 (* 0.0909091 = 0.0894468 loss) | |
| I0623 18:52:52.139245 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.75725 (* 0.0909091 = 0.0688409 loss) | |
| I0623 18:52:52.139258 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.56604 (* 0.0909091 = 0.0514582 loss) | |
| I0623 18:52:52.139272 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 1.00809 (* 0.0909091 = 0.0916445 loss) | |
| I0623 18:52:52.139286 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 1.33245 (* 0.0909091 = 0.121132 loss) | |
| I0623 18:52:52.139299 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.629925 (* 0.0909091 = 0.0572659 loss) | |
| I0623 18:52:52.139313 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 1.42446 (* 0.0909091 = 0.129497 loss) | |
| I0623 18:52:52.139327 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.896901 (* 0.0909091 = 0.0815364 loss) | |
| I0623 18:52:52.139340 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 1.68088 (* 0.0909091 = 0.152808 loss) | |
| I0623 18:52:52.139353 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 1.87448 (* 0.0909091 = 0.170407 loss) | |
| I0623 18:52:52.139367 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.14501 (* 0.0909091 = 0.104092 loss) | |
| I0623 18:52:52.139380 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.730746 (* 0.0909091 = 0.0664314 loss) | |
| I0623 18:52:52.139394 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.08032 (* 0.0909091 = 0.0982112 loss) | |
| I0623 18:52:52.139407 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.824737 (* 0.0909091 = 0.0749761 loss) | |
| I0623 18:52:52.139421 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.386723 (* 0.0909091 = 0.0351566 loss) | |
| I0623 18:52:52.139436 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.049316 (* 0.0909091 = 0.00448327 loss) | |
| I0623 18:52:52.139449 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00235865 (* 0.0909091 = 0.000214423 loss) | |
| I0623 18:52:52.139463 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000580105 (* 0.0909091 = 5.27368e-05 loss) | |
| I0623 18:52:52.139477 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000102701 (* 0.0909091 = 9.33645e-06 loss) | |
| I0623 18:52:52.139490 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 3.78511e-05 (* 0.0909091 = 3.44101e-06 loss) | |
| I0623 18:52:52.139504 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 5.40915e-06 (* 0.0909091 = 4.91741e-07 loss) | |
| I0623 18:52:52.139516 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.125 | |
| I0623 18:52:52.139528 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0 | |
| I0623 18:52:52.139539 10365 solver.cpp:245] Train net output #149: total_confidence = 0.0793492 | |
| I0623 18:52:52.139561 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.0418083 | |
| I0623 18:52:52.139575 10365 sgd_solver.cpp:106] Iteration 21500, lr = 0.001 | |
| I0623 18:53:44.604889 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.8602 > 30) by scale factor 0.836582 | |
| I0623 18:54:06.857218 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.1109 > 30) by scale factor 0.747926 | |
| I0623 18:56:19.697372 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.6634 > 30) by scale factor 0.818254 | |
| I0623 18:56:28.135658 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.417 > 30) by scale factor 0.986289 | |
| I0623 18:58:28.463407 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 44.199 > 30) by scale factor 0.678748 | |
| I0623 18:58:48.398300 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 40.5815 > 30) by scale factor 0.739254 | |
| I0623 18:59:15.645901 10365 solver.cpp:229] Iteration 22000, loss = 4.44112 | |
| I0623 18:59:15.646026 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.494505 | |
| I0623 18:59:15.646047 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.75 | |
| I0623 18:59:15.646060 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 1 | |
| I0623 18:59:15.646073 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.375 | |
| I0623 18:59:15.646085 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 18:59:15.646098 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.5 | |
| I0623 18:59:15.646111 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.625 | |
| I0623 18:59:15.646123 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 18:59:15.646136 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.25 | |
| I0623 18:59:15.646147 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 18:59:15.646159 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.375 | |
| I0623 18:59:15.646172 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.625 | |
| I0623 18:59:15.646184 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.625 | |
| I0623 18:59:15.646195 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.5 | |
| I0623 18:59:15.646208 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.875 | |
| I0623 18:59:15.646219 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 18:59:15.646230 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 18:59:15.646242 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 18:59:15.646253 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 18:59:15.646268 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 18:59:15.646281 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 18:59:15.646292 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 18:59:15.646304 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 18:59:15.646316 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.721591 | |
| I0623 18:59:15.646327 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.813187 | |
| I0623 18:59:15.646343 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.45741 (* 0.3 = 0.437224 loss) | |
| I0623 18:59:15.646358 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.829831 (* 0.3 = 0.248949 loss) | |
| I0623 18:59:15.646373 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.353194 (* 0.0272727 = 0.00963256 loss) | |
| I0623 18:59:15.646386 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.244439 (* 0.0272727 = 0.00666652 loss) | |
| I0623 18:59:15.646401 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.80816 (* 0.0272727 = 0.0493135 loss) | |
| I0623 18:59:15.646414 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.60866 (* 0.0272727 = 0.0438725 loss) | |
| I0623 18:59:15.646428 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.56577 (* 0.0272727 = 0.0427029 loss) | |
| I0623 18:59:15.646441 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.24621 (* 0.0272727 = 0.0339875 loss) | |
| I0623 18:59:15.646456 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.04443 (* 0.0272727 = 0.0557571 loss) | |
| I0623 18:59:15.646469 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.90102 (* 0.0272727 = 0.0518459 loss) | |
| I0623 18:59:15.646482 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.14351 (* 0.0272727 = 0.0584594 loss) | |
| I0623 18:59:15.646497 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.7185 (* 0.0272727 = 0.0468682 loss) | |
| I0623 18:59:15.646510 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.50016 (* 0.0272727 = 0.0409135 loss) | |
| I0623 18:59:15.646524 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.2844 (* 0.0272727 = 0.0350292 loss) | |
| I0623 18:59:15.646555 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.46361 (* 0.0272727 = 0.0399168 loss) | |
| I0623 18:59:15.646570 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.704772 (* 0.0272727 = 0.0192211 loss) | |
| I0623 18:59:15.646584 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.674432 (* 0.0272727 = 0.0183936 loss) | |
| I0623 18:59:15.646598 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.678684 (* 0.0272727 = 0.0185096 loss) | |
| I0623 18:59:15.646612 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.578502 (* 0.0272727 = 0.0157773 loss) | |
| I0623 18:59:15.646625 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.152778 (* 0.0272727 = 0.00416666 loss) | |
| I0623 18:59:15.646639 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0469317 (* 0.0272727 = 0.00127996 loss) | |
| I0623 18:59:15.646653 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.0264072 (* 0.0272727 = 0.000720196 loss) | |
| I0623 18:59:15.646667 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.00108324 (* 0.0272727 = 2.95428e-05 loss) | |
| I0623 18:59:15.646682 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 0.000334542 (* 0.0272727 = 9.12389e-06 loss) | |
| I0623 18:59:15.646693 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.571429 | |
| I0623 18:59:15.646706 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 18:59:15.646718 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 18:59:15.646730 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 18:59:15.646741 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 18:59:15.646754 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 18:59:15.646764 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.875 | |
| I0623 18:59:15.646776 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 18:59:15.646787 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.25 | |
| I0623 18:59:15.646800 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 18:59:15.646811 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 18:59:15.646821 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 18:59:15.646833 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 18:59:15.646844 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 18:59:15.646855 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.875 | |
| I0623 18:59:15.646867 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 18:59:15.646878 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 18:59:15.646889 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 18:59:15.646901 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 18:59:15.646913 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 18:59:15.646924 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 18:59:15.646935 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 18:59:15.646947 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 18:59:15.646958 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.767045 | |
| I0623 18:59:15.646970 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.857143 | |
| I0623 18:59:15.646983 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.16852 (* 0.3 = 0.350555 loss) | |
| I0623 18:59:15.646997 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.640259 (* 0.3 = 0.192078 loss) | |
| I0623 18:59:15.647011 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.249899 (* 0.0272727 = 0.00681542 loss) | |
| I0623 18:59:15.647024 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.280358 (* 0.0272727 = 0.00764614 loss) | |
| I0623 18:59:15.647053 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.270061 (* 0.0272727 = 0.00736531 loss) | |
| I0623 18:59:15.647069 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.00829 (* 0.0272727 = 0.0274987 loss) | |
| I0623 18:59:15.647083 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.26145 (* 0.0272727 = 0.0344033 loss) | |
| I0623 18:59:15.647096 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 0.669832 (* 0.0272727 = 0.0182681 loss) | |
| I0623 18:59:15.647110 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.69446 (* 0.0272727 = 0.0462127 loss) | |
| I0623 18:59:15.647124 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 2.44115 (* 0.0272727 = 0.0665768 loss) | |
| I0623 18:59:15.647137 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.95911 (* 0.0272727 = 0.0534302 loss) | |
| I0623 18:59:15.647151 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.51756 (* 0.0272727 = 0.0413881 loss) | |
| I0623 18:59:15.647166 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.37008 (* 0.0272727 = 0.0373657 loss) | |
| I0623 18:59:15.647179 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.48284 (* 0.0272727 = 0.0404411 loss) | |
| I0623 18:59:15.647192 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.03982 (* 0.0272727 = 0.0283587 loss) | |
| I0623 18:59:15.647207 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.580037 (* 0.0272727 = 0.0158192 loss) | |
| I0623 18:59:15.647219 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.663401 (* 0.0272727 = 0.0180928 loss) | |
| I0623 18:59:15.647233 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.453559 (* 0.0272727 = 0.0123698 loss) | |
| I0623 18:59:15.647248 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.357531 (* 0.0272727 = 0.00975086 loss) | |
| I0623 18:59:15.647261 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.217957 (* 0.0272727 = 0.0059443 loss) | |
| I0623 18:59:15.647275 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0617434 (* 0.0272727 = 0.00168391 loss) | |
| I0623 18:59:15.647289 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.011146 (* 0.0272727 = 0.000303981 loss) | |
| I0623 18:59:15.647302 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00081854 (* 0.0272727 = 2.23238e-05 loss) | |
| I0623 18:59:15.647320 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 5.53176e-05 (* 0.0272727 = 1.50866e-06 loss) | |
| I0623 18:59:15.647332 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.868132 | |
| I0623 18:59:15.647344 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 18:59:15.647356 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 18:59:15.647367 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 18:59:15.647378 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 18:59:15.647390 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 18:59:15.647402 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 18:59:15.647413 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 18:59:15.647424 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 18:59:15.647436 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 18:59:15.647447 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 18:59:15.647459 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.875 | |
| I0623 18:59:15.647470 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 18:59:15.647481 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 18:59:15.647493 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.875 | |
| I0623 18:59:15.647505 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 18:59:15.647516 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.75 | |
| I0623 18:59:15.647537 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 18:59:15.647550 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 18:59:15.647562 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 18:59:15.647573 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 18:59:15.647585 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 18:59:15.647608 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 18:59:15.647622 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.926136 | |
| I0623 18:59:15.647634 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.989011 | |
| I0623 18:59:15.647649 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.419914 (* 1 = 0.419914 loss) | |
| I0623 18:59:15.647661 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.230087 (* 1 = 0.230087 loss) | |
| I0623 18:59:15.647675 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0217514 (* 0.0909091 = 0.0019774 loss) | |
| I0623 18:59:15.647691 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0338301 (* 0.0909091 = 0.00307546 loss) | |
| I0623 18:59:15.647703 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0512309 (* 0.0909091 = 0.00465735 loss) | |
| I0623 18:59:15.647717 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0580617 (* 0.0909091 = 0.00527833 loss) | |
| I0623 18:59:15.647732 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.371956 (* 0.0909091 = 0.0338142 loss) | |
| I0623 18:59:15.647745 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.164658 (* 0.0909091 = 0.0149689 loss) | |
| I0623 18:59:15.647759 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.344298 (* 0.0909091 = 0.0312998 loss) | |
| I0623 18:59:15.647773 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.364353 (* 0.0909091 = 0.033123 loss) | |
| I0623 18:59:15.647786 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.663841 (* 0.0909091 = 0.0603492 loss) | |
| I0623 18:59:15.647799 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.691802 (* 0.0909091 = 0.0628911 loss) | |
| I0623 18:59:15.647814 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.466534 (* 0.0909091 = 0.0424122 loss) | |
| I0623 18:59:15.647826 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.883513 (* 0.0909091 = 0.0803194 loss) | |
| I0623 18:59:15.647840 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.798368 (* 0.0909091 = 0.0725789 loss) | |
| I0623 18:59:15.647853 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.522093 (* 0.0909091 = 0.047463 loss) | |
| I0623 18:59:15.647867 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.579114 (* 0.0909091 = 0.0526468 loss) | |
| I0623 18:59:15.647881 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.347747 (* 0.0909091 = 0.0316133 loss) | |
| I0623 18:59:15.647894 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.378029 (* 0.0909091 = 0.0343662 loss) | |
| I0623 18:59:15.647908 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0830841 (* 0.0909091 = 0.0075531 loss) | |
| I0623 18:59:15.647922 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.0101494 (* 0.0909091 = 0.000922677 loss) | |
| I0623 18:59:15.647935 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000801559 (* 0.0909091 = 7.2869e-05 loss) | |
| I0623 18:59:15.647949 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 0.000250531 (* 0.0909091 = 2.27755e-05 loss) | |
| I0623 18:59:15.647964 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 2.03415e-05 (* 0.0909091 = 1.84922e-06 loss) | |
| I0623 18:59:15.647975 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.5 | |
| I0623 18:59:15.647987 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.375 | |
| I0623 18:59:15.648010 10365 solver.cpp:245] Train net output #149: total_confidence = 0.277416 | |
| I0623 18:59:15.648023 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.244458 | |
| I0623 18:59:15.648036 10365 sgd_solver.cpp:106] Iteration 22000, lr = 0.001 | |
| I0623 19:04:25.096328 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.7714 > 30) by scale factor 0.838658 | |
| I0623 19:04:25.864925 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 30.0928 > 30) by scale factor 0.996918 | |
| I0623 19:04:55.037201 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 31.5863 > 30) by scale factor 0.94978 | |
| I0623 19:05:39.176476 10365 solver.cpp:229] Iteration 22500, loss = 4.47208 | |
| I0623 19:05:39.176614 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.530612 | |
| I0623 19:05:39.176635 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.625 | |
| I0623 19:05:39.176648 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.5 | |
| I0623 19:05:39.176661 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.5 | |
| I0623 19:05:39.176673 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 19:05:39.176686 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 19:05:39.176698 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.5 | |
| I0623 19:05:39.176710 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 19:05:39.176723 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.875 | |
| I0623 19:05:39.176735 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.5 | |
| I0623 19:05:39.176748 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 19:05:39.176759 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 19:05:39.176771 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.5 | |
| I0623 19:05:39.176784 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.375 | |
| I0623 19:05:39.176795 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.375 | |
| I0623 19:05:39.176806 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.375 | |
| I0623 19:05:39.176817 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.875 | |
| I0623 19:05:39.176829 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 1 | |
| I0623 19:05:39.176841 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 19:05:39.176852 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 19:05:39.176864 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 19:05:39.176875 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 19:05:39.176887 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 19:05:39.176898 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.727273 | |
| I0623 19:05:39.176909 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.816327 | |
| I0623 19:05:39.176925 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.47393 (* 0.3 = 0.442179 loss) | |
| I0623 19:05:39.176939 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.882052 (* 0.3 = 0.264616 loss) | |
| I0623 19:05:39.176954 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.685437 (* 0.0272727 = 0.0186937 loss) | |
| I0623 19:05:39.176969 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.18043 (* 0.0272727 = 0.0321936 loss) | |
| I0623 19:05:39.176981 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.89567 (* 0.0272727 = 0.0517001 loss) | |
| I0623 19:05:39.176995 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.52782 (* 0.0272727 = 0.0416678 loss) | |
| I0623 19:05:39.177009 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 2.03625 (* 0.0272727 = 0.0555341 loss) | |
| I0623 19:05:39.177023 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.216 (* 0.0272727 = 0.0331636 loss) | |
| I0623 19:05:39.177037 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.84322 (* 0.0272727 = 0.0502696 loss) | |
| I0623 19:05:39.177050 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 0.891403 (* 0.0272727 = 0.024311 loss) | |
| I0623 19:05:39.177064 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.48251 (* 0.0272727 = 0.040432 loss) | |
| I0623 19:05:39.177078 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.28092 (* 0.0272727 = 0.0349342 loss) | |
| I0623 19:05:39.177091 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.9262 (* 0.0272727 = 0.0525326 loss) | |
| I0623 19:05:39.177105 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.6377 (* 0.0272727 = 0.0446645 loss) | |
| I0623 19:05:39.177136 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.71092 (* 0.0272727 = 0.0466614 loss) | |
| I0623 19:05:39.177151 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 2.0118 (* 0.0272727 = 0.0548671 loss) | |
| I0623 19:05:39.177165 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 1.30564 (* 0.0272727 = 0.0356084 loss) | |
| I0623 19:05:39.177178 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.534951 (* 0.0272727 = 0.0145896 loss) | |
| I0623 19:05:39.177192 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.177005 (* 0.0272727 = 0.0048274 loss) | |
| I0623 19:05:39.177206 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.0484574 (* 0.0272727 = 0.00132157 loss) | |
| I0623 19:05:39.177220 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0174067 (* 0.0272727 = 0.000474728 loss) | |
| I0623 19:05:39.177234 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00176914 (* 0.0272727 = 4.82492e-05 loss) | |
| I0623 19:05:39.177248 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000338998 (* 0.0272727 = 9.24541e-06 loss) | |
| I0623 19:05:39.177265 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 9.1927e-05 (* 0.0272727 = 2.5071e-06 loss) | |
| I0623 19:05:39.177278 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.602041 | |
| I0623 19:05:39.177290 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 19:05:39.177302 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 19:05:39.177314 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.875 | |
| I0623 19:05:39.177325 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.75 | |
| I0623 19:05:39.177337 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.875 | |
| I0623 19:05:39.177348 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 19:05:39.177361 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.625 | |
| I0623 19:05:39.177371 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 19:05:39.177383 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.625 | |
| I0623 19:05:39.177394 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.5 | |
| I0623 19:05:39.177405 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.5 | |
| I0623 19:05:39.177417 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.375 | |
| I0623 19:05:39.177428 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.375 | |
| I0623 19:05:39.177440 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 19:05:39.177451 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.875 | |
| I0623 19:05:39.177462 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 1 | |
| I0623 19:05:39.177474 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 1 | |
| I0623 19:05:39.177484 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 19:05:39.177496 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 19:05:39.177507 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 19:05:39.177518 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 19:05:39.177530 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 19:05:39.177541 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.767045 | |
| I0623 19:05:39.177552 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.857143 | |
| I0623 19:05:39.177566 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.22619 (* 0.3 = 0.367858 loss) | |
| I0623 19:05:39.177579 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.708885 (* 0.3 = 0.212665 loss) | |
| I0623 19:05:39.177593 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.463806 (* 0.0272727 = 0.0126493 loss) | |
| I0623 19:05:39.177608 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.350479 (* 0.0272727 = 0.00955852 loss) | |
| I0623 19:05:39.177635 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.378758 (* 0.0272727 = 0.0103298 loss) | |
| I0623 19:05:39.177651 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 0.868203 (* 0.0272727 = 0.0236783 loss) | |
| I0623 19:05:39.177665 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.15459 (* 0.0272727 = 0.0314889 loss) | |
| I0623 19:05:39.177680 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 0.95984 (* 0.0272727 = 0.0261775 loss) | |
| I0623 19:05:39.177692 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.50235 (* 0.0272727 = 0.0409733 loss) | |
| I0623 19:05:39.177706 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 0.775321 (* 0.0272727 = 0.0211451 loss) | |
| I0623 19:05:39.177721 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.04545 (* 0.0272727 = 0.0285122 loss) | |
| I0623 19:05:39.177733 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.21613 (* 0.0272727 = 0.0331673 loss) | |
| I0623 19:05:39.177747 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.34778 (* 0.0272727 = 0.0367577 loss) | |
| I0623 19:05:39.177760 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.67619 (* 0.0272727 = 0.0457144 loss) | |
| I0623 19:05:39.177773 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.63737 (* 0.0272727 = 0.0446554 loss) | |
| I0623 19:05:39.177786 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.95833 (* 0.0272727 = 0.0534091 loss) | |
| I0623 19:05:39.177800 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.784183 (* 0.0272727 = 0.0213868 loss) | |
| I0623 19:05:39.177814 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.243443 (* 0.0272727 = 0.00663935 loss) | |
| I0623 19:05:39.177829 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.0352275 (* 0.0272727 = 0.00096075 loss) | |
| I0623 19:05:39.177842 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.00617683 (* 0.0272727 = 0.000168459 loss) | |
| I0623 19:05:39.177856 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00114236 (* 0.0272727 = 3.11552e-05 loss) | |
| I0623 19:05:39.177870 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00087112 (* 0.0272727 = 2.37578e-05 loss) | |
| I0623 19:05:39.177884 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 7.84436e-05 (* 0.0272727 = 2.13937e-06 loss) | |
| I0623 19:05:39.177898 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.63322e-05 (* 0.0272727 = 4.45424e-07 loss) | |
| I0623 19:05:39.177911 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.867347 | |
| I0623 19:05:39.177922 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 19:05:39.177933 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 19:05:39.177945 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 19:05:39.177956 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 19:05:39.177968 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 19:05:39.177979 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 1 | |
| I0623 19:05:39.177990 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 1 | |
| I0623 19:05:39.178002 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 19:05:39.178014 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 19:05:39.178025 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 19:05:39.178036 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 19:05:39.178047 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 19:05:39.178059 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.625 | |
| I0623 19:05:39.178071 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.625 | |
| I0623 19:05:39.178081 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 19:05:39.178093 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 19:05:39.178114 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 1 | |
| I0623 19:05:39.178128 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 19:05:39.178139 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 19:05:39.178150 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 19:05:39.178161 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 19:05:39.178174 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 19:05:39.178184 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.920455 | |
| I0623 19:05:39.178196 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.969388 | |
| I0623 19:05:39.178210 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.45821 (* 1 = 0.45821 loss) | |
| I0623 19:05:39.178223 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.268844 (* 1 = 0.268844 loss) | |
| I0623 19:05:39.178237 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.114515 (* 0.0909091 = 0.0104104 loss) | |
| I0623 19:05:39.178251 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0641355 (* 0.0909091 = 0.0058305 loss) | |
| I0623 19:05:39.178266 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0541238 (* 0.0909091 = 0.00492035 loss) | |
| I0623 19:05:39.178278 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.030764 (* 0.0909091 = 0.00279672 loss) | |
| I0623 19:05:39.178292 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.0994836 (* 0.0909091 = 0.00904396 loss) | |
| I0623 19:05:39.178306 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.261978 (* 0.0909091 = 0.0238162 loss) | |
| I0623 19:05:39.178323 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.0679585 (* 0.0909091 = 0.00617805 loss) | |
| I0623 19:05:39.178338 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.149213 (* 0.0909091 = 0.0135649 loss) | |
| I0623 19:05:39.178351 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.186312 (* 0.0909091 = 0.0169374 loss) | |
| I0623 19:05:39.178364 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.511739 (* 0.0909091 = 0.0465217 loss) | |
| I0623 19:05:39.178378 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.867202 (* 0.0909091 = 0.0788365 loss) | |
| I0623 19:05:39.178392 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.13752 (* 0.0909091 = 0.103411 loss) | |
| I0623 19:05:39.178406 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 1.01403 (* 0.0909091 = 0.0921849 loss) | |
| I0623 19:05:39.178419 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 2.15029 (* 0.0909091 = 0.19548 loss) | |
| I0623 19:05:39.178432 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.627284 (* 0.0909091 = 0.0570258 loss) | |
| I0623 19:05:39.178447 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.166836 (* 0.0909091 = 0.0151669 loss) | |
| I0623 19:05:39.178460 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.00513969 (* 0.0909091 = 0.000467245 loss) | |
| I0623 19:05:39.178474 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.000176837 (* 0.0909091 = 1.6076e-05 loss) | |
| I0623 19:05:39.178488 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 3.82975e-05 (* 0.0909091 = 3.48159e-06 loss) | |
| I0623 19:05:39.178503 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 2.89539e-05 (* 0.0909091 = 2.63218e-06 loss) | |
| I0623 19:05:39.178516 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 2.09665e-05 (* 0.0909091 = 1.90604e-06 loss) | |
| I0623 19:05:39.178529 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 4.87272e-06 (* 0.0909091 = 4.42975e-07 loss) | |
| I0623 19:05:39.178541 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.25 | |
| I0623 19:05:39.178553 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.25 | |
| I0623 19:05:39.178565 10365 solver.cpp:245] Train net output #149: total_confidence = 0.198386 | |
| I0623 19:05:39.178586 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.179517 | |
| I0623 19:05:39.178599 10365 sgd_solver.cpp:106] Iteration 22500, lr = 0.001 | |
| I0623 19:07:24.577302 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 38.0769 > 30) by scale factor 0.787879 | |
| I0623 19:10:05.843128 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.6384 > 30) by scale factor 0.866091 | |
| I0623 19:11:57.020308 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 42.9467 > 30) by scale factor 0.698541 | |
| I0623 19:12:02.390028 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.5678 > 30) by scale factor 0.843459 | |
| I0623 19:12:02.796979 10365 solver.cpp:229] Iteration 23000, loss = 4.39844 | |
| I0623 19:12:02.797046 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.419355 | |
| I0623 19:12:02.797065 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 1 | |
| I0623 19:12:02.797080 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.75 | |
| I0623 19:12:02.797093 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.125 | |
| I0623 19:12:02.797106 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0 | |
| I0623 19:12:02.797118 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.375 | |
| I0623 19:12:02.797134 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.25 | |
| I0623 19:12:02.797147 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 19:12:02.797160 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.625 | |
| I0623 19:12:02.797173 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 19:12:02.797185 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 19:12:02.797199 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.5 | |
| I0623 19:12:02.797211 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.375 | |
| I0623 19:12:02.797224 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 19:12:02.797235 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.625 | |
| I0623 19:12:02.797248 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.625 | |
| I0623 19:12:02.797260 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 19:12:02.797271 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 19:12:02.797283 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 19:12:02.797297 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 19:12:02.797308 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 19:12:02.797320 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 19:12:02.797333 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 19:12:02.797344 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.670455 | |
| I0623 19:12:02.797358 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.763441 | |
| I0623 19:12:02.797374 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.67255 (* 0.3 = 0.501765 loss) | |
| I0623 19:12:02.797387 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.950513 (* 0.3 = 0.285154 loss) | |
| I0623 19:12:02.797404 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.13566 (* 0.0272727 = 0.00369981 loss) | |
| I0623 19:12:02.797417 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.16426 (* 0.0272727 = 0.0317526 loss) | |
| I0623 19:12:02.797431 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.56544 (* 0.0272727 = 0.0426937 loss) | |
| I0623 19:12:02.797446 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 2.44679 (* 0.0272727 = 0.0667307 loss) | |
| I0623 19:12:02.797461 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.67114 (* 0.0272727 = 0.0455764 loss) | |
| I0623 19:12:02.797473 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 2.35478 (* 0.0272727 = 0.0642212 loss) | |
| I0623 19:12:02.797487 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.78622 (* 0.0272727 = 0.0487152 loss) | |
| I0623 19:12:02.797502 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.19396 (* 0.0272727 = 0.0325626 loss) | |
| I0623 19:12:02.797515 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.44642 (* 0.0272727 = 0.0394477 loss) | |
| I0623 19:12:02.797529 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.50822 (* 0.0272727 = 0.0411334 loss) | |
| I0623 19:12:02.797544 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.24799 (* 0.0272727 = 0.0340362 loss) | |
| I0623 19:12:02.797607 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.34975 (* 0.0272727 = 0.0368113 loss) | |
| I0623 19:12:02.797623 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.23953 (* 0.0272727 = 0.0338053 loss) | |
| I0623 19:12:02.797637 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.828895 (* 0.0272727 = 0.0226062 loss) | |
| I0623 19:12:02.797652 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 2.10044 (* 0.0272727 = 0.0572846 loss) | |
| I0623 19:12:02.797665 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.767193 (* 0.0272727 = 0.0209235 loss) | |
| I0623 19:12:02.797679 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.779365 (* 0.0272727 = 0.0212554 loss) | |
| I0623 19:12:02.797693 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.00390422 (* 0.0272727 = 0.000106479 loss) | |
| I0623 19:12:02.797708 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.000789079 (* 0.0272727 = 2.15203e-05 loss) | |
| I0623 19:12:02.797722 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 8.28066e-05 (* 0.0272727 = 2.25836e-06 loss) | |
| I0623 19:12:02.797736 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 3.93351e-05 (* 0.0272727 = 1.07278e-06 loss) | |
| I0623 19:12:02.797750 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 3.57629e-06 (* 0.0272727 = 9.75353e-08 loss) | |
| I0623 19:12:02.797763 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.569892 | |
| I0623 19:12:02.797776 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 1 | |
| I0623 19:12:02.797788 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.75 | |
| I0623 19:12:02.797801 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 19:12:02.797812 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.625 | |
| I0623 19:12:02.797824 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.75 | |
| I0623 19:12:02.797837 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.625 | |
| I0623 19:12:02.797848 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.375 | |
| I0623 19:12:02.797860 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 19:12:02.797871 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 19:12:02.797883 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 19:12:02.797895 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.625 | |
| I0623 19:12:02.797907 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.5 | |
| I0623 19:12:02.797919 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 19:12:02.797931 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 19:12:02.797942 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 19:12:02.797955 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 19:12:02.797966 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.75 | |
| I0623 19:12:02.797978 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 19:12:02.797991 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 19:12:02.798002 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 19:12:02.798014 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 19:12:02.798027 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 19:12:02.798038 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.761364 | |
| I0623 19:12:02.798050 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.870968 | |
| I0623 19:12:02.798064 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.23414 (* 0.3 = 0.370243 loss) | |
| I0623 19:12:02.798079 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.703861 (* 0.3 = 0.211158 loss) | |
| I0623 19:12:02.801827 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.166004 (* 0.0272727 = 0.00452737 loss) | |
| I0623 19:12:02.801851 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.511346 (* 0.0272727 = 0.0139458 loss) | |
| I0623 19:12:02.801867 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.219341 (* 0.0272727 = 0.00598204 loss) | |
| I0623 19:12:02.801882 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.75275 (* 0.0272727 = 0.0478021 loss) | |
| I0623 19:12:02.801898 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 0.76263 (* 0.0272727 = 0.020799 loss) | |
| I0623 19:12:02.801911 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 2.38262 (* 0.0272727 = 0.0649807 loss) | |
| I0623 19:12:02.801925 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.38905 (* 0.0272727 = 0.0378831 loss) | |
| I0623 19:12:02.801939 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.03152 (* 0.0272727 = 0.0281323 loss) | |
| I0623 19:12:02.801954 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.73115 (* 0.0272727 = 0.0472132 loss) | |
| I0623 19:12:02.801969 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.82418 (* 0.0272727 = 0.0497503 loss) | |
| I0623 19:12:02.801982 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.30321 (* 0.0272727 = 0.0355422 loss) | |
| I0623 19:12:02.801996 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.29307 (* 0.0272727 = 0.0352656 loss) | |
| I0623 19:12:02.802011 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 0.962024 (* 0.0272727 = 0.026237 loss) | |
| I0623 19:12:02.802026 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.853564 (* 0.0272727 = 0.023279 loss) | |
| I0623 19:12:02.802039 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.977178 (* 0.0272727 = 0.0266503 loss) | |
| I0623 19:12:02.802053 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.511563 (* 0.0272727 = 0.0139517 loss) | |
| I0623 19:12:02.802067 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.570589 (* 0.0272727 = 0.0155615 loss) | |
| I0623 19:12:02.802083 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0160505 (* 0.0272727 = 0.000437741 loss) | |
| I0623 19:12:02.802098 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00149845 (* 0.0272727 = 4.08667e-05 loss) | |
| I0623 19:12:02.802112 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00107177 (* 0.0272727 = 2.923e-05 loss) | |
| I0623 19:12:02.802127 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.00019308 (* 0.0272727 = 5.26582e-06 loss) | |
| I0623 19:12:02.802142 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.15041e-05 (* 0.0272727 = 3.13749e-07 loss) | |
| I0623 19:12:02.802155 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.83871 | |
| I0623 19:12:02.802167 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 19:12:02.802180 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 19:12:02.802191 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 19:12:02.802202 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 19:12:02.802214 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 1 | |
| I0623 19:12:02.802225 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 19:12:02.802237 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 19:12:02.802249 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.875 | |
| I0623 19:12:02.802263 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.875 | |
| I0623 19:12:02.802276 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.75 | |
| I0623 19:12:02.802289 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.75 | |
| I0623 19:12:02.802299 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 19:12:02.802311 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.75 | |
| I0623 19:12:02.802337 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.75 | |
| I0623 19:12:02.802367 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.75 | |
| I0623 19:12:02.802381 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 19:12:02.802392 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 19:12:02.802405 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 19:12:02.802417 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 19:12:02.802429 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 19:12:02.802441 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 19:12:02.802453 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 19:12:02.802464 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.903409 | |
| I0623 19:12:02.802476 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.978495 | |
| I0623 19:12:02.802490 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.597536 (* 1 = 0.597536 loss) | |
| I0623 19:12:02.802505 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.338335 (* 1 = 0.338335 loss) | |
| I0623 19:12:02.802520 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.0285183 (* 0.0909091 = 0.00259258 loss) | |
| I0623 19:12:02.802533 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.0645436 (* 0.0909091 = 0.0058676 loss) | |
| I0623 19:12:02.802548 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.0340053 (* 0.0909091 = 0.00309139 loss) | |
| I0623 19:12:02.802562 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.0262009 (* 0.0909091 = 0.0023819 loss) | |
| I0623 19:12:02.802577 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.062336 (* 0.0909091 = 0.00566691 loss) | |
| I0623 19:12:02.802592 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 1.07759 (* 0.0909091 = 0.0979626 loss) | |
| I0623 19:12:02.802605 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 1.01326 (* 0.0909091 = 0.0921149 loss) | |
| I0623 19:12:02.802619 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.594182 (* 0.0909091 = 0.0540165 loss) | |
| I0623 19:12:02.802634 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.740084 (* 0.0909091 = 0.0672803 loss) | |
| I0623 19:12:02.802649 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.606144 (* 0.0909091 = 0.055104 loss) | |
| I0623 19:12:02.802661 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.942097 (* 0.0909091 = 0.0856451 loss) | |
| I0623 19:12:02.802676 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 1.29077 (* 0.0909091 = 0.117343 loss) | |
| I0623 19:12:02.802690 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.503419 (* 0.0909091 = 0.0457654 loss) | |
| I0623 19:12:02.802703 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 0.712482 (* 0.0909091 = 0.0647711 loss) | |
| I0623 19:12:02.802717 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.948499 (* 0.0909091 = 0.0862272 loss) | |
| I0623 19:12:02.802731 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.283417 (* 0.0909091 = 0.0257652 loss) | |
| I0623 19:12:02.802744 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.393268 (* 0.0909091 = 0.0357516 loss) | |
| I0623 19:12:02.802759 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.00459031 (* 0.0909091 = 0.000417301 loss) | |
| I0623 19:12:02.802773 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000744758 (* 0.0909091 = 6.77053e-05 loss) | |
| I0623 19:12:02.802788 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000105301 (* 0.0909091 = 9.5728e-06 loss) | |
| I0623 19:12:02.802806 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 4.16151e-05 (* 0.0909091 = 3.78319e-06 loss) | |
| I0623 19:12:02.802821 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 2.87594e-06 (* 0.0909091 = 2.61449e-07 loss) | |
| I0623 19:12:02.802845 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.5 | |
| I0623 19:12:02.802860 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.5 | |
| I0623 19:12:02.802871 10365 solver.cpp:245] Train net output #149: total_confidence = 0.362685 | |
| I0623 19:12:02.802883 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.315538 | |
| I0623 19:12:02.802896 10365 sgd_solver.cpp:106] Iteration 23000, lr = 0.001 | |
| I0623 19:12:06.994076 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 33.0276 > 30) by scale factor 0.908331 | |
| I0623 19:13:12.910956 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 45.3563 > 30) by scale factor 0.661429 | |
| I0623 19:13:36.683379 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 32.89 > 30) by scale factor 0.912131 | |
| I0623 19:13:44.348840 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.2214 > 30) by scale factor 0.876645 | |
| I0623 19:14:02.735530 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 35.6101 > 30) by scale factor 0.842458 | |
| I0623 19:14:36.450295 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 36.3027 > 30) by scale factor 0.826384 | |
| I0623 19:18:26.055400 10365 solver.cpp:229] Iteration 23500, loss = 4.40985 | |
| I0623 19:18:26.055497 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.438095 | |
| I0623 19:18:26.055517 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.625 | |
| I0623 19:18:26.055531 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 19:18:26.055543 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.875 | |
| I0623 19:18:26.055557 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.5 | |
| I0623 19:18:26.055569 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 19:18:26.055582 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 19:18:26.055594 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 19:18:26.055621 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.5 | |
| I0623 19:18:26.055634 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.625 | |
| I0623 19:18:26.055647 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.625 | |
| I0623 19:18:26.055660 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.375 | |
| I0623 19:18:26.055672 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.25 | |
| I0623 19:18:26.055685 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 19:18:26.055696 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.5 | |
| I0623 19:18:26.055708 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.75 | |
| I0623 19:18:26.055721 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.625 | |
| I0623 19:18:26.055732 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 19:18:26.055743 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 1 | |
| I0623 19:18:26.055755 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 19:18:26.055768 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 19:18:26.055779 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 19:18:26.055791 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 19:18:26.055804 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.653409 | |
| I0623 19:18:26.055815 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.733333 | |
| I0623 19:18:26.055832 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.60697 (* 0.3 = 0.482092 loss) | |
| I0623 19:18:26.055846 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.998127 (* 0.3 = 0.299438 loss) | |
| I0623 19:18:26.055861 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 1.37518 (* 0.0272727 = 0.037505 loss) | |
| I0623 19:18:26.055876 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 1.24358 (* 0.0272727 = 0.0339159 loss) | |
| I0623 19:18:26.055889 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.05907 (* 0.0272727 = 0.0288838 loss) | |
| I0623 19:18:26.055903 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.38325 (* 0.0272727 = 0.037725 loss) | |
| I0623 19:18:26.055917 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.80911 (* 0.0272727 = 0.0493393 loss) | |
| I0623 19:18:26.055932 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.96566 (* 0.0272727 = 0.053609 loss) | |
| I0623 19:18:26.055945 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 2.00591 (* 0.0272727 = 0.0547065 loss) | |
| I0623 19:18:26.055959 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.31421 (* 0.0272727 = 0.0358422 loss) | |
| I0623 19:18:26.055974 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 2.00078 (* 0.0272727 = 0.0545666 loss) | |
| I0623 19:18:26.055986 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 1.81488 (* 0.0272727 = 0.0494966 loss) | |
| I0623 19:18:26.056000 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 2.2259 (* 0.0272727 = 0.0607063 loss) | |
| I0623 19:18:26.056015 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 2.06259 (* 0.0272727 = 0.0562524 loss) | |
| I0623 19:18:26.056047 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.10687 (* 0.0272727 = 0.0301875 loss) | |
| I0623 19:18:26.056062 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 1.4998 (* 0.0272727 = 0.0409037 loss) | |
| I0623 19:18:26.056076 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.693693 (* 0.0272727 = 0.0189189 loss) | |
| I0623 19:18:26.056089 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.864703 (* 0.0272727 = 0.0235828 loss) | |
| I0623 19:18:26.056103 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.301733 (* 0.0272727 = 0.00822907 loss) | |
| I0623 19:18:26.056118 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.102518 (* 0.0272727 = 0.00279595 loss) | |
| I0623 19:18:26.056131 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.014983 (* 0.0272727 = 0.000408628 loss) | |
| I0623 19:18:26.056146 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00215807 (* 0.0272727 = 5.88563e-05 loss) | |
| I0623 19:18:26.056160 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.000174623 (* 0.0272727 = 4.76244e-06 loss) | |
| I0623 19:18:26.056174 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 5.88536e-05 (* 0.0272727 = 1.6051e-06 loss) | |
| I0623 19:18:26.056190 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.504762 | |
| I0623 19:18:26.056202 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.75 | |
| I0623 19:18:26.056215 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 0.875 | |
| I0623 19:18:26.056226 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 0.625 | |
| I0623 19:18:26.056238 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.375 | |
| I0623 19:18:26.056249 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.25 | |
| I0623 19:18:26.056262 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 19:18:26.056273 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.625 | |
| I0623 19:18:26.056285 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 19:18:26.056296 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.5 | |
| I0623 19:18:26.056313 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.375 | |
| I0623 19:18:26.056324 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.25 | |
| I0623 19:18:26.056337 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.375 | |
| I0623 19:18:26.056349 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.625 | |
| I0623 19:18:26.056360 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.625 | |
| I0623 19:18:26.056371 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 19:18:26.056383 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.875 | |
| I0623 19:18:26.056394 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 19:18:26.056406 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 1 | |
| I0623 19:18:26.056418 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 19:18:26.056429 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 19:18:26.056442 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 19:18:26.056452 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 19:18:26.056464 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.698864 | |
| I0623 19:18:26.056475 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.847619 | |
| I0623 19:18:26.056490 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.33479 (* 0.3 = 0.400437 loss) | |
| I0623 19:18:26.056504 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.826063 (* 0.3 = 0.247819 loss) | |
| I0623 19:18:26.056517 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.827653 (* 0.0272727 = 0.0225724 loss) | |
| I0623 19:18:26.056531 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.78218 (* 0.0272727 = 0.0213322 loss) | |
| I0623 19:18:26.056556 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.751358 (* 0.0272727 = 0.0204916 loss) | |
| I0623 19:18:26.056571 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.71874 (* 0.0272727 = 0.0468748 loss) | |
| I0623 19:18:26.056586 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.64669 (* 0.0272727 = 0.0449098 loss) | |
| I0623 19:18:26.056599 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.66989 (* 0.0272727 = 0.0455425 loss) | |
| I0623 19:18:26.056612 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.4097 (* 0.0272727 = 0.0384465 loss) | |
| I0623 19:18:26.056627 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 0.977195 (* 0.0272727 = 0.0266508 loss) | |
| I0623 19:18:26.056640 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.91375 (* 0.0272727 = 0.0521933 loss) | |
| I0623 19:18:26.056653 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 1.58857 (* 0.0272727 = 0.0433246 loss) | |
| I0623 19:18:26.056668 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.82003 (* 0.0272727 = 0.0496371 loss) | |
| I0623 19:18:26.056681 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.91412 (* 0.0272727 = 0.0522034 loss) | |
| I0623 19:18:26.056694 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.26022 (* 0.0272727 = 0.0343696 loss) | |
| I0623 19:18:26.056710 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 1.15334 (* 0.0272727 = 0.0314548 loss) | |
| I0623 19:18:26.056722 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.795329 (* 0.0272727 = 0.0216908 loss) | |
| I0623 19:18:26.056736 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.539874 (* 0.0272727 = 0.0147238 loss) | |
| I0623 19:18:26.056751 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.314986 (* 0.0272727 = 0.00859053 loss) | |
| I0623 19:18:26.056763 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.0367135 (* 0.0272727 = 0.00100128 loss) | |
| I0623 19:18:26.056777 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.00411071 (* 0.0272727 = 0.00011211 loss) | |
| I0623 19:18:26.056792 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.000675559 (* 0.0272727 = 1.84243e-05 loss) | |
| I0623 19:18:26.056805 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 2.73898e-05 (* 0.0272727 = 7.46994e-07 loss) | |
| I0623 19:18:26.056819 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 1.27706e-05 (* 0.0272727 = 3.48289e-07 loss) | |
| I0623 19:18:26.056831 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.771429 | |
| I0623 19:18:26.056843 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 0.75 | |
| I0623 19:18:26.056855 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 0.875 | |
| I0623 19:18:26.056867 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 19:18:26.056879 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 1 | |
| I0623 19:18:26.056890 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 19:18:26.056902 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.875 | |
| I0623 19:18:26.056913 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.875 | |
| I0623 19:18:26.056926 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.75 | |
| I0623 19:18:26.056936 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.625 | |
| I0623 19:18:26.056948 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.875 | |
| I0623 19:18:26.056960 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.5 | |
| I0623 19:18:26.056972 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.875 | |
| I0623 19:18:26.056983 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.875 | |
| I0623 19:18:26.056995 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 0.5 | |
| I0623 19:18:26.057006 10365 solver.cpp:245] Train net output #113: loss3/accuracy15 = 0.875 | |
| I0623 19:18:26.057018 10365 solver.cpp:245] Train net output #114: loss3/accuracy16 = 0.875 | |
| I0623 19:18:26.057039 10365 solver.cpp:245] Train net output #115: loss3/accuracy17 = 0.875 | |
| I0623 19:18:26.057052 10365 solver.cpp:245] Train net output #116: loss3/accuracy18 = 1 | |
| I0623 19:18:26.057065 10365 solver.cpp:245] Train net output #117: loss3/accuracy19 = 1 | |
| I0623 19:18:26.057076 10365 solver.cpp:245] Train net output #118: loss3/accuracy20 = 1 | |
| I0623 19:18:26.057087 10365 solver.cpp:245] Train net output #119: loss3/accuracy21 = 1 | |
| I0623 19:18:26.057099 10365 solver.cpp:245] Train net output #120: loss3/accuracy22 = 1 | |
| I0623 19:18:26.057111 10365 solver.cpp:245] Train net output #121: loss3/accuracy_incl_empty = 0.863636 | |
| I0623 19:18:26.057122 10365 solver.cpp:245] Train net output #122: loss3/accuracy_top3 = 0.990476 | |
| I0623 19:18:26.057137 10365 solver.cpp:245] Train net output #123: loss3/cross_entropy_loss = 0.642197 (* 1 = 0.642197 loss) | |
| I0623 19:18:26.057149 10365 solver.cpp:245] Train net output #124: loss3/cross_entropy_loss_incl_empty = 0.387462 (* 1 = 0.387462 loss) | |
| I0623 19:18:26.057163 10365 solver.cpp:245] Train net output #125: loss3/loss01 = 0.746961 (* 0.0909091 = 0.0679056 loss) | |
| I0623 19:18:26.057178 10365 solver.cpp:245] Train net output #126: loss3/loss02 = 0.427255 (* 0.0909091 = 0.0388413 loss) | |
| I0623 19:18:26.057191 10365 solver.cpp:245] Train net output #127: loss3/loss03 = 0.254393 (* 0.0909091 = 0.0231267 loss) | |
| I0623 19:18:26.057204 10365 solver.cpp:245] Train net output #128: loss3/loss04 = 0.155302 (* 0.0909091 = 0.0141183 loss) | |
| I0623 19:18:26.057219 10365 solver.cpp:245] Train net output #129: loss3/loss05 = 0.295794 (* 0.0909091 = 0.0268904 loss) | |
| I0623 19:18:26.057235 10365 solver.cpp:245] Train net output #130: loss3/loss06 = 0.523211 (* 0.0909091 = 0.0475646 loss) | |
| I0623 19:18:26.057250 10365 solver.cpp:245] Train net output #131: loss3/loss07 = 0.73759 (* 0.0909091 = 0.0670536 loss) | |
| I0623 19:18:26.057262 10365 solver.cpp:245] Train net output #132: loss3/loss08 = 0.730804 (* 0.0909091 = 0.0664368 loss) | |
| I0623 19:18:26.057276 10365 solver.cpp:245] Train net output #133: loss3/loss09 = 0.993098 (* 0.0909091 = 0.0902816 loss) | |
| I0623 19:18:26.057291 10365 solver.cpp:245] Train net output #134: loss3/loss10 = 0.495618 (* 0.0909091 = 0.0450562 loss) | |
| I0623 19:18:26.057303 10365 solver.cpp:245] Train net output #135: loss3/loss11 = 0.959931 (* 0.0909091 = 0.0872664 loss) | |
| I0623 19:18:26.057317 10365 solver.cpp:245] Train net output #136: loss3/loss12 = 0.500809 (* 0.0909091 = 0.0455281 loss) | |
| I0623 19:18:26.057332 10365 solver.cpp:245] Train net output #137: loss3/loss13 = 0.615762 (* 0.0909091 = 0.0559784 loss) | |
| I0623 19:18:26.057344 10365 solver.cpp:245] Train net output #138: loss3/loss14 = 1.22065 (* 0.0909091 = 0.110968 loss) | |
| I0623 19:18:26.057361 10365 solver.cpp:245] Train net output #139: loss3/loss15 = 0.426821 (* 0.0909091 = 0.0388019 loss) | |
| I0623 19:18:26.057375 10365 solver.cpp:245] Train net output #140: loss3/loss16 = 0.256596 (* 0.0909091 = 0.0233269 loss) | |
| I0623 19:18:26.057389 10365 solver.cpp:245] Train net output #141: loss3/loss17 = 0.150697 (* 0.0909091 = 0.0136997 loss) | |
| I0623 19:18:26.057404 10365 solver.cpp:245] Train net output #142: loss3/loss18 = 0.0182152 (* 0.0909091 = 0.00165593 loss) | |
| I0623 19:18:26.057417 10365 solver.cpp:245] Train net output #143: loss3/loss19 = 0.000589617 (* 0.0909091 = 5.36015e-05 loss) | |
| I0623 19:18:26.057431 10365 solver.cpp:245] Train net output #144: loss3/loss20 = 0.000115198 (* 0.0909091 = 1.04726e-05 loss) | |
| I0623 19:18:26.057446 10365 solver.cpp:245] Train net output #145: loss3/loss21 = 2.76428e-05 (* 0.0909091 = 2.51298e-06 loss) | |
| I0623 19:18:26.057459 10365 solver.cpp:245] Train net output #146: loss3/loss22 = 3.15906e-06 (* 0.0909091 = 2.87187e-07 loss) | |
| I0623 19:18:26.057472 10365 solver.cpp:245] Train net output #147: total_accuracy = 0.375 | |
| I0623 19:18:26.057484 10365 solver.cpp:245] Train net output #148: total_accuracy_not_rec = 0.25 | |
| I0623 19:18:26.057512 10365 solver.cpp:245] Train net output #149: total_confidence = 0.157386 | |
| I0623 19:18:26.057524 10365 solver.cpp:245] Train net output #150: total_confidence_not_rec = 0.145633 | |
| I0623 19:18:26.057538 10365 sgd_solver.cpp:106] Iteration 23500, lr = 0.001 | |
| I0623 19:18:26.416527 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 51.0978 > 30) by scale factor 0.587109 | |
| I0623 19:18:48.653311 10365 sgd_solver.cpp:92] Gradient clipping: scaling down gradients (L2 norm 34.4378 > 30) by scale factor 0.871136 | |
| I0623 19:24:49.240486 10365 solver.cpp:229] Iteration 24000, loss = 4.42424 | |
| I0623 19:24:49.240617 10365 solver.cpp:245] Train net output #0: loss1/accuracy = 0.434783 | |
| I0623 19:24:49.240638 10365 solver.cpp:245] Train net output #1: loss1/accuracy01 = 0.625 | |
| I0623 19:24:49.240651 10365 solver.cpp:245] Train net output #2: loss1/accuracy02 = 0.625 | |
| I0623 19:24:49.240664 10365 solver.cpp:245] Train net output #3: loss1/accuracy03 = 0.625 | |
| I0623 19:24:49.240677 10365 solver.cpp:245] Train net output #4: loss1/accuracy04 = 0.625 | |
| I0623 19:24:49.240689 10365 solver.cpp:245] Train net output #5: loss1/accuracy05 = 0.25 | |
| I0623 19:24:49.240702 10365 solver.cpp:245] Train net output #6: loss1/accuracy06 = 0.375 | |
| I0623 19:24:49.240715 10365 solver.cpp:245] Train net output #7: loss1/accuracy07 = 0.5 | |
| I0623 19:24:49.240727 10365 solver.cpp:245] Train net output #8: loss1/accuracy08 = 0.375 | |
| I0623 19:24:49.240739 10365 solver.cpp:245] Train net output #9: loss1/accuracy09 = 0.375 | |
| I0623 19:24:49.240753 10365 solver.cpp:245] Train net output #10: loss1/accuracy10 = 0.25 | |
| I0623 19:24:49.240766 10365 solver.cpp:245] Train net output #11: loss1/accuracy11 = 0.25 | |
| I0623 19:24:49.240777 10365 solver.cpp:245] Train net output #12: loss1/accuracy12 = 0.5 | |
| I0623 19:24:49.240788 10365 solver.cpp:245] Train net output #13: loss1/accuracy13 = 0.625 | |
| I0623 19:24:49.240800 10365 solver.cpp:245] Train net output #14: loss1/accuracy14 = 0.75 | |
| I0623 19:24:49.240813 10365 solver.cpp:245] Train net output #15: loss1/accuracy15 = 0.875 | |
| I0623 19:24:49.240824 10365 solver.cpp:245] Train net output #16: loss1/accuracy16 = 0.75 | |
| I0623 19:24:49.240835 10365 solver.cpp:245] Train net output #17: loss1/accuracy17 = 0.875 | |
| I0623 19:24:49.240847 10365 solver.cpp:245] Train net output #18: loss1/accuracy18 = 0.875 | |
| I0623 19:24:49.240859 10365 solver.cpp:245] Train net output #19: loss1/accuracy19 = 1 | |
| I0623 19:24:49.240871 10365 solver.cpp:245] Train net output #20: loss1/accuracy20 = 1 | |
| I0623 19:24:49.240882 10365 solver.cpp:245] Train net output #21: loss1/accuracy21 = 1 | |
| I0623 19:24:49.240895 10365 solver.cpp:245] Train net output #22: loss1/accuracy22 = 1 | |
| I0623 19:24:49.240906 10365 solver.cpp:245] Train net output #23: loss1/accuracy_incl_empty = 0.681818 | |
| I0623 19:24:49.240918 10365 solver.cpp:245] Train net output #24: loss1/accuracy_top3 = 0.76087 | |
| I0623 19:24:49.240934 10365 solver.cpp:245] Train net output #25: loss1/cross_entropy_loss = 1.66097 (* 0.3 = 0.498292 loss) | |
| I0623 19:24:49.240948 10365 solver.cpp:245] Train net output #26: loss1/cross_entropy_loss_incl_empty = 0.964563 (* 0.3 = 0.289369 loss) | |
| I0623 19:24:49.240963 10365 solver.cpp:245] Train net output #27: loss1/loss01 = 0.837711 (* 0.0272727 = 0.0228467 loss) | |
| I0623 19:24:49.240978 10365 solver.cpp:245] Train net output #28: loss1/loss02 = 0.726464 (* 0.0272727 = 0.0198127 loss) | |
| I0623 19:24:49.240991 10365 solver.cpp:245] Train net output #29: loss1/loss03 = 1.04157 (* 0.0272727 = 0.0284064 loss) | |
| I0623 19:24:49.241005 10365 solver.cpp:245] Train net output #30: loss1/loss04 = 1.56231 (* 0.0272727 = 0.0426086 loss) | |
| I0623 19:24:49.241019 10365 solver.cpp:245] Train net output #31: loss1/loss05 = 1.97246 (* 0.0272727 = 0.0537944 loss) | |
| I0623 19:24:49.241034 10365 solver.cpp:245] Train net output #32: loss1/loss06 = 1.45372 (* 0.0272727 = 0.039647 loss) | |
| I0623 19:24:49.241047 10365 solver.cpp:245] Train net output #33: loss1/loss07 = 1.79345 (* 0.0272727 = 0.0489124 loss) | |
| I0623 19:24:49.241061 10365 solver.cpp:245] Train net output #34: loss1/loss08 = 1.93212 (* 0.0272727 = 0.0526942 loss) | |
| I0623 19:24:49.241075 10365 solver.cpp:245] Train net output #35: loss1/loss09 = 1.88003 (* 0.0272727 = 0.0512735 loss) | |
| I0623 19:24:49.241088 10365 solver.cpp:245] Train net output #36: loss1/loss10 = 2.8772 (* 0.0272727 = 0.078469 loss) | |
| I0623 19:24:49.241102 10365 solver.cpp:245] Train net output #37: loss1/loss11 = 1.9336 (* 0.0272727 = 0.0527345 loss) | |
| I0623 19:24:49.241116 10365 solver.cpp:245] Train net output #38: loss1/loss12 = 1.6189 (* 0.0272727 = 0.0441518 loss) | |
| I0623 19:24:49.241147 10365 solver.cpp:245] Train net output #39: loss1/loss13 = 1.5861 (* 0.0272727 = 0.0432574 loss) | |
| I0623 19:24:49.241163 10365 solver.cpp:245] Train net output #40: loss1/loss14 = 0.638677 (* 0.0272727 = 0.0174185 loss) | |
| I0623 19:24:49.241176 10365 solver.cpp:245] Train net output #41: loss1/loss15 = 0.513882 (* 0.0272727 = 0.014015 loss) | |
| I0623 19:24:49.241190 10365 solver.cpp:245] Train net output #42: loss1/loss16 = 0.716304 (* 0.0272727 = 0.0195356 loss) | |
| I0623 19:24:49.241204 10365 solver.cpp:245] Train net output #43: loss1/loss17 = 0.346612 (* 0.0272727 = 0.00945306 loss) | |
| I0623 19:24:49.241219 10365 solver.cpp:245] Train net output #44: loss1/loss18 = 0.312853 (* 0.0272727 = 0.00853237 loss) | |
| I0623 19:24:49.241232 10365 solver.cpp:245] Train net output #45: loss1/loss19 = 0.0522305 (* 0.0272727 = 0.00142447 loss) | |
| I0623 19:24:49.241246 10365 solver.cpp:245] Train net output #46: loss1/loss20 = 0.00682698 (* 0.0272727 = 0.00018619 loss) | |
| I0623 19:24:49.241263 10365 solver.cpp:245] Train net output #47: loss1/loss21 = 0.00339161 (* 0.0272727 = 9.24984e-05 loss) | |
| I0623 19:24:49.241278 10365 solver.cpp:245] Train net output #48: loss1/loss22 = 6.46573e-05 (* 0.0272727 = 1.76338e-06 loss) | |
| I0623 19:24:49.241291 10365 solver.cpp:245] Train net output #49: loss2/accuracy = 0.521739 | |
| I0623 19:24:49.241303 10365 solver.cpp:245] Train net output #50: loss2/accuracy01 = 0.875 | |
| I0623 19:24:49.241314 10365 solver.cpp:245] Train net output #51: loss2/accuracy02 = 1 | |
| I0623 19:24:49.241327 10365 solver.cpp:245] Train net output #52: loss2/accuracy03 = 1 | |
| I0623 19:24:49.241338 10365 solver.cpp:245] Train net output #53: loss2/accuracy04 = 0.375 | |
| I0623 19:24:49.241349 10365 solver.cpp:245] Train net output #54: loss2/accuracy05 = 0.5 | |
| I0623 19:24:49.241361 10365 solver.cpp:245] Train net output #55: loss2/accuracy06 = 0.375 | |
| I0623 19:24:49.241372 10365 solver.cpp:245] Train net output #56: loss2/accuracy07 = 0.5 | |
| I0623 19:24:49.241384 10365 solver.cpp:245] Train net output #57: loss2/accuracy08 = 0.75 | |
| I0623 19:24:49.241395 10365 solver.cpp:245] Train net output #58: loss2/accuracy09 = 0.375 | |
| I0623 19:24:49.241407 10365 solver.cpp:245] Train net output #59: loss2/accuracy10 = 0.25 | |
| I0623 19:24:49.241418 10365 solver.cpp:245] Train net output #60: loss2/accuracy11 = 0.375 | |
| I0623 19:24:49.241430 10365 solver.cpp:245] Train net output #61: loss2/accuracy12 = 0.75 | |
| I0623 19:24:49.241441 10365 solver.cpp:245] Train net output #62: loss2/accuracy13 = 0.75 | |
| I0623 19:24:49.241452 10365 solver.cpp:245] Train net output #63: loss2/accuracy14 = 0.75 | |
| I0623 19:24:49.241464 10365 solver.cpp:245] Train net output #64: loss2/accuracy15 = 0.75 | |
| I0623 19:24:49.241475 10365 solver.cpp:245] Train net output #65: loss2/accuracy16 = 0.75 | |
| I0623 19:24:49.241487 10365 solver.cpp:245] Train net output #66: loss2/accuracy17 = 0.875 | |
| I0623 19:24:49.241498 10365 solver.cpp:245] Train net output #67: loss2/accuracy18 = 0.875 | |
| I0623 19:24:49.241510 10365 solver.cpp:245] Train net output #68: loss2/accuracy19 = 1 | |
| I0623 19:24:49.241521 10365 solver.cpp:245] Train net output #69: loss2/accuracy20 = 1 | |
| I0623 19:24:49.241533 10365 solver.cpp:245] Train net output #70: loss2/accuracy21 = 1 | |
| I0623 19:24:49.241545 10365 solver.cpp:245] Train net output #71: loss2/accuracy22 = 1 | |
| I0623 19:24:49.241556 10365 solver.cpp:245] Train net output #72: loss2/accuracy_incl_empty = 0.721591 | |
| I0623 19:24:49.241569 10365 solver.cpp:245] Train net output #73: loss2/accuracy_top3 = 0.73913 | |
| I0623 19:24:49.241582 10365 solver.cpp:245] Train net output #74: loss2/cross_entropy_loss = 1.38263 (* 0.3 = 0.414788 loss) | |
| I0623 19:24:49.241595 10365 solver.cpp:245] Train net output #75: loss2/cross_entropy_loss_incl_empty = 0.834249 (* 0.3 = 0.250275 loss) | |
| I0623 19:24:49.241610 10365 solver.cpp:245] Train net output #76: loss2/loss01 = 0.430849 (* 0.0272727 = 0.0117504 loss) | |
| I0623 19:24:49.241623 10365 solver.cpp:245] Train net output #77: loss2/loss02 = 0.138123 (* 0.0272727 = 0.00376698 loss) | |
| I0623 19:24:49.241654 10365 solver.cpp:245] Train net output #78: loss2/loss03 = 0.219288 (* 0.0272727 = 0.0059806 loss) | |
| I0623 19:24:49.241670 10365 solver.cpp:245] Train net output #79: loss2/loss04 = 1.06938 (* 0.0272727 = 0.029165 loss) | |
| I0623 19:24:49.241684 10365 solver.cpp:245] Train net output #80: loss2/loss05 = 1.32321 (* 0.0272727 = 0.0360876 loss) | |
| I0623 19:24:49.241698 10365 solver.cpp:245] Train net output #81: loss2/loss06 = 1.79521 (* 0.0272727 = 0.0489603 loss) | |
| I0623 19:24:49.241711 10365 solver.cpp:245] Train net output #82: loss2/loss07 = 1.79414 (* 0.0272727 = 0.0489311 loss) | |
| I0623 19:24:49.241725 10365 solver.cpp:245] Train net output #83: loss2/loss08 = 1.39233 (* 0.0272727 = 0.0379726 loss) | |
| I0623 19:24:49.241739 10365 solver.cpp:245] Train net output #84: loss2/loss09 = 1.85248 (* 0.0272727 = 0.0505222 loss) | |
| I0623 19:24:49.241752 10365 solver.cpp:245] Train net output #85: loss2/loss10 = 2.55568 (* 0.0272727 = 0.0697005 loss) | |
| I0623 19:24:49.241766 10365 solver.cpp:245] Train net output #86: loss2/loss11 = 1.84348 (* 0.0272727 = 0.0502769 loss) | |
| I0623 19:24:49.241780 10365 solver.cpp:245] Train net output #87: loss2/loss12 = 1.22456 (* 0.0272727 = 0.0333971 loss) | |
| I0623 19:24:49.241793 10365 solver.cpp:245] Train net output #88: loss2/loss13 = 1.12343 (* 0.0272727 = 0.030639 loss) | |
| I0623 19:24:49.241807 10365 solver.cpp:245] Train net output #89: loss2/loss14 = 0.653864 (* 0.0272727 = 0.0178327 loss) | |
| I0623 19:24:49.241821 10365 solver.cpp:245] Train net output #90: loss2/loss15 = 0.670272 (* 0.0272727 = 0.0182802 loss) | |
| I0623 19:24:49.241834 10365 solver.cpp:245] Train net output #91: loss2/loss16 = 0.928457 (* 0.0272727 = 0.0253216 loss) | |
| I0623 19:24:49.241848 10365 solver.cpp:245] Train net output #92: loss2/loss17 = 0.414092 (* 0.0272727 = 0.0112934 loss) | |
| I0623 19:24:49.241863 10365 solver.cpp:245] Train net output #93: loss2/loss18 = 0.525441 (* 0.0272727 = 0.0143302 loss) | |
| I0623 19:24:49.241876 10365 solver.cpp:245] Train net output #94: loss2/loss19 = 0.0193555 (* 0.0272727 = 0.000527878 loss) | |
| I0623 19:24:49.241890 10365 solver.cpp:245] Train net output #95: loss2/loss20 = 0.00260787 (* 0.0272727 = 7.11237e-05 loss) | |
| I0623 19:24:49.241904 10365 solver.cpp:245] Train net output #96: loss2/loss21 = 0.000755905 (* 0.0272727 = 2.06156e-05 loss) | |
| I0623 19:24:49.241917 10365 solver.cpp:245] Train net output #97: loss2/loss22 = 0.000131481 (* 0.0272727 = 3.58584e-06 loss) | |
| I0623 19:24:49.241930 10365 solver.cpp:245] Train net output #98: loss3/accuracy = 0.804348 | |
| I0623 19:24:49.241942 10365 solver.cpp:245] Train net output #99: loss3/accuracy01 = 1 | |
| I0623 19:24:49.241955 10365 solver.cpp:245] Train net output #100: loss3/accuracy02 = 1 | |
| I0623 19:24:49.241966 10365 solver.cpp:245] Train net output #101: loss3/accuracy03 = 1 | |
| I0623 19:24:49.241977 10365 solver.cpp:245] Train net output #102: loss3/accuracy04 = 0.875 | |
| I0623 19:24:49.241989 10365 solver.cpp:245] Train net output #103: loss3/accuracy05 = 0.875 | |
| I0623 19:24:49.242000 10365 solver.cpp:245] Train net output #104: loss3/accuracy06 = 0.75 | |
| I0623 19:24:49.242012 10365 solver.cpp:245] Train net output #105: loss3/accuracy07 = 0.75 | |
| I0623 19:24:49.242024 10365 solver.cpp:245] Train net output #106: loss3/accuracy08 = 0.625 | |
| I0623 19:24:49.242035 10365 solver.cpp:245] Train net output #107: loss3/accuracy09 = 0.5 | |
| I0623 19:24:49.242048 10365 solver.cpp:245] Train net output #108: loss3/accuracy10 = 0.625 | |
| I0623 19:24:49.242058 10365 solver.cpp:245] Train net output #109: loss3/accuracy11 = 0.625 | |
| I0623 19:24:49.242070 10365 solver.cpp:245] Train net output #110: loss3/accuracy12 = 0.625 | |
| I0623 19:24:49.242081 10365 solver.cpp:245] Train net output #111: loss3/accuracy13 = 0.875 | |
| I0623 19:24:49.242094 10365 solver.cpp:245] Train net output #112: loss3/accuracy14 = 1 | |
| I0623 19:24:49.242 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment