Last active
July 9, 2018 05:11
-
-
Save bearpelican/4cb7f2c812c3c59d04f9692cfe56c2ba to your computer and use it in GitHub Desktop.
Training logs for 8 machines - p3.16xlarge. High learning rate warmup with batchnorm set to 0.
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
~~epoch hours top1Accuracy | |
Distributed: init_process_group success | |
Loaded model | |
Defined loss and optimizer | |
Created data loaders | |
Begin training | |
Changing LR from None to 1.5 | |
~~0 0.015269303611111111 8.536 | |
* Prec@1 2.754 Prec@5 8.536 | |
Changing LR from 1.7961783439490446 to 1.8 | |
~~1 0.023635661111111112 21.258 | |
* Prec@1 8.432 Prec@5 21.258 | |
Changing LR from 2.0961783439490445 to 2.1 | |
~~2 0.03173788527777778 37.690 | |
* Prec@1 17.334 Prec@5 37.690 | |
Changing LR from 2.3961783439490447 to 2.4000000000000004 | |
~~3 0.039703370277777775 33.122 | |
* Prec@1 14.496 Prec@5 33.122 | |
Changing LR from 2.6961783439490445 to 2.6999999999999997 | |
~~4 0.04787378527777778 50.588 | |
* Prec@1 26.480 Prec@5 50.588 | |
Changing LR from 2.996178343949045 to 2.0 | |
~~5 0.05568475555555556 59.004 | |
* Prec@1 33.008 Prec@5 59.004 | |
~~6 0.0639391811111111 53.928 | |
* Prec@1 29.774 Prec@5 53.928 | |
~~7 0.07236078416666668 51.474 | |
* Prec@1 27.276 Prec@5 51.474 | |
~~8 0.08073456555555555 60.892 | |
* Prec@1 35.056 Prec@5 60.892 | |
~~9 0.08893140777777778 50.422 | |
* Prec@1 26.872 Prec@5 50.422 | |
~~10 0.09690940777777778 65.224 | |
* Prec@1 38.542 Prec@5 65.224 | |
~~11 0.10507692777777779 52.642 | |
* Prec@1 29.386 Prec@5 52.642 | |
~~12 0.11343941944444444 67.358 | |
* Prec@1 41.262 Prec@5 67.358 | |
~~13 0.12170321777777778 61.786 | |
* Prec@1 35.736 Prec@5 61.786 | |
DataManager changing image size to 244 | |
~~14 0.13926569777777778 71.286 | |
* Prec@1 45.048 Prec@5 71.286 | |
~~15 0.15423443805555556 72.606 | |
* Prec@1 45.718 Prec@5 72.606 | |
~~16 0.16906809694444444 72.932 | |
* Prec@1 46.532 Prec@5 72.932 | |
Changing LR from 2.0 to 0.2 | |
~~17 0.18378840666666668 87.702 | |
* Prec@1 66.650 Prec@5 87.702 | |
~~18 0.19871511277777779 88.390 | |
* Prec@1 67.792 Prec@5 88.390 | |
~~19 0.21360281750000001 88.516 | |
* Prec@1 68.104 Prec@5 88.516 | |
~~20 0.22842486722222222 88.766 | |
* Prec@1 68.110 Prec@5 88.766 | |
~~21 0.24325293527777778 88.824 | |
* Prec@1 68.288 Prec@5 88.824 | |
~~22 0.25836048833333336 88.840 | |
* Prec@1 68.718 Prec@5 88.840 | |
~~23 0.27334701805555556 88.814 | |
* Prec@1 68.286 Prec@5 88.814 | |
~~24 0.28863768055555555 89.032 | |
* Prec@1 68.870 Prec@5 89.032 | |
~~25 0.3036785363888889 89.240 | |
* Prec@1 69.358 Prec@5 89.240 | |
~~26 0.3184671144444445 88.690 | |
* Prec@1 68.220 Prec@5 88.690 | |
~~27 0.3334224694444444 89.278 | |
* Prec@1 69.288 Prec@5 89.278 | |
~~28 0.34860874722222224 89.014 | |
* Prec@1 68.592 Prec@5 89.014 | |
~~29 0.3633554663888889 88.702 | |
* Prec@1 68.442 Prec@5 88.702 | |
Changing LR from 0.2 to 0.02 | |
~~30 0.37831854333333337 90.902 | |
* Prec@1 72.466 Prec@5 90.902 | |
~~31 0.39322078027777774 91.084 | |
* Prec@1 72.672 Prec@5 91.084 | |
~~32 0.4083793536111111 91.112 | |
* Prec@1 72.900 Prec@5 91.112 | |
~~33 0.42323898694444445 91.008 | |
* Prec@1 72.914 Prec@5 91.008 | |
~~34 0.4382319338888889 91.164 | |
* Prec@1 72.864 Prec@5 91.164 | |
DataManager changing image size to 288 | |
~~35 0.47201463000000005 92.836 | |
* Prec@1 75.322 Prec@5 92.836 | |
~~36 0.4919849927777778 92.966 | |
* Prec@1 75.368 Prec@5 92.966 | |
~~37 0.512011628888889 93.072 | |
* Prec@1 75.484 Prec@5 93.072 | |
Changing LR from 0.02 to 0.002 | |
~~38 0.5319756361111111 93.102 | |
* Prec@1 75.562 Prec@5 93.102 | |
~~39 0.5515782158333333 93.090 | |
* Prec@1 75.596 Prec@5 93.090 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment