tree attn multinli

tree-attn-multinli.log

17-03-26 23:12:58 [1] Flag Values:
{
    "?": null,
    "actively_decay_learning_rate": true,
    "attention_dim": 128,
    "batch_size": 32,
    "branch_name": "mrdrozdov/tree-attn",
    "bucket_eval": true,
    "ckpt_interval_steps": 5000,
    "ckpt_on_best_dev_error": true,
    "ckpt_path": "/home/dexter/logs/spinn",
    "ckpt_step": 1000,
    "clipping_max_value": 5.0,
    "data_type": "snli",
    "debug": false,
    "deque_length": null,
    "embedding_data_path": "/home/dexter/data/glove/glove.840B.300d.txt",
    "embedding_keep_rate": 0.9,
    "encode_bidirectional": false,
    "encode_num_layers": 1,
    "encode_reverse": false,
    "encode_style": null,
    "eval_data_limit": -1,
    "eval_data_path": "/home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl",
    "eval_interval_steps": 500,
    "eval_report_use_preds": true,
    "eval_seq_length": null,
    "evalb": false,
    "expanded_eval_only_mode": false,
    "experiment_name": "spinn-multisnli-attn-eid_04",
    "gen_h": true,
    "gpu": 0,
    "help": null,
    "helpshort": null,
    "helpxml": null,
    "init_range": 0.005,
    "l2_lambda": 2.75e-05,
    "lateral_tracking": true,
    "learning_rate": 0.0003,
    "learning_rate_decay_per_10k_steps": 0.75,
    "load_best": false,
    "log_path": "/home/dexter/logs/spinn",
    "lowercase": false,
    "metrics_interval_steps": 10,
    "metrics_path": "/home/dexter/logs/spinn-runs",
    "mlp_bn": true,
    "mlp_dim": 1024,
    "model_dim": 600,
    "model_type": "SPINN",
    "num_mlp_layers": 2,
    "num_samples": 0,
    "optimizer_type": "RMSprop",
    "predict_leaf": true,
    "predict_use_cell": true,
    "rl_baseline": "ema",
    "rl_entropy": false,
    "rl_entropy_beta": 0.001,
    "rl_epsilon": 1.0,
    "rl_epsilon_decay": 50000.0,
    "rl_mu": 0.1,
    "rl_reward": "standard",
    "rl_weight": 1.0,
    "rl_whiten": false,
    "semantic_classifier_keep_rate": 0.9,
    "seq_length": 500,
    "sha": "3e4f3c0218bb66b52175deb154db7aeb169ea6bb",
    "show_progress_bar": true,
    "shuffle_eval": false,
    "shuffle_eval_seed": 123,
    "smart_batching": true,
    "statistics_interval_steps": 100,
    "tracking_lstm_hidden_dim": 40,
    "training_data_path": "/home/dexter/data/multinli_0.2/multinli_0.2_train.jsonl",
    "training_steps": 250000,
    "transition_weight": 0.6,
    "use_attention": true,
    "use_difference_feature": true,
    "use_encode": false,
    "use_internal_parser": true,
    "use_l2_cost": true,
    "use_lengths": false,
    "use_peano": true,
    "use_product_feature": true,
    "use_tracking_in_composition": true,
    "validate_transitions": true,
    "word_embedding_dim": 300,
    "write_eval_report": false
}
17-03-26 23:13:24 [1] In open vocabulary mode. Using loaded embeddings without fine-tuning.
17-03-26 23:13:24 [1] Constructing vocabulary...
17-03-26 23:13:25 [1] Found 89840 word types.
17-03-26 23:13:39 [1] Loading vocabulary with 79321 words from /home/dexter/data/glove/glove.840B.300d.txt
17-03-26 23:14:10 [1] Preprocessing training data.
17-03-26 23:15:00 [1] Preprocessing eval data: /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl
17-03-26 23:15:12 [1] Building model.
17-03-26 23:15:13 [1] Architecture: BaseModel (
  (spinn): SPINN (
    (reduce): Reduce (
      (left): CustomLinear (300 -> 1500)
      (right): CustomLinear (300 -> 1500)
      (track): CustomLinear (40 -> 1500)
    )
    (tracker): Tracker (
      (buf): CustomLinear (300 -> 160)
      (stack1): CustomLinear (300 -> 160)
      (stack2): CustomLinear (300 -> 160)
      (lateral): CustomLinear (40 -> 160)
    )
    (transition_net): Linear (80 -> 2)
  )
  (mlp): MLP (
    (bn_inp): BatchNorm1d(1200, eps=1e-05, momentum=0.1, affine=True)
    (l0): CustomLinear (1200 -> 1024)
    (bn0): BatchNorm1d(1024, eps=1e-05, momentum=0.1, affine=True)
    (l1): CustomLinear (1024 -> 1024)
    (bn1): BatchNorm1d(1024, eps=1e-05, momentum=0.1, affine=True)
    (l2): CustomLinear (1024 -> 3)
  )
  (embed): Embed (
    (projection): Linear (300 -> 600)
  )
  (attention_keys): CustomLinear (300 -> 128)
  (attention_query): CustomLinear (40 -> 128)
  (attention_mix): Linear (600 -> 300)
)
17-03-26 23:15:13 [1] Total params: 3805765.0
17-03-26 23:15:18 [1]
17-03-26 23:15:18 [1] # ----- BEGIN: Log Configuration ----- #
17-03-26 23:15:18 [1] Train-Format: Step: {step} Acc: {class_acc:.5f} {transition_acc:.5f} Cost: {total_cost:.5f} {xent_cost:.5f} {transition_cost:.5f} {l2_cost:.5f} Time: {time:.5f}
17-03-26 23:15:18 [1] Train-Extra-Format: Train Extra: lr={learning_rate:.7f} inv={invalid:.7f} sub={struct:.7f}
17-03-26 23:15:18 [1] Eval-Format: Step: {step} Eval acc: {class_acc:.5f} {transition_acc:.5f} {filename} Time: {time:.5f}
17-03-26 23:15:18 [1] Eval-Extra-Format: Eval Extra: inv={inv:.7f}
17-03-26 23:15:18 [1] # ----- END: Log Configuration ----- #
17-03-26 23:15:18 [1]
17-03-26 23:15:18 [1] Training.
17-03-26 23:15:19 [1] Step: 0 Acc: 0.37500 0.69837 Cost: 1.92000 1.27875 0.41662 0.22463 Time: 0.00045
17-03-26 23:15:19 [1] Train Extra: lr=0.0003000 inv=0.9843750 sub=0.0000000
17-03-26 23:17:06 [1] Step: 100 Acc: 0.40844 0.76095 Cost: 1.79488 1.18069 0.38868 0.22551 Time: 0.00091
17-03-26 23:17:06 [1] Train Extra: lr=0.0002991 inv=0.7179688 sub=0.0000000
17-03-26 23:18:35 [1] Step: 200 Acc: 0.42094 0.76245 Cost: 1.64252 1.06373 0.35483 0.22396 Time: 0.00084
17-03-26 23:18:35 [1] Train Extra: lr=0.0002983 inv=0.4087500 sub=0.0000000
17-03-26 23:19:55 [1] Step: 300 Acc: 0.43812 0.76168 Cost: 1.68262 1.16200 0.29840 0.22222 Time: 0.00080
17-03-26 23:19:55 [1] Train Extra: lr=0.0002974 inv=0.4989062 sub=0.0000000
17-03-26 23:21:23 [1] Step: 400 Acc: 0.43719 0.76573 Cost: 1.70363 1.13186 0.35151 0.22026 Time: 0.00082
17-03-26 23:21:23 [1] Train Extra: lr=0.0002966 inv=0.5176563 sub=0.0000000
17-03-26 23:23:12 [1] Step: 500 Acc: 0.45781 0.78855 Cost: 1.78751 1.20929 0.36023 0.21799 Time: 0.00092
17-03-26 23:23:12 [1] Train Extra: lr=0.0002957 inv=0.5781250 sub=0.0000000
17-03-26 23:24:19 [1] Step: 500 Eval acc: 0.47937 0.77666 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-26 23:24:19 [1] Eval Extra: inv=0.5402369
17-03-26 23:25:49 [1] Step: 600 Acc: 0.45937 0.77700 Cost: 1.59304 1.02498 0.35264 0.21541 Time: 0.00086
17-03-26 23:25:49 [1] Train Extra: lr=0.0002949 inv=0.5998438 sub=0.0000000
17-03-26 23:27:16 [1] Step: 700 Acc: 0.47687 0.77447 Cost: 1.71280 1.14358 0.35656 0.21266 Time: 0.00084
17-03-26 23:27:16 [1] Train Extra: lr=0.0002940 inv=0.6054688 sub=0.0000000
17-03-26 23:28:51 [1] Step: 800 Acc: 0.46656 0.77781 Cost: 1.45826 0.93971 0.30893 0.20963 Time: 0.00084
17-03-26 23:28:51 [1] Train Extra: lr=0.0002932 inv=0.6282813 sub=0.0000000
17-03-26 23:30:16 [1] Step: 900 Acc: 0.47656 0.77166 Cost: 1.49145 0.95946 0.32550 0.20648 Time: 0.00083
17-03-26 23:30:16 [1] Train Extra: lr=0.0002923 inv=0.6100000 sub=0.0000000
17-03-26 23:31:44 [1] Step: 1000 Acc: 0.46563 0.77809 Cost: 1.62421 1.07939 0.34169 0.20314 Time: 0.00084
17-03-26 23:31:44 [1] Train Extra: lr=0.0002915 inv=0.6456250 sub=0.0000000
17-03-26 23:32:54 [1] Step: 1000 Eval acc: 0.46181 0.78233 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-26 23:32:54 [1] Eval Extra: inv=0.6400123
17-03-26 23:34:25 [1] Step: 1100 Acc: 0.47594 0.78853 Cost: 1.45417 0.94265 0.31196 0.19955 Time: 0.00084
17-03-26 23:34:25 [1] Train Extra: lr=0.0002907 inv=0.6514062 sub=0.0000000
17-03-26 23:36:04 [1] Step: 1200 Acc: 0.47687 0.79699 Cost: 1.50034 1.04392 0.26051 0.19591 Time: 0.00092
17-03-26 23:36:04 [1] Train Extra: lr=0.0002898 inv=0.6371875 sub=0.0000000
17-03-26 23:37:31 [1] Step: 1300 Acc: 0.49094 0.77480 Cost: 1.65202 1.15363 0.30608 0.19231 Time: 0.00083
17-03-26 23:37:31 [1] Train Extra: lr=0.0002890 inv=0.6506250 sub=0.0000000
17-03-26 23:39:13 [1] Step: 1400 Acc: 0.50062 0.79705 Cost: 1.35542 0.85983 0.30671 0.18888 Time: 0.00090
17-03-26 23:39:13 [1] Train Extra: lr=0.0002882 inv=0.6668750 sub=0.0000000
17-03-26 23:40:48 [1] Step: 1500 Acc: 0.49938 0.79309 Cost: 1.64015 1.15314 0.30156 0.18545 Time: 0.00086
17-03-26 23:40:48 [1] Train Extra: lr=0.0002873 inv=0.6865625 sub=0.0000000
17-03-26 23:41:54 [1] Step: 1500 Eval acc: 0.51093 0.79871 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-26 23:41:54 [1] Eval Extra: inv=0.6516544
17-03-26 23:41:54 [1] Checkpointing with new best dev accuracy of 0.510927
17-03-26 23:43:25 [1] Step: 1600 Acc: 0.49375 0.79873 Cost: 1.48404 0.98808 0.31395 0.18201 Time: 0.00089
17-03-26 23:43:25 [1] Train Extra: lr=0.0002865 inv=0.7018750 sub=0.0000000
17-03-26 23:44:54 [1] Step: 1700 Acc: 0.50844 0.79319 Cost: 1.30830 0.89533 0.23439 0.17858 Time: 0.00085
17-03-26 23:44:54 [1] Train Extra: lr=0.0002857 inv=0.7090625 sub=0.0000000
17-03-26 23:46:21 [1] Step: 1800 Acc: 0.51313 0.79540 Cost: 1.48180 1.01798 0.28858 0.17524 Time: 0.00085
17-03-26 23:46:21 [1] Train Extra: lr=0.0002849 inv=0.7260937 sub=0.0000000
17-03-26 23:47:54 [1] Step: 1900 Acc: 0.50906 0.80755 Cost: 1.41914 0.93908 0.30805 0.17202 Time: 0.00090
17-03-26 23:47:54 [1] Train Extra: lr=0.0002840 inv=0.7206250 sub=0.0000000
17-03-26 23:49:15 [1] Step: 2000 Acc: 0.51625 0.79306 Cost: 1.23804 0.83542 0.23377 0.16885 Time: 0.00081
17-03-26 23:49:15 [1] Train Extra: lr=0.0002832 inv=0.6956250 sub=0.0000000
17-03-26 23:50:26 [1] Step: 2000 Eval acc: 0.55249 0.80891 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-26 23:50:26 [1] Eval Extra: inv=0.6656965
17-03-26 23:50:26 [1] Checkpointing with new best dev accuracy of 0.552492
17-03-26 23:51:53 [1] Step: 2100 Acc: 0.50719 0.79725 Cost: 1.43621 0.95947 0.31092 0.16581 Time: 0.00084
17-03-26 23:51:53 [1] Train Extra: lr=0.0002824 inv=0.6515625 sub=0.0000000
17-03-26 23:53:25 [1] Step: 2200 Acc: 0.50219 0.79653 Cost: 1.67968 1.25989 0.25695 0.16284 Time: 0.00085
17-03-26 23:53:25 [1] Train Extra: lr=0.0002816 inv=0.6118750 sub=0.0000000
17-03-26 23:55:03 [1] Step: 2300 Acc: 0.50406 0.80785 Cost: 1.32052 0.87869 0.28189 0.15994 Time: 0.00087
17-03-26 23:55:03 [1] Train Extra: lr=0.0002808 inv=0.6110937 sub=0.0000000
17-03-26 23:56:28 [1] Step: 2400 Acc: 0.52531 0.80152 Cost: 1.42830 0.94309 0.32805 0.15716 Time: 0.00083
17-03-26 23:56:28 [1] Train Extra: lr=0.0002800 inv=0.5784375 sub=0.0000000
17-03-26 23:57:54 [1] Step: 2500 Acc: 0.51719 0.81279 Cost: 1.29288 0.95534 0.18310 0.15444 Time: 0.00085
17-03-26 23:57:54 [1] Train Extra: lr=0.0002792 inv=0.5240625 sub=0.0000000
17-03-26 23:59:04 [1] Step: 2500 Eval acc: 0.54841 0.82250 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-26 23:59:04 [1] Eval Extra: inv=0.4863664
17-03-27 00:00:34 [1] Step: 2600 Acc: 0.50781 0.82277 Cost: 1.44379 1.02793 0.26402 0.15184 Time: 0.00085
17-03-27 00:00:34 [1] Train Extra: lr=0.0002784 inv=0.5100000 sub=0.0000000
17-03-27 00:02:08 [1] Step: 2700 Acc: 0.50469 0.80108 Cost: 1.25434 0.88634 0.21845 0.14955 Time: 0.00084
17-03-27 00:02:08 [1] Train Extra: lr=0.0002776 inv=0.5457812 sub=0.0000000
17-03-27 00:03:35 [1] Step: 2800 Acc: 0.51812 0.80962 Cost: 1.19550 0.84949 0.19890 0.14711 Time: 0.00084
17-03-27 00:03:35 [1] Train Extra: lr=0.0002768 inv=0.4839062 sub=0.0000000
17-03-27 00:05:01 [1] Step: 2900 Acc: 0.53187 0.80632 Cost: 1.43468 1.00733 0.28257 0.14479 Time: 0.00084
17-03-27 00:05:01 [1] Train Extra: lr=0.0002760 inv=0.4879688 sub=0.0000000
17-03-27 00:06:34 [1] Step: 3000 Acc: 0.51656 0.82924 Cost: 1.35418 0.88582 0.32584 0.14252 Time: 0.00090
17-03-27 00:06:34 [1] Train Extra: lr=0.0002752 inv=0.4675000 sub=0.0000000
17-03-27 00:07:42 [1] Step: 3000 Eval acc: 0.54984 0.81693 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 00:07:42 [1] Eval Extra: inv=0.4538909
17-03-27 00:09:07 [1] Step: 3100 Acc: 0.52781 0.81514 Cost: 1.30468 0.87671 0.28783 0.14014 Time: 0.00085
17-03-27 00:09:07 [1] Train Extra: lr=0.0002744 inv=0.4732812 sub=0.0000000
17-03-27 00:10:32 [1] Step: 3200 Acc: 0.53969 0.81446 Cost: 1.39455 1.05854 0.19819 0.13782 Time: 0.00085
17-03-27 00:10:32 [1] Train Extra: lr=0.0002736 inv=0.4440625 sub=0.0000000
17-03-27 00:12:02 [1] Step: 3300 Acc: 0.53063 0.80893 Cost: 1.42092 0.98096 0.30434 0.13562 Time: 0.00083
17-03-27 00:12:02 [1] Train Extra: lr=0.0002728 inv=0.4914062 sub=0.0000000
17-03-27 00:13:36 [1] Step: 3400 Acc: 0.53375 0.82419 Cost: 1.52599 1.06023 0.33215 0.13361 Time: 0.00088
17-03-27 00:13:36 [1] Train Extra: lr=0.0002720 inv=0.4395312 sub=0.0000000
17-03-27 00:15:10 [1] Step: 3500 Acc: 0.52687 0.82538 Cost: 1.46170 1.06632 0.26374 0.13164 Time: 0.00088
17-03-27 00:15:10 [1] Train Extra: lr=0.0002713 inv=0.4606250 sub=0.0000000
17-03-27 00:16:16 [1] Step: 3500 Eval acc: 0.55750 0.82897 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 00:16:16 [1] Eval Extra: inv=0.4727839
17-03-27 00:16:16 [1] Checkpointing with new best dev accuracy of 0.557496
17-03-27 00:17:53 [1] Step: 3600 Acc: 0.53125 0.82111 Cost: 1.37947 0.99675 0.25294 0.12978 Time: 0.00088
17-03-27 00:17:53 [1] Train Extra: lr=0.0002705 inv=0.4556250 sub=0.0000000
17-03-27 00:19:26 [1] Step: 3700 Acc: 0.53875 0.82094 Cost: 1.26574 0.87169 0.26605 0.12800 Time: 0.00087
17-03-27 00:19:26 [1] Train Extra: lr=0.0002697 inv=0.4523437 sub=0.0000000
17-03-27 00:20:57 [1] Step: 3800 Acc: 0.54406 0.81714 Cost: 1.32154 0.89473 0.30064 0.12618 Time: 0.00087
17-03-27 00:20:57 [1] Train Extra: lr=0.0002689 inv=0.4303125 sub=0.0000000
17-03-27 00:22:31 [1] Step: 3900 Acc: 0.52938 0.81257 Cost: 1.28414 0.85846 0.30136 0.12432 Time: 0.00086
17-03-27 00:22:31 [1] Train Extra: lr=0.0002682 inv=0.4737500 sub=0.0000000
17-03-27 00:24:01 [1] Step: 4000 Acc: 0.54469 0.81835 Cost: 1.55567 1.13532 0.29766 0.12268 Time: 0.00084
17-03-27 00:24:01 [1] Train Extra: lr=0.0002674 inv=0.4660937 sub=0.0000000
17-03-27 00:25:14 [1] Step: 4000 Eval acc: 0.56883 0.82730 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 00:25:14 [1] Eval Extra: inv=0.4518484
17-03-27 00:25:14 [1] Checkpointing with new best dev accuracy of 0.568832
17-03-27 00:26:44 [1] Step: 4100 Acc: 0.52312 0.81800 Cost: 1.09692 0.86144 0.11449 0.12098 Time: 0.00085
17-03-27 00:26:44 [1] Train Extra: lr=0.0002666 inv=0.4670313 sub=0.0000000
17-03-27 00:28:24 [1] Step: 4200 Acc: 0.54688 0.82684 Cost: 1.55940 1.04785 0.39217 0.11938 Time: 0.00091
17-03-27 00:28:24 [1] Train Extra: lr=0.0002659 inv=0.4664063 sub=0.0000000
17-03-27 00:29:52 [1] Step: 4300 Acc: 0.53312 0.82111 Cost: 1.31647 0.95856 0.24004 0.11787 Time: 0.00085
17-03-27 00:29:52 [1] Train Extra: lr=0.0002651 inv=0.4632812 sub=0.0000000
17-03-27 00:31:18 [1] Step: 4400 Acc: 0.53125 0.81998 Cost: 1.41965 0.97782 0.32544 0.11639 Time: 0.00084
17-03-27 00:31:18 [1] Train Extra: lr=0.0002643 inv=0.4421875 sub=0.0000000
17-03-27 00:32:57 [1] Step: 4500 Acc: 0.52312 0.82104 Cost: 1.34084 1.06645 0.15955 0.11484 Time: 0.00085
17-03-27 00:32:57 [1] Train Extra: lr=0.0002636 inv=0.4637500 sub=0.0000000
17-03-27 00:34:08 [1] Step: 4500 Eval acc: 0.55862 0.83021 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 00:34:08 [1] Eval Extra: inv=0.4114583
17-03-27 00:35:41 [1] Step: 4600 Acc: 0.53312 0.82773 Cost: 1.32014 0.96626 0.24057 0.11331 Time: 0.00087
17-03-27 00:35:41 [1] Train Extra: lr=0.0002628 inv=0.4704687 sub=0.0000000
17-03-27 00:37:08 [1] Step: 4700 Acc: 0.55531 0.82410 Cost: 1.30808 0.93776 0.25837 0.11195 Time: 0.00081
17-03-27 00:37:08 [1] Train Extra: lr=0.0002621 inv=0.4381250 sub=0.0000000
17-03-27 00:38:44 [1] Step: 4800 Acc: 0.54000 0.83663 Cost: 1.21335 0.83044 0.27232 0.11059 Time: 0.00090
17-03-27 00:38:44 [1] Train Extra: lr=0.0002613 inv=0.4623437 sub=0.0000000
17-03-27 00:40:10 [1] Step: 4900 Acc: 0.54906 0.81461 Cost: 1.16319 0.82800 0.22582 0.10937 Time: 0.00081
17-03-27 00:40:10 [1] Train Extra: lr=0.0002606 inv=0.4562500 sub=0.0000000
17-03-27 00:41:37 [1] Step: 5000 Acc: 0.53469 0.82764 Cost: 1.41416 1.04030 0.26576 0.10810 Time: 0.00085
17-03-27 00:41:37 [1] Train Extra: lr=0.0002598 inv=0.4167188 sub=0.0000000
17-03-27 00:42:46 [1] Step: 5000 Eval acc: 0.55607 0.83141 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 00:42:46 [1] Eval Extra: inv=0.4300449
17-03-27 00:42:46 [1] Checkpointing.
17-03-27 00:44:19 [1] Step: 5100 Acc: 0.55500 0.83020 Cost: 1.20718 0.93964 0.16070 0.10684 Time: 0.00087
17-03-27 00:44:19 [1] Train Extra: lr=0.0002591 inv=0.4412500 sub=0.0000000
17-03-27 00:45:46 [1] Step: 5200 Acc: 0.56063 0.82344 Cost: 1.21540 0.84552 0.26406 0.10582 Time: 0.00084
17-03-27 00:45:46 [1] Train Extra: lr=0.0002583 inv=0.4207812 sub=0.0000000
17-03-27 00:47:07 [1] Step: 5300 Acc: 0.56500 0.83389 Cost: 1.32252 1.00038 0.21731 0.10483 Time: 0.00081
17-03-27 00:47:07 [1] Train Extra: lr=0.0002576 inv=0.4062500 sub=0.0000000
17-03-27 00:48:39 [1] Step: 5400 Acc: 0.55375 0.82044 Cost: 1.22988 0.88052 0.24561 0.10375 Time: 0.00085
17-03-27 00:48:39 [1] Train Extra: lr=0.0002568 inv=0.4414062 sub=0.0000000
17-03-27 00:50:04 [1] Step: 5500 Acc: 0.55094 0.82208 Cost: 1.24423 0.98836 0.15315 0.10272 Time: 0.00084
17-03-27 00:50:04 [1] Train Extra: lr=0.0002561 inv=0.4446875 sub=0.0000000
17-03-27 00:51:13 [1] Step: 5500 Eval acc: 0.55556 0.82903 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 00:51:13 [1] Eval Extra: inv=0.4460784
17-03-27 00:52:39 [1] Step: 5600 Acc: 0.53687 0.81768 Cost: 1.45089 1.12261 0.22662 0.10167 Time: 0.00080
17-03-27 00:52:39 [1] Train Extra: lr=0.0002554 inv=0.4603125 sub=0.0000000
17-03-27 00:54:05 [1] Step: 5700 Acc: 0.55812 0.82074 Cost: 1.27704 0.90229 0.27405 0.10070 Time: 0.00081
17-03-27 00:54:05 [1] Train Extra: lr=0.0002546 inv=0.4479688 sub=0.0000000
17-03-27 00:55:35 [1] Step: 5800 Acc: 0.54969 0.82817 Cost: 1.30312 0.90465 0.29877 0.09970 Time: 0.00085
17-03-27 00:55:35 [1] Train Extra: lr=0.0002539 inv=0.4518750 sub=0.0000000
17-03-27 00:57:01 [1] Step: 5900 Acc: 0.53406 0.81926 Cost: 1.28822 0.96476 0.22471 0.09875 Time: 0.00080
17-03-27 00:57:01 [1] Train Extra: lr=0.0002532 inv=0.4420312 sub=0.0000000
17-03-27 00:58:31 [1] Step: 6000 Acc: 0.56125 0.82472 Cost: 1.25943 0.87206 0.28947 0.09790 Time: 0.00084
17-03-27 00:58:31 [1] Train Extra: lr=0.0002524 inv=0.4492188 sub=0.0000000
17-03-27 00:59:36 [1] Step: 6000 Eval acc: 0.56965 0.83831 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-27 00:59:36 [1] Eval Extra: inv=0.4226409
17-03-27 01:01:10 [1] Step: 6100 Acc: 0.54063 0.81796 Cost: 1.35555 0.93755 0.32095 0.09705 Time: 0.00082
17-03-27 01:01:10 [1] Train Extra: lr=0.0002517 inv=0.4654687 sub=0.0000000
17-03-27 01:02:42 [1] Step: 6200 Acc: 0.57031 0.82969 Cost: 1.48710 1.09744 0.29357 0.09608 Time: 0.00088
17-03-27 01:02:42 [1] Train Extra: lr=0.0002510 inv=0.4042188 sub=0.0000000
17-03-27 01:04:02 [1] Step: 6300 Acc: 0.56125 0.82773 Cost: 1.02412 0.77619 0.15253 0.09539 Time: 0.00079
17-03-27 01:04:02 [1] Train Extra: lr=0.0002503 inv=0.4159375 sub=0.0000000
17-03-27 01:05:32 [1] Step: 6400 Acc: 0.56969 0.82596 Cost: 1.19627 0.85179 0.24994 0.09455 Time: 0.00086
17-03-27 01:05:32 [1] Train Extra: lr=0.0002496 inv=0.4423437 sub=0.0000000
17-03-27 01:06:58 [1] Step: 6500 Acc: 0.55594 0.82442 Cost: 1.36422 0.95346 0.31702 0.09373 Time: 0.00083
17-03-27 01:06:58 [1] Train Extra: lr=0.0002488 inv=0.4006250 sub=0.0000000
17-03-27 01:08:03 [1] Step: 6500 Eval acc: 0.57864 0.83961 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 01:08:03 [1] Eval Extra: inv=0.4214665
17-03-27 01:08:03 [1] Checkpointing with new best dev accuracy of 0.578636
17-03-27 01:09:36 [1] Step: 6600 Acc: 0.55656 0.82261 Cost: 1.55326 1.33294 0.12745 0.09287 Time: 0.00083
17-03-27 01:09:36 [1] Train Extra: lr=0.0002481 inv=0.4162500 sub=0.0000000
17-03-27 01:11:00 [1] Step: 6700 Acc: 0.54688 0.82711 Cost: 1.36603 0.93480 0.33916 0.09208 Time: 0.00082
17-03-27 01:11:00 [1] Train Extra: lr=0.0002474 inv=0.4035937 sub=0.0000000
17-03-27 01:12:33 [1] Step: 6800 Acc: 0.55875 0.82389 Cost: 1.47652 1.11585 0.26933 0.09135 Time: 0.00083
17-03-27 01:12:33 [1] Train Extra: lr=0.0002467 inv=0.4075000 sub=0.0000000
17-03-27 01:14:12 [1] Step: 6900 Acc: 0.55844 0.83611 Cost: 1.10285 0.79546 0.21673 0.09067 Time: 0.00088
17-03-27 01:14:12 [1] Train Extra: lr=0.0002460 inv=0.4109375 sub=0.0000000
17-03-27 01:15:34 [1] Step: 7000 Acc: 0.57219 0.83019 Cost: 1.07977 0.75498 0.23480 0.08999 Time: 0.00085
17-03-27 01:15:34 [1] Train Extra: lr=0.0002453 inv=0.3853125 sub=0.0000000
17-03-27 01:16:39 [1] Step: 7000 Eval acc: 0.57128 0.83904 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-27 01:16:39 [1] Eval Extra: inv=0.4097733
17-03-27 01:18:18 [1] Step: 7100 Acc: 0.55000 0.83282 Cost: 1.06796 0.80214 0.17654 0.08928 Time: 0.00087
17-03-27 01:18:18 [1] Train Extra: lr=0.0002446 inv=0.4275000 sub=0.0000000
17-03-27 01:19:50 [1] Step: 7200 Acc: 0.55812 0.82591 Cost: 1.36419 1.03706 0.23849 0.08864 Time: 0.00084
17-03-27 01:19:50 [1] Train Extra: lr=0.0002439 inv=0.4214062 sub=0.0000000
17-03-27 01:21:22 [1] Step: 7300 Acc: 0.55844 0.82911 Cost: 1.05008 0.82560 0.13652 0.08795 Time: 0.00084
17-03-27 01:21:22 [1] Train Extra: lr=0.0002432 inv=0.4410938 sub=0.0000000
17-03-27 01:22:50 [1] Step: 7400 Acc: 0.56969 0.82546 Cost: 1.10783 0.85504 0.16541 0.08738 Time: 0.00080
17-03-27 01:22:50 [1] Train Extra: lr=0.0002425 inv=0.4315625 sub=0.0000000
17-03-27 01:24:14 [1] Step: 7500 Acc: 0.57656 0.83045 Cost: 1.17671 0.91478 0.17519 0.08674 Time: 0.00082
17-03-27 01:24:14 [1] Train Extra: lr=0.0002418 inv=0.4140625 sub=0.0000000
17-03-27 01:25:23 [1] Step: 7500 Eval acc: 0.57251 0.82540 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 01:25:23 [1] Eval Extra: inv=0.4391850
17-03-27 01:27:01 [1] Step: 7600 Acc: 0.55094 0.82938 Cost: 1.13709 0.75484 0.29608 0.08617 Time: 0.00085
17-03-27 01:27:01 [1] Train Extra: lr=0.0002411 inv=0.4515625 sub=0.0000000
17-03-27 01:28:28 [1] Step: 7700 Acc: 0.56844 0.83794 Cost: 1.08326 0.74761 0.25008 0.08557 Time: 0.00083
17-03-27 01:28:28 [1] Train Extra: lr=0.0002404 inv=0.3981250 sub=0.0000000
17-03-27 01:29:54 [1] Step: 7800 Acc: 0.55937 0.82955 Cost: 1.43338 1.10049 0.24792 0.08498 Time: 0.00083
17-03-27 01:29:54 [1] Train Extra: lr=0.0002397 inv=0.4134375 sub=0.0000000
17-03-27 01:31:28 [1] Step: 7900 Acc: 0.55844 0.82841 Cost: 1.20482 0.84612 0.27422 0.08448 Time: 0.00085
17-03-27 01:31:28 [1] Train Extra: lr=0.0002390 inv=0.4337500 sub=0.0000000
17-03-27 01:33:01 [1] Step: 8000 Acc: 0.55563 0.82412 Cost: 1.38970 1.04559 0.26012 0.08399 Time: 0.00082
17-03-27 01:33:01 [1] Train Extra: lr=0.0002383 inv=0.4376563 sub=0.0000000
17-03-27 01:34:10 [1] Step: 8000 Eval acc: 0.58619 0.83558 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 01:34:10 [1] Eval Extra: inv=0.3851103
17-03-27 01:34:10 [1] Checkpointing with new best dev accuracy of 0.586193
17-03-27 01:35:37 [1] Step: 8100 Acc: 0.56812 0.82346 Cost: 1.26292 0.94366 0.23575 0.08351 Time: 0.00081
17-03-27 01:35:37 [1] Train Extra: lr=0.0002376 inv=0.4165625 sub=0.0000000
17-03-27 01:37:00 [1] Step: 8200 Acc: 0.56188 0.83255 Cost: 0.91391 0.72318 0.10769 0.08304 Time: 0.00084
17-03-27 01:37:00 [1] Train Extra: lr=0.0002370 inv=0.4000000 sub=0.0000000
17-03-27 01:38:24 [1] Step: 8300 Acc: 0.59094 0.82732 Cost: 1.00757 0.73458 0.19034 0.08264 Time: 0.00081
17-03-27 01:38:24 [1] Train Extra: lr=0.0002363 inv=0.4246875 sub=0.0000000
17-03-27 01:39:56 [1] Step: 8400 Acc: 0.57500 0.82808 Cost: 1.33627 0.95804 0.29597 0.08227 Time: 0.00083
17-03-27 01:39:56 [1] Train Extra: lr=0.0002356 inv=0.4375000 sub=0.0000000
17-03-27 01:41:23 [1] Step: 8500 Acc: 0.55563 0.83439 Cost: 1.40788 1.09116 0.23493 0.08178 Time: 0.00083
17-03-27 01:41:23 [1] Train Extra: lr=0.0002349 inv=0.4073438 sub=0.0000000
17-03-27 01:42:32 [1] Step: 8500 Eval acc: 0.59007 0.83836 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 01:42:32 [1] Eval Extra: inv=0.3995098
17-03-27 01:44:05 [1] Step: 8600 Acc: 0.55375 0.82297 Cost: 1.13974 0.83465 0.22378 0.08131 Time: 0.00081
17-03-27 01:44:05 [1] Train Extra: lr=0.0002342 inv=0.4326563 sub=0.0000000
17-03-27 01:45:36 [1] Step: 8700 Acc: 0.57656 0.83497 Cost: 1.35595 0.88444 0.39056 0.08095 Time: 0.00086
17-03-27 01:45:36 [1] Train Extra: lr=0.0002336 inv=0.3965625 sub=0.0000000
17-03-27 01:46:56 [1] Step: 8800 Acc: 0.57500 0.83093 Cost: 1.26395 1.04273 0.14070 0.08052 Time: 0.00079
17-03-27 01:46:56 [1] Train Extra: lr=0.0002329 inv=0.3971875 sub=0.0000000
17-03-27 01:48:27 [1] Step: 8900 Acc: 0.58469 0.83445 Cost: 1.11027 0.81305 0.21715 0.08007 Time: 0.00086
17-03-27 01:48:27 [1] Train Extra: lr=0.0002322 inv=0.4053125 sub=0.0000000
17-03-27 01:49:53 [1] Step: 9000 Acc: 0.58719 0.82668 Cost: 1.09751 0.85313 0.16476 0.07963 Time: 0.00080
17-03-27 01:49:53 [1] Train Extra: lr=0.0002316 inv=0.4253125 sub=0.0000000
17-03-27 01:51:02 [1] Step: 9000 Eval acc: 0.59538 0.83095 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 01:51:02 [1] Eval Extra: inv=0.4081393
17-03-27 01:51:02 [1] Checkpointing with new best dev accuracy of 0.595384
17-03-27 01:52:36 [1] Step: 9100 Acc: 0.58844 0.83268 Cost: 1.26974 1.04421 0.14621 0.07932 Time: 0.00084
17-03-27 01:52:36 [1] Train Extra: lr=0.0002309 inv=0.4162500 sub=0.0000000
17-03-27 01:54:01 [1] Step: 9200 Acc: 0.56219 0.83660 Cost: 1.09971 0.77026 0.25058 0.07887 Time: 0.00085
17-03-27 01:54:01 [1] Train Extra: lr=0.0002302 inv=0.3870312 sub=0.0000000
17-03-27 01:55:51 [1] Step: 9300 Acc: 0.56500 0.84434 Cost: 1.31463 0.97088 0.26533 0.07843 Time: 0.00090
17-03-27 01:55:51 [1] Train Extra: lr=0.0002296 inv=0.4303125 sub=0.0000000
17-03-27 01:57:22 [1] Step: 9400 Acc: 0.57344 0.83152 Cost: 1.17776 0.92112 0.17866 0.07798 Time: 0.00085
17-03-27 01:57:22 [1] Train Extra: lr=0.0002289 inv=0.4132812 sub=0.0000000
17-03-27 01:58:48 [1] Step: 9500 Acc: 0.55625 0.84321 Cost: 1.20588 0.93284 0.19552 0.07752 Time: 0.00085
17-03-27 01:58:48 [1] Train Extra: lr=0.0002283 inv=0.3832813 sub=0.0000000
17-03-27 01:59:57 [1] Step: 9500 Eval acc: 0.59630 0.83628 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 01:59:57 [1] Eval Extra: inv=0.4546058
17-03-27 02:01:30 [1] Step: 9600 Acc: 0.58250 0.83905 Cost: 1.38063 1.07663 0.22680 0.07720 Time: 0.00086
17-03-27 02:01:30 [1] Train Extra: lr=0.0002276 inv=0.4159375 sub=0.0000000
17-03-27 02:03:06 [1] Step: 9700 Acc: 0.55625 0.83587 Cost: 1.16238 0.82418 0.26135 0.07685 Time: 0.00084
17-03-27 02:03:06 [1] Train Extra: lr=0.0002270 inv=0.4117188 sub=0.0000000
17-03-27 02:04:48 [1] Step: 9800 Acc: 0.57500 0.83415 Cost: 1.30693 0.95178 0.27869 0.07645 Time: 0.00091
17-03-27 02:04:48 [1] Train Extra: lr=0.0002263 inv=0.4239062 sub=0.0000000
17-03-27 02:06:15 [1] Step: 9900 Acc: 0.59156 0.83560 Cost: 1.23820 0.92714 0.23496 0.07609 Time: 0.00083
17-03-27 02:06:15 [1] Train Extra: lr=0.0002256 inv=0.4025000 sub=0.0000000
17-03-27 02:07:39 [1] Step: 10000 Acc: 0.58813 0.83324 Cost: 1.25532 0.87409 0.30550 0.07573 Time: 0.00084
17-03-27 02:07:39 [1] Train Extra: lr=0.0002250 inv=0.4129687 sub=0.0000000
17-03-27 02:08:46 [1] Step: 10000 Eval acc: 0.59375 0.84253 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 02:08:46 [1] Eval Extra: inv=0.4448019
17-03-27 02:08:46 [1] Checkpointing.
17-03-27 02:10:24 [1] Step: 10100 Acc: 0.58500 0.83450 Cost: 1.29906 1.00131 0.22223 0.07552 Time: 0.00088
17-03-27 02:10:24 [1] Train Extra: lr=0.0002244 inv=0.4414062 sub=0.0000000
17-03-27 02:11:51 [1] Step: 10200 Acc: 0.57312 0.84674 Cost: 1.26362 0.93109 0.25726 0.07528 Time: 0.00087
17-03-27 02:11:51 [1] Train Extra: lr=0.0002237 inv=0.3581250 sub=0.0000000
17-03-27 02:13:27 [1] Step: 10300 Acc: 0.59156 0.83607 Cost: 1.25930 0.89784 0.28644 0.07502 Time: 0.00086
17-03-27 02:13:27 [1] Train Extra: lr=0.0002231 inv=0.4203125 sub=0.0000000
17-03-27 02:15:02 [1] Step: 10400 Acc: 0.60313 0.82911 Cost: 1.31022 0.95479 0.28066 0.07477 Time: 0.00085
17-03-27 02:15:02 [1] Train Extra: lr=0.0002224 inv=0.4100000 sub=0.0000000
17-03-27 02:16:20 [1] Step: 10500 Acc: 0.60219 0.83393 Cost: 1.14833 0.80609 0.26772 0.07452 Time: 0.00082
17-03-27 02:16:20 [1] Train Extra: lr=0.0002218 inv=0.3715625 sub=0.0000000
17-03-27 02:17:30 [1] Step: 10500 Eval acc: 0.59732 0.83874 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 02:17:30 [1] Eval Extra: inv=0.3980801
17-03-27 02:19:05 [1] Step: 10600 Acc: 0.57125 0.84039 Cost: 1.02581 0.83579 0.11572 0.07429 Time: 0.00086
17-03-27 02:19:05 [1] Train Extra: lr=0.0002211 inv=0.4154687 sub=0.0000000
17-03-27 02:20:33 [1] Step: 10700 Acc: 0.58844 0.82832 Cost: 1.17383 0.76938 0.33047 0.07398 Time: 0.00081
17-03-27 02:20:33 [1] Train Extra: lr=0.0002205 inv=0.4235937 sub=0.0000000
17-03-27 02:22:13 [1] Step: 10800 Acc: 0.58719 0.83428 Cost: 1.53067 1.14267 0.31426 0.07374 Time: 0.00088
17-03-27 02:22:13 [1] Train Extra: lr=0.0002199 inv=0.3993750 sub=0.0000000
17-03-27 02:23:39 [1] Step: 10900 Acc: 0.57250 0.82693 Cost: 1.26489 0.88642 0.30482 0.07365 Time: 0.00081
17-03-27 02:23:39 [1] Train Extra: lr=0.0002192 inv=0.4110937 sub=0.0000000
17-03-27 02:25:07 [1] Step: 11000 Acc: 0.57625 0.83465 Cost: 1.21203 0.88755 0.25111 0.07338 Time: 0.00084
17-03-27 02:25:07 [1] Train Extra: lr=0.0002186 inv=0.4020313 sub=0.0000000
17-03-27 02:26:18 [1] Step: 11000 Eval acc: 0.60243 0.83449 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 02:26:18 [1] Eval Extra: inv=0.4081393
17-03-27 02:26:18 [1] Checkpointing with new best dev accuracy of 0.602431
17-03-27 02:27:50 [1] Step: 11100 Acc: 0.57625 0.82556 Cost: 1.05350 0.85290 0.12743 0.07317 Time: 0.00084
17-03-27 02:27:50 [1] Train Extra: lr=0.0002180 inv=0.4314062 sub=0.0000000
17-03-27 02:29:26 [1] Step: 11200 Acc: 0.58281 0.84135 Cost: 1.13704 0.92230 0.14182 0.07293 Time: 0.00085
17-03-27 02:29:26 [1] Train Extra: lr=0.0002174 inv=0.4134375 sub=0.0000000
17-03-27 02:30:51 [1] Step: 11300 Acc: 0.58469 0.83655 Cost: 1.21956 0.90721 0.23968 0.07267 Time: 0.00083
17-03-27 02:30:51 [1] Train Extra: lr=0.0002167 inv=0.4060937 sub=0.0000000
17-03-27 02:32:26 [1] Step: 11400 Acc: 0.58375 0.83791 Cost: 1.24281 0.91064 0.25970 0.07247 Time: 0.00086
17-03-27 02:32:26 [1] Train Extra: lr=0.0002161 inv=0.4112500 sub=0.0000000
17-03-27 02:34:09 [1] Step: 11500 Acc: 0.56312 0.82871 Cost: 1.19996 0.92035 0.20737 0.07224 Time: 0.00085
17-03-27 02:34:09 [1] Train Extra: lr=0.0002155 inv=0.4539063 sub=0.0000000
17-03-27 02:35:19 [1] Step: 11500 Eval acc: 0.60886 0.84506 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 02:35:19 [1] Eval Extra: inv=0.3298611
17-03-27 02:35:19 [1] Checkpointing with new best dev accuracy of 0.608864
17-03-27 02:36:53 [1] Step: 11600 Acc: 0.57812 0.83279 Cost: 0.99649 0.76782 0.15672 0.07196 Time: 0.00084
17-03-27 02:36:53 [1] Train Extra: lr=0.0002149 inv=0.4012500 sub=0.0000000
17-03-27 02:38:26 [1] Step: 11700 Acc: 0.58688 0.82586 Cost: 1.07039 0.79953 0.19908 0.07178 Time: 0.00084
17-03-27 02:38:26 [1] Train Extra: lr=0.0002143 inv=0.4078125 sub=0.0000000
17-03-27 02:39:46 [1] Step: 11800 Acc: 0.58594 0.83980 Cost: 1.20309 0.92333 0.20817 0.07159 Time: 0.00081
17-03-27 02:39:46 [1] Train Extra: lr=0.0002136 inv=0.3960938 sub=0.0000000
17-03-27 02:41:22 [1] Step: 11900 Acc: 0.58125 0.83461 Cost: 1.19254 0.89731 0.22384 0.07138 Time: 0.00086
17-03-27 02:41:22 [1] Train Extra: lr=0.0002130 inv=0.4412500 sub=0.0000000
17-03-27 02:43:03 [1] Step: 12000 Acc: 0.58344 0.84074 Cost: 1.13906 0.79727 0.27057 0.07122 Time: 0.00087
17-03-27 02:43:03 [1] Train Extra: lr=0.0002124 inv=0.4182812 sub=0.0000000
17-03-27 02:44:13 [1] Step: 12000 Eval acc: 0.60948 0.84541 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 02:44:13 [1] Eval Extra: inv=0.3278186
17-03-27 02:45:52 [1] Step: 12100 Acc: 0.57188 0.83813 Cost: 0.96281 0.72146 0.17033 0.07102 Time: 0.00084
17-03-27 02:45:52 [1] Train Extra: lr=0.0002118 inv=0.4192187 sub=0.0000000
17-03-27 02:47:18 [1] Step: 12200 Acc: 0.59656 0.83837 Cost: 1.17401 0.87709 0.22601 0.07091 Time: 0.00085
17-03-27 02:47:18 [1] Train Extra: lr=0.0002112 inv=0.4051562 sub=0.0000000
17-03-27 02:48:51 [1] Step: 12300 Acc: 0.59813 0.84145 Cost: 1.20889 0.88334 0.25489 0.07066 Time: 0.00088
17-03-27 02:48:51 [1] Train Extra: lr=0.0002106 inv=0.3982812 sub=0.0000000
17-03-27 02:50:16 [1] Step: 12400 Acc: 0.57969 0.83232 Cost: 1.16823 0.98430 0.11351 0.07043 Time: 0.00082
17-03-27 02:50:16 [1] Train Extra: lr=0.0002100 inv=0.3878125 sub=0.0000000
17-03-27 02:51:43 [1] Step: 12500 Acc: 0.58469 0.83900 Cost: 1.18914 0.89950 0.21944 0.07020 Time: 0.00084
17-03-27 02:51:43 [1] Train Extra: lr=0.0002094 inv=0.3832813 sub=0.0000000
17-03-27 02:52:54 [1] Step: 12500 Eval acc: 0.60958 0.83682 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 02:52:54 [1] Eval Extra: inv=0.4094669
17-03-27 02:54:25 [1] Step: 12600 Acc: 0.59875 0.83992 Cost: 1.09940 0.81513 0.21422 0.07006 Time: 0.00088
17-03-27 02:54:25 [1] Train Extra: lr=0.0002088 inv=0.3807813 sub=0.0000000
17-03-27 02:55:45 [1] Step: 12700 Acc: 0.58781 0.83862 Cost: 1.27573 0.98101 0.22475 0.06997 Time: 0.00081
17-03-27 02:55:45 [1] Train Extra: lr=0.0002082 inv=0.3698438 sub=0.0000000
17-03-27 02:57:27 [1] Step: 12800 Acc: 0.59094 0.84142 Cost: 1.28367 1.01510 0.19882 0.06975 Time: 0.00089
17-03-27 02:57:27 [1] Train Extra: lr=0.0002076 inv=0.3826562 sub=0.0000000
17-03-27 02:58:52 [1] Step: 12900 Acc: 0.59875 0.83580 Cost: 1.37824 1.09754 0.21110 0.06961 Time: 0.00084
17-03-27 02:58:52 [1] Train Extra: lr=0.0002070 inv=0.3810938 sub=0.0000000
17-03-27 03:00:35 [1] Step: 13000 Acc: 0.55906 0.84374 Cost: 1.08050 0.80631 0.20477 0.06942 Time: 0.00090
17-03-27 03:00:35 [1] Train Extra: lr=0.0002064 inv=0.4007812 sub=0.0000000
17-03-27 03:01:45 [1] Step: 13000 Eval acc: 0.61040 0.84796 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 03:01:45 [1] Eval Extra: inv=0.3889910
17-03-27 03:03:10 [1] Step: 13100 Acc: 0.59281 0.83197 Cost: 0.93606 0.77190 0.09495 0.06921 Time: 0.00081
17-03-27 03:03:10 [1] Train Extra: lr=0.0002058 inv=0.4104687 sub=0.0000000
17-03-27 03:04:32 [1] Step: 13200 Acc: 0.59719 0.83812 Cost: 1.10344 0.91721 0.11720 0.06902 Time: 0.00079
17-03-27 03:04:32 [1] Train Extra: lr=0.0002052 inv=0.3676563 sub=0.0000000
17-03-27 03:06:02 [1] Step: 13300 Acc: 0.59250 0.84116 Cost: 1.01733 0.76325 0.18522 0.06887 Time: 0.00088
17-03-27 03:06:02 [1] Train Extra: lr=0.0002046 inv=0.3701563 sub=0.0000000
17-03-27 03:07:29 [1] Step: 13400 Acc: 0.59719 0.84391 Cost: 1.13198 0.91440 0.14888 0.06870 Time: 0.00086
17-03-27 03:07:29 [1] Train Extra: lr=0.0002040 inv=0.3693750 sub=0.0000000
17-03-27 03:09:09 [1] Step: 13500 Acc: 0.58875 0.84099 Cost: 1.26711 0.89092 0.30764 0.06855 Time: 0.00090
17-03-27 03:09:09 [1] Train Extra: lr=0.0002034 inv=0.3881250 sub=0.0000000
17-03-27 03:10:16 [1] Step: 13500 Eval acc: 0.61080 0.84713 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 03:10:16 [1] Eval Extra: inv=0.4101307
17-03-27 03:11:57 [1] Step: 13600 Acc: 0.59281 0.84278 Cost: 1.33657 1.00600 0.26218 0.06839 Time: 0.00089
17-03-27 03:11:57 [1] Train Extra: lr=0.0002029 inv=0.3934375 sub=0.0000000
17-03-27 03:13:33 [1] Step: 13700 Acc: 0.58688 0.83798 Cost: 1.29340 0.95596 0.26914 0.06829 Time: 0.00085
17-03-27 03:13:33 [1] Train Extra: lr=0.0002023 inv=0.4095313 sub=0.0000000
17-03-27 03:15:07 [1] Step: 13800 Acc: 0.59781 0.83980 Cost: 1.41414 1.12943 0.21659 0.06812 Time: 0.00087
17-03-27 03:15:07 [1] Train Extra: lr=0.0002017 inv=0.4045313 sub=0.0000000
17-03-27 03:16:39 [1] Step: 13900 Acc: 0.58688 0.84175 Cost: 0.93469 0.70609 0.16060 0.06800 Time: 0.00086
17-03-27 03:16:39 [1] Train Extra: lr=0.0002011 inv=0.3818750 sub=0.0000000
17-03-27 03:18:10 [1] Step: 14000 Acc: 0.58906 0.82915 Cost: 1.27625 0.93876 0.26968 0.06781 Time: 0.00083
17-03-27 03:18:10 [1] Train Extra: lr=0.0002005 inv=0.4345312 sub=0.0000000
17-03-27 03:19:18 [1] Step: 14000 Eval acc: 0.61305 0.84890 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 03:19:18 [1] Eval Extra: inv=0.3619792
17-03-27 03:19:18 [1] Checkpointing with new best dev accuracy of 0.613051
17-03-27 03:20:37 [1] Step: 14100 Acc: 0.59500 0.84367 Cost: 1.37824 0.97916 0.33141 0.06767 Time: 0.00083
17-03-27 03:20:37 [1] Train Extra: lr=0.0002000 inv=0.3965625 sub=0.0000000
17-03-27 03:22:15 [1] Step: 14200 Acc: 0.58937 0.83522 Cost: 1.15125 0.84215 0.24154 0.06756 Time: 0.00087
17-03-27 03:22:15 [1] Train Extra: lr=0.0001994 inv=0.4185937 sub=0.0000000
17-03-27 03:23:42 [1] Step: 14300 Acc: 0.60219 0.83627 Cost: 1.19567 0.87488 0.25330 0.06748 Time: 0.00082
17-03-27 03:23:42 [1] Train Extra: lr=0.0001988 inv=0.3776563 sub=0.0000000
17-03-27 03:25:07 [1] Step: 14400 Acc: 0.58344 0.83910 Cost: 1.06982 0.84280 0.15961 0.06741 Time: 0.00084
17-03-27 03:25:07 [1] Train Extra: lr=0.0001982 inv=0.3734375 sub=0.0000000
17-03-27 03:26:34 [1] Step: 14500 Acc: 0.60750 0.83349 Cost: 1.06539 0.75483 0.24333 0.06723 Time: 0.00082
17-03-27 03:26:34 [1] Train Extra: lr=0.0001977 inv=0.3745312 sub=0.0000000
17-03-27 03:27:41 [1] Step: 14500 Eval acc: 0.62163 0.84686 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 03:27:41 [1] Eval Extra: inv=0.3577921
17-03-27 03:27:41 [1] Checkpointing with new best dev accuracy of 0.621630
17-03-27 03:29:07 [1] Step: 14600 Acc: 0.59719 0.84259 Cost: 1.31901 1.03670 0.21515 0.06716 Time: 0.00084
17-03-27 03:29:07 [1] Train Extra: lr=0.0001971 inv=0.3776563 sub=0.0000000
17-03-27 03:30:30 [1] Step: 14700 Acc: 0.59562 0.85023 Cost: 1.55549 1.18569 0.30274 0.06706 Time: 0.00087
17-03-27 03:30:30 [1] Train Extra: lr=0.0001965 inv=0.3657813 sub=0.0000000
17-03-27 03:32:03 [1] Step: 14800 Acc: 0.60219 0.82928 Cost: 1.29300 1.01047 0.21561 0.06693 Time: 0.00084
17-03-27 03:32:03 [1] Train Extra: lr=0.0001960 inv=0.4292187 sub=0.0000000
17-03-27 03:33:37 [1] Step: 14900 Acc: 0.59281 0.84617 Cost: 1.18334 0.77766 0.33890 0.06678 Time: 0.00089
17-03-27 03:33:37 [1] Train Extra: lr=0.0001954 inv=0.3564062 sub=0.0000000
17-03-27 03:35:12 [1] Step: 15000 Acc: 0.57625 0.84005 Cost: 1.28570 0.97458 0.24444 0.06668 Time: 0.00086
17-03-27 03:35:12 [1] Train Extra: lr=0.0001949 inv=0.4103125 sub=0.0000000
17-03-27 03:36:19 [1] Step: 15000 Eval acc: 0.62214 0.84313 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 03:36:19 [1] Eval Extra: inv=0.4219771
17-03-27 03:36:19 [1] Checkpointing.
17-03-27 03:38:01 [1] Step: 15100 Acc: 0.59406 0.84931 Cost: 1.61466 1.16835 0.37975 0.06656 Time: 0.00090
17-03-27 03:38:01 [1] Train Extra: lr=0.0001943 inv=0.3960938 sub=0.0000000
17-03-27 03:39:33 [1] Step: 15200 Acc: 0.59562 0.84367 Cost: 1.38445 1.02366 0.29433 0.06646 Time: 0.00087
17-03-27 03:39:33 [1] Train Extra: lr=0.0001937 inv=0.3810938 sub=0.0000000
17-03-27 03:40:53 [1] Step: 15300 Acc: 0.58031 0.83975 Cost: 1.23126 0.99470 0.17019 0.06637 Time: 0.00082
17-03-27 03:40:53 [1] Train Extra: lr=0.0001932 inv=0.3639062 sub=0.0000000
17-03-27 03:42:40 [1] Step: 15400 Acc: 0.59719 0.83841 Cost: 1.14270 0.77403 0.30248 0.06619 Time: 0.00088
17-03-27 03:42:40 [1] Train Extra: lr=0.0001926 inv=0.4310937 sub=0.0000000
17-03-27 03:44:13 [1] Step: 15500 Acc: 0.58156 0.82750 Cost: 0.91019 0.72417 0.11994 0.06607 Time: 0.00083
17-03-27 03:44:13 [1] Train Extra: lr=0.0001921 inv=0.3979687 sub=0.0000000
17-03-27 03:45:24 [1] Step: 15500 Eval acc: 0.61846 0.84547 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 03:45:24 [1] Eval Extra: inv=0.3200061
17-03-27 03:46:57 [1] Step: 15600 Acc: 0.58281 0.84354 Cost: 1.29716 0.98057 0.25061 0.06598 Time: 0.00086
17-03-27 03:46:57 [1] Train Extra: lr=0.0001915 inv=0.3787500 sub=0.0000000
17-03-27 03:48:31 [1] Step: 15700 Acc: 0.59188 0.83072 Cost: 1.37678 1.11470 0.19623 0.06585 Time: 0.00083
17-03-27 03:48:31 [1] Train Extra: lr=0.0001910 inv=0.4260937 sub=0.0000000
17-03-27 03:50:00 [1] Step: 15800 Acc: 0.58781 0.83446 Cost: 1.29450 0.98261 0.24613 0.06575 Time: 0.00083
17-03-27 03:50:00 [1] Train Extra: lr=0.0001904 inv=0.3895312 sub=0.0000000
17-03-27 03:51:23 [1] Step: 15900 Acc: 0.59031 0.83830 Cost: 1.16191 0.93924 0.15696 0.06571 Time: 0.00086
17-03-27 03:51:23 [1] Train Extra: lr=0.0001899 inv=0.3746875 sub=0.0000000
17-03-27 03:52:52 [1] Step: 16000 Acc: 0.59156 0.83754 Cost: 0.78015 0.61077 0.10384 0.06554 Time: 0.00083
17-03-27 03:52:52 [1] Train Extra: lr=0.0001893 inv=0.3993750 sub=0.0000000
17-03-27 03:54:02 [1] Step: 16000 Eval acc: 0.62275 0.84985 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 03:54:02 [1] Eval Extra: inv=0.3295037
17-03-27 03:55:23 [1] Step: 16100 Acc: 0.58750 0.83808 Cost: 0.89814 0.64366 0.18903 0.06544 Time: 0.00080
17-03-27 03:55:23 [1] Train Extra: lr=0.0001888 inv=0.3900000 sub=0.0000000
17-03-27 03:56:48 [1] Step: 16200 Acc: 0.60031 0.84173 Cost: 1.30215 0.99049 0.24634 0.06533 Time: 0.00083
17-03-27 03:56:48 [1] Train Extra: lr=0.0001882 inv=0.4045313 sub=0.0000000
17-03-27 03:58:27 [1] Step: 16300 Acc: 0.59437 0.84156 Cost: 1.01678 0.79724 0.15431 0.06523 Time: 0.00088
17-03-27 03:58:27 [1] Train Extra: lr=0.0001877 inv=0.3970313 sub=0.0000000
17-03-27 04:00:02 [1] Step: 16400 Acc: 0.58094 0.83530 Cost: 1.19905 1.02416 0.10975 0.06515 Time: 0.00084
17-03-27 04:00:02 [1] Train Extra: lr=0.0001872 inv=0.4150000 sub=0.0000000
17-03-27 04:01:36 [1] Step: 16500 Acc: 0.60906 0.84544 Cost: 1.39973 1.05148 0.28314 0.06511 Time: 0.00087
17-03-27 04:01:36 [1] Train Extra: lr=0.0001866 inv=0.4140625 sub=0.0000000
17-03-27 04:02:48 [1] Step: 16500 Eval acc: 0.62316 0.84735 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 04:02:48 [1] Eval Extra: inv=0.3743873
17-03-27 04:04:09 [1] Step: 16600 Acc: 0.59688 0.83461 Cost: 1.09842 0.86698 0.16639 0.06505 Time: 0.00080
17-03-27 04:04:09 [1] Train Extra: lr=0.0001861 inv=0.3606250 sub=0.0000000
17-03-27 04:05:40 [1] Step: 16700 Acc: 0.59000 0.84047 Cost: 1.24964 0.91480 0.26989 0.06495 Time: 0.00088
17-03-27 04:05:40 [1] Train Extra: lr=0.0001856 inv=0.3723437 sub=0.0000000
17-03-27 04:07:09 [1] Step: 16800 Acc: 0.60094 0.83674 Cost: 1.27560 0.97150 0.23927 0.06483 Time: 0.00081
17-03-27 04:07:09 [1] Train Extra: lr=0.0001850 inv=0.4100000 sub=0.0000000
17-03-27 04:08:47 [1] Step: 16900 Acc: 0.59094 0.84467 Cost: 0.99956 0.77009 0.16474 0.06473 Time: 0.00090
17-03-27 04:08:47 [1] Train Extra: lr=0.0001845 inv=0.3920312 sub=0.0000000
17-03-27 04:10:22 [1] Step: 17000 Acc: 0.58562 0.83790 Cost: 1.17610 0.87079 0.24074 0.06457 Time: 0.00083
17-03-27 04:10:22 [1] Train Extra: lr=0.0001840 inv=0.4406250 sub=0.0000000
17-03-27 04:11:32 [1] Step: 17000 Eval acc: 0.62173 0.84647 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 04:11:32 [1] Eval Extra: inv=0.3752042
17-03-27 04:13:05 [1] Step: 17100 Acc: 0.59000 0.84336 Cost: 1.27699 0.97979 0.23272 0.06448 Time: 0.00087
17-03-27 04:13:05 [1] Train Extra: lr=0.0001834 inv=0.3931250 sub=0.0000000
17-03-27 04:14:32 [1] Step: 17200 Acc: 0.61594 0.83508 Cost: 0.99433 0.77450 0.15547 0.06435 Time: 0.00081
17-03-27 04:14:32 [1] Train Extra: lr=0.0001829 inv=0.4128125 sub=0.0000000
17-03-27 04:16:05 [1] Step: 17300 Acc: 0.60156 0.84262 Cost: 1.36747 1.00953 0.29353 0.06441 Time: 0.00087
17-03-27 04:16:05 [1] Train Extra: lr=0.0001824 inv=0.3835938 sub=0.0000000
17-03-27 04:17:34 [1] Step: 17400 Acc: 0.59469 0.84755 Cost: 1.35093 0.98428 0.30229 0.06435 Time: 0.00085
17-03-27 04:17:34 [1] Train Extra: lr=0.0001819 inv=0.3787500 sub=0.0000000
17-03-27 04:19:05 [1] Step: 17500 Acc: 0.58156 0.83535 Cost: 1.05875 0.87142 0.12316 0.06417 Time: 0.00085
17-03-27 04:19:05 [1] Train Extra: lr=0.0001813 inv=0.4110937 sub=0.0000000
17-03-27 04:20:12 [1] Step: 17500 Eval acc: 0.61928 0.84515 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 04:20:12 [1] Eval Extra: inv=0.3744894
17-03-27 04:21:39 [1] Step: 17600 Acc: 0.59437 0.83748 Cost: 1.25276 0.92766 0.26094 0.06415 Time: 0.00083
17-03-27 04:21:39 [1] Train Extra: lr=0.0001808 inv=0.4075000 sub=0.0000000
17-03-27 04:23:11 [1] Step: 17700 Acc: 0.60313 0.84792 Cost: 1.08096 0.72066 0.29621 0.06410 Time: 0.00089
17-03-27 04:23:11 [1] Train Extra: lr=0.0001803 inv=0.3546875 sub=0.0000000
17-03-27 04:24:49 [1] Step: 17800 Acc: 0.59500 0.84329 Cost: 1.17353 0.90748 0.20199 0.06406 Time: 0.00088
17-03-27 04:24:49 [1] Train Extra: lr=0.0001798 inv=0.3975000 sub=0.0000000
17-03-27 04:26:17 [1] Step: 17900 Acc: 0.59000 0.84786 Cost: 1.52946 1.23020 0.23529 0.06396 Time: 0.00085
17-03-27 04:26:17 [1] Train Extra: lr=0.0001793 inv=0.3756250 sub=0.0000000
17-03-27 04:27:43 [1] Step: 18000 Acc: 0.59000 0.84659 Cost: 1.10129 0.83361 0.20373 0.06396 Time: 0.00084
17-03-27 04:27:43 [1] Train Extra: lr=0.0001787 inv=0.3687500 sub=0.0000000
17-03-27 04:28:54 [1] Step: 18000 Eval acc: 0.62510 0.84709 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 04:28:54 [1] Eval Extra: inv=0.3737745
17-03-27 04:30:20 [1] Step: 18100 Acc: 0.60469 0.84304 Cost: 0.96299 0.74328 0.15581 0.06391 Time: 0.00084
17-03-27 04:30:20 [1] Train Extra: lr=0.0001782 inv=0.3803125 sub=0.0000000
17-03-27 04:31:55 [1] Step: 18200 Acc: 0.60406 0.83776 Cost: 1.14092 0.81909 0.25796 0.06387 Time: 0.00084
17-03-27 04:31:55 [1] Train Extra: lr=0.0001777 inv=0.3832813 sub=0.0000000
17-03-27 04:33:20 [1] Step: 18300 Acc: 0.61187 0.84672 Cost: 1.41947 1.11619 0.23947 0.06381 Time: 0.00086
17-03-27 04:33:20 [1] Train Extra: lr=0.0001772 inv=0.3659375 sub=0.0000000
17-03-27 04:34:54 [1] Step: 18400 Acc: 0.60906 0.83887 Cost: 1.01298 0.69603 0.25316 0.06379 Time: 0.00084
17-03-27 04:34:54 [1] Train Extra: lr=0.0001767 inv=0.4204688 sub=0.0000000
17-03-27 04:36:26 [1] Step: 18500 Acc: 0.60594 0.83566 Cost: 1.10497 0.78312 0.25814 0.06370 Time: 0.00084
17-03-27 04:36:26 [1] Train Extra: lr=0.0001762 inv=0.4132812 sub=0.0000000
17-03-27 04:37:33 [1] Step: 18500 Eval acc: 0.61887 0.84895 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 04:37:33 [1] Eval Extra: inv=0.3807700
17-03-27 04:39:07 [1] Step: 18600 Acc: 0.61219 0.83555 Cost: 1.29948 0.93098 0.30486 0.06365 Time: 0.00083
17-03-27 04:39:07 [1] Train Extra: lr=0.0001757 inv=0.4073438 sub=0.0000000
17-03-27 04:40:32 [1] Step: 18700 Acc: 0.60406 0.84465 Cost: 1.11807 0.88708 0.16743 0.06357 Time: 0.00085
17-03-27 04:40:32 [1] Train Extra: lr=0.0001752 inv=0.3828125 sub=0.0000000
17-03-27 04:42:05 [1] Step: 18800 Acc: 0.61187 0.84311 Cost: 1.32322 1.10014 0.15949 0.06358 Time: 0.00085
17-03-27 04:42:05 [1] Train Extra: lr=0.0001747 inv=0.3957812 sub=0.0000000
17-03-27 04:43:34 [1] Step: 18900 Acc: 0.60219 0.85082 Cost: 1.31323 0.94599 0.30361 0.06363 Time: 0.00085
17-03-27 04:43:34 [1] Train Extra: lr=0.0001742 inv=0.3623438 sub=0.0000000
17-03-27 04:45:06 [1] Step: 19000 Acc: 0.61219 0.84593 Cost: 0.97003 0.67794 0.22850 0.06359 Time: 0.00088
17-03-27 04:45:06 [1] Train Extra: lr=0.0001737 inv=0.3873437 sub=0.0000000
17-03-27 04:46:17 [1] Step: 19000 Eval acc: 0.61897 0.85133 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 04:46:17 [1] Eval Extra: inv=0.3661152
17-03-27 04:47:43 [1] Step: 19100 Acc: 0.59031 0.84978 Cost: 0.93338 0.71868 0.15120 0.06350 Time: 0.00086
17-03-27 04:47:43 [1] Train Extra: lr=0.0001732 inv=0.3664062 sub=0.0000000
17-03-27 04:49:10 [1] Step: 19200 Acc: 0.59813 0.83547 Cost: 0.95601 0.76207 0.13046 0.06348 Time: 0.00082
17-03-27 04:49:10 [1] Train Extra: lr=0.0001727 inv=0.3753125 sub=0.0000000
17-03-27 04:50:36 [1] Step: 19300 Acc: 0.59375 0.83799 Cost: 1.13812 0.84304 0.23169 0.06338 Time: 0.00082
17-03-27 04:50:36 [1] Train Extra: lr=0.0001722 inv=0.3898437 sub=0.0000000
17-03-27 04:52:04 [1] Step: 19400 Acc: 0.59313 0.84076 Cost: 1.34534 0.99392 0.28803 0.06339 Time: 0.00083
17-03-27 04:52:04 [1] Train Extra: lr=0.0001717 inv=0.3818750 sub=0.0000000
17-03-27 04:53:30 [1] Step: 19500 Acc: 0.60531 0.84596 Cost: 0.88724 0.62287 0.20097 0.06340 Time: 0.00085
17-03-27 04:53:30 [1] Train Extra: lr=0.0001712 inv=0.3704688 sub=0.0000000
17-03-27 04:54:41 [1] Step: 19500 Eval acc: 0.62275 0.84331 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 04:54:41 [1] Eval Extra: inv=0.3779105
17-03-27 04:56:04 [1] Step: 19600 Acc: 0.61938 0.84122 Cost: 1.04601 0.79455 0.18804 0.06342 Time: 0.00084
17-03-27 04:56:04 [1] Train Extra: lr=0.0001707 inv=0.3707813 sub=0.0000000
17-03-27 04:57:35 [1] Step: 19700 Acc: 0.61969 0.84638 Cost: 0.81582 0.62519 0.12717 0.06345 Time: 0.00089
17-03-27 04:57:35 [1] Train Extra: lr=0.0001702 inv=0.3796875 sub=0.0000000
17-03-27 04:59:11 [1] Step: 19800 Acc: 0.61750 0.84415 Cost: 1.06242 0.77952 0.21945 0.06345 Time: 0.00086
17-03-27 04:59:11 [1] Train Extra: lr=0.0001697 inv=0.3981250 sub=0.0000000
17-03-27 05:00:30 [1] Step: 19900 Acc: 0.61531 0.84547 Cost: 0.91382 0.71281 0.13761 0.06340 Time: 0.00082
17-03-27 05:00:30 [1] Train Extra: lr=0.0001692 inv=0.3628125 sub=0.0000000
17-03-27 05:02:03 [1] Step: 20000 Acc: 0.61750 0.83571 Cost: 1.26724 0.93268 0.27110 0.06346 Time: 0.00084
17-03-27 05:02:03 [1] Train Extra: lr=0.0001687 inv=0.4006250 sub=0.0000000
17-03-27 05:03:14 [1] Step: 20000 Eval acc: 0.62439 0.84911 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 05:03:14 [1] Eval Extra: inv=0.4018587
17-03-27 05:03:14 [1] Checkpointing.
17-03-27 05:04:40 [1] Step: 20100 Acc: 0.61844 0.83788 Cost: 0.97589 0.72489 0.18747 0.06354 Time: 0.00083
17-03-27 05:04:40 [1] Train Extra: lr=0.0001683 inv=0.4046875 sub=0.0000000
17-03-27 05:06:09 [1] Step: 20200 Acc: 0.60781 0.83762 Cost: 1.22688 0.86637 0.29686 0.06365 Time: 0.00082
17-03-27 05:06:09 [1] Train Extra: lr=0.0001678 inv=0.3973437 sub=0.0000000
17-03-27 05:07:48 [1] Step: 20300 Acc: 0.61406 0.84600 Cost: 1.28789 0.94929 0.27503 0.06357 Time: 0.00090
17-03-27 05:07:48 [1] Train Extra: lr=0.0001673 inv=0.3829688 sub=0.0000000
17-03-27 05:09:15 [1] Step: 20400 Acc: 0.61875 0.83749 Cost: 1.18923 0.87160 0.25407 0.06356 Time: 0.00082
17-03-27 05:09:15 [1] Train Extra: lr=0.0001668 inv=0.3967188 sub=0.0000000
17-03-27 05:10:44 [1] Step: 20500 Acc: 0.63281 0.84240 Cost: 0.89773 0.69680 0.13736 0.06357 Time: 0.00084
17-03-27 05:10:44 [1] Train Extra: lr=0.0001663 inv=0.3726563 sub=0.0000000
17-03-27 05:11:56 [1] Step: 20500 Eval acc: 0.63092 0.85064 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 05:11:56 [1] Eval Extra: inv=0.3808211
17-03-27 05:11:56 [1] Checkpointing with new best dev accuracy of 0.630923
17-03-27 05:13:18 [1] Step: 20600 Acc: 0.61281 0.83656 Cost: 1.18486 0.90896 0.21229 0.06362 Time: 0.00080
17-03-27 05:13:18 [1] Train Extra: lr=0.0001659 inv=0.4035937 sub=0.0000000
17-03-27 05:14:49 [1] Step: 20700 Acc: 0.60219 0.83716 Cost: 0.95175 0.73386 0.15427 0.06362 Time: 0.00086
17-03-27 05:14:49 [1] Train Extra: lr=0.0001654 inv=0.3998437 sub=0.0000000
17-03-27 05:16:24 [1] Step: 20800 Acc: 0.60813 0.83640 Cost: 1.10853 0.77741 0.26762 0.06350 Time: 0.00083
17-03-27 05:16:24 [1] Train Extra: lr=0.0001649 inv=0.4406250 sub=0.0000000
17-03-27 05:18:04 [1] Step: 20900 Acc: 0.61813 0.84827 Cost: 1.18379 0.87871 0.24158 0.06350 Time: 0.00090
17-03-27 05:18:04 [1] Train Extra: lr=0.0001644 inv=0.3906250 sub=0.0000000
17-03-27 05:19:32 [1] Step: 21000 Acc: 0.59844 0.83369 Cost: 1.23410 0.93120 0.23940 0.06350 Time: 0.00082
17-03-27 05:19:32 [1] Train Extra: lr=0.0001640 inv=0.3996875 sub=0.0000000
17-03-27 05:20:40 [1] Step: 21000 Eval acc: 0.62561 0.84759 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 05:20:40 [1] Eval Extra: inv=0.3731618
17-03-27 05:22:21 [1] Step: 21100 Acc: 0.62000 0.84463 Cost: 1.10987 0.85188 0.19440 0.06359 Time: 0.00088
17-03-27 05:22:21 [1] Train Extra: lr=0.0001635 inv=0.3954687 sub=0.0000000
17-03-27 05:23:54 [1] Step: 21200 Acc: 0.62094 0.84249 Cost: 1.02614 0.79883 0.16373 0.06358 Time: 0.00085
17-03-27 05:23:54 [1] Train Extra: lr=0.0001630 inv=0.4003125 sub=0.0000000
17-03-27 05:25:15 [1] Step: 21300 Acc: 0.61062 0.84305 Cost: 1.16006 0.85438 0.24215 0.06354 Time: 0.00081
17-03-27 05:25:15 [1] Train Extra: lr=0.0001626 inv=0.3815625 sub=0.0000000
17-03-27 05:26:42 [1] Step: 21400 Acc: 0.61156 0.83996 Cost: 1.11215 0.83244 0.21622 0.06348 Time: 0.00084
17-03-27 05:26:42 [1] Train Extra: lr=0.0001621 inv=0.3898437 sub=0.0000000
17-03-27 05:28:24 [1] Step: 21500 Acc: 0.61906 0.85289 Cost: 1.00576 0.69462 0.24766 0.06348 Time: 0.00093
17-03-27 05:28:24 [1] Train Extra: lr=0.0001616 inv=0.3760938 sub=0.0000000
17-03-27 05:29:34 [1] Step: 21500 Eval acc: 0.62224 0.84682 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 05:29:34 [1] Eval Extra: inv=0.4145221
17-03-27 05:30:56 [1] Step: 21600 Acc: 0.60313 0.84641 Cost: 1.27140 0.93758 0.27040 0.06342 Time: 0.00081
17-03-27 05:30:56 [1] Train Extra: lr=0.0001612 inv=0.3529687 sub=0.0000000
17-03-27 05:32:22 [1] Step: 21700 Acc: 0.63844 0.83933 Cost: 1.19604 0.90428 0.22827 0.06349 Time: 0.00085
17-03-27 05:32:22 [1] Train Extra: lr=0.0001607 inv=0.3665625 sub=0.0000000
17-03-27 05:34:02 [1] Step: 21800 Acc: 0.61031 0.83758 Cost: 1.01527 0.70810 0.24363 0.06354 Time: 0.00087
17-03-27 05:34:02 [1] Train Extra: lr=0.0001602 inv=0.4284375 sub=0.0000000
17-03-27 05:35:42 [1] Step: 21900 Acc: 0.60062 0.84889 Cost: 1.12666 0.81475 0.24843 0.06348 Time: 0.00089
17-03-27 05:35:42 [1] Train Extra: lr=0.0001598 inv=0.3798437 sub=0.0000000
17-03-27 05:37:09 [1] Step: 22000 Acc: 0.61094 0.84853 Cost: 1.07315 0.84213 0.16762 0.06340 Time: 0.00086
17-03-27 05:37:09 [1] Train Extra: lr=0.0001593 inv=0.3648438 sub=0.0000000
17-03-27 05:38:20 [1] Step: 22000 Eval acc: 0.62531 0.85118 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 05:38:20 [1] Eval Extra: inv=0.3849060
17-03-27 05:39:46 [1] Step: 22100 Acc: 0.60125 0.84476 Cost: 1.17037 0.87749 0.22949 0.06338 Time: 0.00085
17-03-27 05:39:46 [1] Train Extra: lr=0.0001589 inv=0.3576563 sub=0.0000000
17-03-27 05:41:20 [1] Step: 22200 Acc: 0.61125 0.84299 Cost: 1.21383 0.96060 0.18981 0.06342 Time: 0.00085
17-03-27 05:41:20 [1] Train Extra: lr=0.0001584 inv=0.3987500 sub=0.0000000
17-03-27 05:42:45 [1] Step: 22300 Acc: 0.61156 0.85428 Cost: 1.15791 0.82553 0.26896 0.06341 Time: 0.00087
17-03-27 05:42:45 [1] Train Extra: lr=0.0001579 inv=0.3553125 sub=0.0000000
17-03-27 05:44:19 [1] Step: 22400 Acc: 0.61594 0.84105 Cost: 1.17180 0.85973 0.24867 0.06340 Time: 0.00086
17-03-27 05:44:19 [1] Train Extra: lr=0.0001575 inv=0.4006250 sub=0.0000000
17-03-27 05:45:54 [1] Step: 22500 Acc: 0.60813 0.83957 Cost: 1.41299 1.04263 0.30700 0.06335 Time: 0.00086
17-03-27 05:45:54 [1] Train Extra: lr=0.0001570 inv=0.3904687 sub=0.0000000
17-03-27 05:47:07 [1] Step: 22500 Eval acc: 0.63511 0.85131 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 05:47:07 [1] Eval Extra: inv=0.3551368
17-03-27 05:47:07 [1] Checkpointing with new best dev accuracy of 0.635110
17-03-27 05:48:51 [1] Step: 22600 Acc: 0.60406 0.84651 Cost: 1.23581 1.02818 0.14422 0.06341 Time: 0.00089
17-03-27 05:48:51 [1] Train Extra: lr=0.0001566 inv=0.3689062 sub=0.0000000
17-03-27 05:50:20 [1] Step: 22700 Acc: 0.60281 0.84255 Cost: 1.04646 0.69501 0.28804 0.06341 Time: 0.00083
17-03-27 05:50:20 [1] Train Extra: lr=0.0001561 inv=0.3993750 sub=0.0000000
17-03-27 05:51:54 [1] Step: 22800 Acc: 0.61625 0.84410 Cost: 1.16113 0.86419 0.23348 0.06346 Time: 0.00086
17-03-27 05:51:54 [1] Train Extra: lr=0.0001557 inv=0.3789062 sub=0.0000000
17-03-27 05:53:28 [1] Step: 22900 Acc: 0.61906 0.83737 Cost: 1.06786 0.78791 0.21644 0.06352 Time: 0.00084
17-03-27 05:53:28 [1] Train Extra: lr=0.0001552 inv=0.4031250 sub=0.0000000
17-03-27 05:54:59 [1] Step: 23000 Acc: 0.61719 0.83786 Cost: 1.08432 0.76661 0.25416 0.06355 Time: 0.00085
17-03-27 05:54:59 [1] Train Extra: lr=0.0001548 inv=0.3637500 sub=0.0000000
17-03-27 05:56:06 [1] Step: 23000 Eval acc: 0.63297 0.85058 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 05:56:06 [1] Eval Extra: inv=0.3808211
17-03-27 05:57:40 [1] Step: 23100 Acc: 0.64187 0.84675 Cost: 1.06160 0.72432 0.27379 0.06349 Time: 0.00087
17-03-27 05:57:40 [1] Train Extra: lr=0.0001544 inv=0.3776563 sub=0.0000000
17-03-27 05:59:12 [1] Step: 23200 Acc: 0.60719 0.84605 Cost: 1.40940 1.13618 0.20975 0.06348 Time: 0.00090
17-03-27 05:59:12 [1] Train Extra: lr=0.0001539 inv=0.3589062 sub=0.0000000
17-03-27 06:00:39 [1] Step: 23300 Acc: 0.61875 0.83981 Cost: 0.96474 0.70049 0.20074 0.06350 Time: 0.00083
17-03-27 06:00:39 [1] Train Extra: lr=0.0001535 inv=0.3798437 sub=0.0000000
17-03-27 06:02:05 [1] Step: 23400 Acc: 0.61719 0.84809 Cost: 1.05936 0.88676 0.10909 0.06351 Time: 0.00086
17-03-27 06:02:05 [1] Train Extra: lr=0.0001530 inv=0.3709375 sub=0.0000000
17-03-27 06:03:40 [1] Step: 23500 Acc: 0.61750 0.84582 Cost: 0.96877 0.81266 0.09262 0.06348 Time: 0.00086
17-03-27 06:03:40 [1] Train Extra: lr=0.0001526 inv=0.4103125 sub=0.0000000
17-03-27 06:04:47 [1] Step: 23500 Eval acc: 0.63031 0.85125 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 06:04:47 [1] Eval Extra: inv=0.3839359
17-03-27 06:06:27 [1] Step: 23600 Acc: 0.62281 0.83717 Cost: 0.86410 0.60691 0.19371 0.06348 Time: 0.00087
17-03-27 06:06:27 [1] Train Extra: lr=0.0001521 inv=0.4160937 sub=0.0000000
17-03-27 06:07:48 [1] Step: 23700 Acc: 0.61219 0.84627 Cost: 1.03714 0.83353 0.14004 0.06358 Time: 0.00081
17-03-27 06:07:48 [1] Train Extra: lr=0.0001517 inv=0.3814063 sub=0.0000000
17-03-27 06:09:27 [1] Step: 23800 Acc: 0.60688 0.84798 Cost: 1.23860 1.03065 0.14442 0.06352 Time: 0.00090
17-03-27 06:09:27 [1] Train Extra: lr=0.0001513 inv=0.3806250 sub=0.0000000
17-03-27 06:10:53 [1] Step: 23900 Acc: 0.61344 0.84349 Cost: 1.07890 0.86428 0.15108 0.06354 Time: 0.00084
17-03-27 06:10:53 [1] Train Extra: lr=0.0001508 inv=0.3810938 sub=0.0000000
17-03-27 06:12:20 [1] Step: 24000 Acc: 0.62969 0.84632 Cost: 0.90668 0.66138 0.18173 0.06357 Time: 0.00087
17-03-27 06:12:20 [1] Train Extra: lr=0.0001504 inv=0.3523438 sub=0.0000000
17-03-27 06:13:28 [1] Step: 24000 Eval acc: 0.62531 0.84514 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 06:13:28 [1] Eval Extra: inv=0.3944547
17-03-27 06:15:10 [1] Step: 24100 Acc: 0.62219 0.84864 Cost: 1.20922 0.85557 0.29009 0.06356 Time: 0.00089
17-03-27 06:15:10 [1] Train Extra: lr=0.0001500 inv=0.3879687 sub=0.0000000
17-03-27 06:16:42 [1] Step: 24200 Acc: 0.63156 0.83847 Cost: 0.99580 0.77461 0.15766 0.06353 Time: 0.00085
17-03-27 06:16:42 [1] Train Extra: lr=0.0001495 inv=0.3934375 sub=0.0000000
17-03-27 06:18:09 [1] Step: 24300 Acc: 0.61719 0.84186 Cost: 1.10183 0.87160 0.16680 0.06343 Time: 0.00084
17-03-27 06:18:09 [1] Train Extra: lr=0.0001491 inv=0.3700000 sub=0.0000000
17-03-27 06:19:35 [1] Step: 24400 Acc: 0.62750 0.84132 Cost: 1.16233 0.83996 0.25888 0.06348 Time: 0.00085
17-03-27 06:19:35 [1] Train Extra: lr=0.0001487 inv=0.3534375 sub=0.0000000
17-03-27 06:21:16 [1] Step: 24500 Acc: 0.60250 0.84800 Cost: 1.14793 0.87519 0.20922 0.06352 Time: 0.00089
17-03-27 06:21:16 [1] Train Extra: lr=0.0001483 inv=0.4064063 sub=0.0000000
17-03-27 06:22:22 [1] Step: 24500 Eval acc: 0.63674 0.85443 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 06:22:22 [1] Eval Extra: inv=0.3954248
17-03-27 06:23:56 [1] Step: 24600 Acc: 0.61125 0.83785 Cost: 1.36422 1.10227 0.19845 0.06351 Time: 0.00086
17-03-27 06:23:56 [1] Train Extra: lr=0.0001478 inv=0.3806250 sub=0.0000000
17-03-27 06:25:11 [1] Step: 24700 Acc: 0.61813 0.84735 Cost: 1.53831 1.20611 0.26871 0.06349 Time: 0.00081
17-03-27 06:25:11 [1] Train Extra: lr=0.0001474 inv=0.3443750 sub=0.0000000
17-03-27 06:26:53 [1] Step: 24800 Acc: 0.61562 0.84550 Cost: 0.88377 0.69848 0.12180 0.06348 Time: 0.00088
17-03-27 06:26:53 [1] Train Extra: lr=0.0001470 inv=0.3865625 sub=0.0000000
17-03-27 06:28:13 [1] Step: 24900 Acc: 0.62094 0.83871 Cost: 1.18497 0.89829 0.22315 0.06352 Time: 0.00079
17-03-27 06:28:13 [1] Train Extra: lr=0.0001466 inv=0.3618750 sub=0.0000000
17-03-27 06:29:45 [1] Step: 25000 Acc: 0.61625 0.83634 Cost: 1.03270 0.82410 0.14514 0.06345 Time: 0.00085
17-03-27 06:29:45 [1] Train Extra: lr=0.0001461 inv=0.4048438 sub=0.0000000
17-03-27 06:30:56 [1] Step: 25000 Eval acc: 0.63470 0.83511 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 06:30:56 [1] Eval Extra: inv=0.4361724
17-03-27 06:30:56 [1] Checkpointing.
17-03-27 06:32:23 [1] Step: 25100 Acc: 0.62469 0.84443 Cost: 1.02553 0.76552 0.19651 0.06350 Time: 0.00085
17-03-27 06:32:23 [1] Train Extra: lr=0.0001457 inv=0.3732813 sub=0.0000000
17-03-27 06:34:01 [1] Step: 25200 Acc: 0.59562 0.84675 Cost: 0.97480 0.77768 0.13368 0.06344 Time: 0.00086
17-03-27 06:34:01 [1] Train Extra: lr=0.0001453 inv=0.4262500 sub=0.0000000
17-03-27 06:35:33 [1] Step: 25300 Acc: 0.61281 0.84102 Cost: 1.17399 0.78317 0.32746 0.06337 Time: 0.00085
17-03-27 06:35:33 [1] Train Extra: lr=0.0001449 inv=0.4101562 sub=0.0000000
17-03-27 06:37:06 [1] Step: 25400 Acc: 0.61938 0.84557 Cost: 1.13051 0.89280 0.17434 0.06337 Time: 0.00086
17-03-27 06:37:06 [1] Train Extra: lr=0.0001445 inv=0.3934375 sub=0.0000000
17-03-27 06:38:50 [1] Step: 25500 Acc: 0.59469 0.83871 Cost: 1.10467 0.81340 0.22785 0.06342 Time: 0.00087
17-03-27 06:38:50 [1] Train Extra: lr=0.0001441 inv=0.4023438 sub=0.0000000
17-03-27 06:40:02 [1] Step: 25500 Eval acc: 0.63603 0.85414 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 06:40:02 [1] Eval Extra: inv=0.3400735
17-03-27 06:41:46 [1] Step: 25600 Acc: 0.60594 0.85170 Cost: 1.05994 0.76113 0.23539 0.06343 Time: 0.00088
17-03-27 06:41:46 [1] Train Extra: lr=0.0001436 inv=0.4076562 sub=0.0000000
17-03-27 06:43:13 [1] Step: 25700 Acc: 0.61500 0.83262 Cost: 1.28279 0.93946 0.27998 0.06335 Time: 0.00081
17-03-27 06:43:13 [1] Train Extra: lr=0.0001432 inv=0.3953125 sub=0.0000000
17-03-27 06:44:43 [1] Step: 25800 Acc: 0.63438 0.84448 Cost: 0.98374 0.82882 0.09153 0.06339 Time: 0.00086
17-03-27 06:44:43 [1] Train Extra: lr=0.0001428 inv=0.4012500 sub=0.0000000
17-03-27 06:46:10 [1] Step: 25900 Acc: 0.61969 0.84723 Cost: 1.26089 0.89636 0.30116 0.06338 Time: 0.00086
17-03-27 06:46:10 [1] Train Extra: lr=0.0001424 inv=0.3535937 sub=0.0000000
17-03-27 06:47:53 [1] Step: 26000 Acc: 0.60656 0.83667 Cost: 0.99257 0.82401 0.10519 0.06337 Time: 0.00086
17-03-27 06:47:53 [1] Train Extra: lr=0.0001420 inv=0.4081250 sub=0.0000000
17-03-27 06:49:05 [1] Step: 26000 Eval acc: 0.63899 0.85289 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 06:49:05 [1] Eval Extra: inv=0.3352737
17-03-27 06:49:05 [1] Checkpointing with new best dev accuracy of 0.638991
17-03-27 06:50:39 [1] Step: 26100 Acc: 0.62062 0.84623 Cost: 1.18917 0.90157 0.22421 0.06339 Time: 0.00087
17-03-27 06:50:39 [1] Train Extra: lr=0.0001416 inv=0.3757813 sub=0.0000000
17-03-27 06:52:01 [1] Step: 26200 Acc: 0.61438 0.84913 Cost: 1.30766 1.05464 0.18963 0.06339 Time: 0.00082
17-03-27 06:52:01 [1] Train Extra: lr=0.0001412 inv=0.3887500 sub=0.0000000
17-03-27 06:53:35 [1] Step: 26300 Acc: 0.61156 0.84797 Cost: 1.09446 0.90465 0.12646 0.06335 Time: 0.00087
17-03-27 06:53:35 [1] Train Extra: lr=0.0001408 inv=0.3742187 sub=0.0000000
17-03-27 06:55:23 [1] Step: 26400 Acc: 0.61969 0.83505 Cost: 1.25673 0.90408 0.28922 0.06343 Time: 0.00087
17-03-27 06:55:23 [1] Train Extra: lr=0.0001404 inv=0.4303125 sub=0.0000000
17-03-27 06:57:11 [1] Step: 26500 Acc: 0.59375 0.85402 Cost: 1.35789 0.96699 0.32753 0.06337 Time: 0.00093
17-03-27 06:57:11 [1] Train Extra: lr=0.0001400 inv=0.3862500 sub=0.0000000
17-03-27 06:58:19 [1] Step: 26500 Eval acc: 0.63848 0.85396 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 06:58:19 [1] Eval Extra: inv=0.4115605
17-03-27 06:59:45 [1] Step: 26600 Acc: 0.61187 0.84245 Cost: 0.89331 0.73061 0.09938 0.06332 Time: 0.00083
17-03-27 06:59:45 [1] Train Extra: lr=0.0001396 inv=0.3926562 sub=0.0000000
17-03-27 07:01:12 [1] Step: 26700 Acc: 0.61500 0.85096 Cost: 0.86350 0.69421 0.10596 0.06333 Time: 0.00086
17-03-27 07:01:12 [1] Train Extra: lr=0.0001392 inv=0.3848437 sub=0.0000000
17-03-27 07:02:44 [1] Step: 26800 Acc: 0.62281 0.84688 Cost: 1.22137 0.91054 0.24742 0.06341 Time: 0.00086
17-03-27 07:02:44 [1] Train Extra: lr=0.0001388 inv=0.3851562 sub=0.0000000
17-03-27 07:04:19 [1] Step: 26900 Acc: 0.61562 0.85247 Cost: 1.16336 0.92642 0.17360 0.06334 Time: 0.00087
17-03-27 07:04:19 [1] Train Extra: lr=0.0001384 inv=0.3762500 sub=0.0000000
17-03-27 07:05:51 [1] Step: 27000 Acc: 0.59688 0.84822 Cost: 1.23095 0.92425 0.24331 0.06339 Time: 0.00087
17-03-27 07:05:51 [1] Train Extra: lr=0.0001380 inv=0.4110937 sub=0.0000000
17-03-27 07:07:03 [1] Step: 27000 Eval acc: 0.63399 0.84859 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 07:07:03 [1] Eval Extra: inv=0.4123775
17-03-27 07:08:23 [1] Step: 27100 Acc: 0.61406 0.84775 Cost: 1.39860 0.97282 0.36240 0.06337 Time: 0.00082
17-03-27 07:08:23 [1] Train Extra: lr=0.0001376 inv=0.3792187 sub=0.0000000
17-03-27 07:09:52 [1] Step: 27200 Acc: 0.61531 0.84500 Cost: 1.04234 0.80170 0.17728 0.06336 Time: 0.00083
17-03-27 07:09:52 [1] Train Extra: lr=0.0001372 inv=0.3995313 sub=0.0000000
17-03-27 07:11:18 [1] Step: 27300 Acc: 0.62062 0.84543 Cost: 1.19282 0.86695 0.26254 0.06332 Time: 0.00085
17-03-27 07:11:18 [1] Train Extra: lr=0.0001368 inv=0.3743750 sub=0.0000000
17-03-27 07:12:50 [1] Step: 27400 Acc: 0.61469 0.84671 Cost: 1.19906 0.97510 0.16061 0.06335 Time: 0.00088
17-03-27 07:12:50 [1] Train Extra: lr=0.0001364 inv=0.3679688 sub=0.0000000
17-03-27 07:14:19 [1] Step: 27500 Acc: 0.59937 0.84314 Cost: 1.28404 0.97843 0.24231 0.06330 Time: 0.00082
17-03-27 07:14:19 [1] Train Extra: lr=0.0001360 inv=0.4081250 sub=0.0000000
17-03-27 07:15:30 [1] Step: 27500 Eval acc: 0.63838 0.85483 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 07:15:30 [1] Eval Extra: inv=0.3540135
17-03-27 07:16:56 [1] Step: 27600 Acc: 0.61781 0.85037 Cost: 1.37966 0.98045 0.33595 0.06326 Time: 0.00087
17-03-27 07:16:56 [1] Train Extra: lr=0.0001356 inv=0.3554688 sub=0.0000000
17-03-27 07:18:30 [1] Step: 27700 Acc: 0.61469 0.85152 Cost: 1.26412 0.90175 0.29919 0.06318 Time: 0.00087
17-03-27 07:18:30 [1] Train Extra: lr=0.0001352 inv=0.4079687 sub=0.0000000
17-03-27 07:19:49 [1] Step: 27800 Acc: 0.62656 0.85219 Cost: 1.03375 0.84923 0.12128 0.06324 Time: 0.00084
17-03-27 07:19:49 [1] Train Extra: lr=0.0001348 inv=0.3715625 sub=0.0000000
17-03-27 07:21:19 [1] Step: 27900 Acc: 0.60281 0.84839 Cost: 1.11300 0.76984 0.27991 0.06325 Time: 0.00084
17-03-27 07:21:19 [1] Train Extra: lr=0.0001344 inv=0.3623438 sub=0.0000000
17-03-27 07:22:53 [1] Step: 28000 Acc: 0.61625 0.85066 Cost: 1.08249 0.81495 0.20428 0.06326 Time: 0.00087
17-03-27 07:22:53 [1] Train Extra: lr=0.0001341 inv=0.3867188 sub=0.0000000
17-03-27 07:24:04 [1] Step: 28000 Eval acc: 0.64103 0.85201 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 07:24:04 [1] Eval Extra: inv=0.3633578
17-03-27 07:25:40 [1] Step: 28100 Acc: 0.61750 0.84256 Cost: 1.20005 0.85201 0.28477 0.06326 Time: 0.00085
17-03-27 07:25:40 [1] Train Extra: lr=0.0001337 inv=0.4017188 sub=0.0000000
17-03-27 07:27:12 [1] Step: 28200 Acc: 0.61781 0.83860 Cost: 1.01817 0.84137 0.11353 0.06327 Time: 0.00083
17-03-27 07:27:12 [1] Train Extra: lr=0.0001333 inv=0.4046875 sub=0.0000000
17-03-27 07:28:52 [1] Step: 28300 Acc: 0.61000 0.84349 Cost: 1.16831 0.94950 0.15551 0.06329 Time: 0.00089
17-03-27 07:28:52 [1] Train Extra: lr=0.0001329 inv=0.3839063 sub=0.0000000
17-03-27 07:30:19 [1] Step: 28400 Acc: 0.61406 0.84094 Cost: 1.32141 0.96416 0.29396 0.06329 Time: 0.00083
17-03-27 07:30:19 [1] Train Extra: lr=0.0001325 inv=0.3765625 sub=0.0000000
17-03-27 07:31:38 [1] Step: 28500 Acc: 0.61406 0.84829 Cost: 0.99989 0.82441 0.11215 0.06333 Time: 0.00082
17-03-27 07:31:38 [1] Train Extra: lr=0.0001321 inv=0.3557812 sub=0.0000000
17-03-27 07:32:50 [1] Step: 28500 Eval acc: 0.64379 0.85107 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 07:32:50 [1] Eval Extra: inv=0.3873570
17-03-27 07:32:50 [1] Checkpointing with new best dev accuracy of 0.643791
17-03-27 07:34:17 [1] Step: 28600 Acc: 0.61719 0.84694 Cost: 1.09741 0.90250 0.13160 0.06332 Time: 0.00084
17-03-27 07:34:17 [1] Train Extra: lr=0.0001318 inv=0.3748437 sub=0.0000000
17-03-27 07:35:37 [1] Step: 28700 Acc: 0.61687 0.85718 Cost: 1.20006 0.94581 0.19094 0.06332 Time: 0.00084
17-03-27 07:35:37 [1] Train Extra: lr=0.0001314 inv=0.3601563 sub=0.0000000
17-03-27 07:37:10 [1] Step: 28800 Acc: 0.62281 0.85249 Cost: 1.39213 1.02240 0.30640 0.06333 Time: 0.00088
17-03-27 07:37:10 [1] Train Extra: lr=0.0001310 inv=0.3743750 sub=0.0000000
17-03-27 07:38:40 [1] Step: 28900 Acc: 0.62031 0.84905 Cost: 1.11418 0.84631 0.20443 0.06344 Time: 0.00085
17-03-27 07:38:40 [1] Train Extra: lr=0.0001306 inv=0.4046875 sub=0.0000000
17-03-27 07:40:17 [1] Step: 29000 Acc: 0.61219 0.84280 Cost: 1.00097 0.74216 0.19536 0.06346 Time: 0.00087
17-03-27 07:40:17 [1] Train Extra: lr=0.0001303 inv=0.4370312 sub=0.0000000
17-03-27 07:41:24 [1] Step: 29000 Eval acc: 0.63756 0.84844 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 07:41:24 [1] Eval Extra: inv=0.3977737
17-03-27 07:42:48 [1] Step: 29100 Acc: 0.61062 0.84552 Cost: 1.25482 0.94188 0.24950 0.06344 Time: 0.00083
17-03-27 07:42:48 [1] Train Extra: lr=0.0001299 inv=0.3950000 sub=0.0000000
17-03-27 07:44:14 [1] Step: 29200 Acc: 0.61719 0.84862 Cost: 1.07444 0.81636 0.19461 0.06346 Time: 0.00086
17-03-27 07:44:14 [1] Train Extra: lr=0.0001295 inv=0.3832813 sub=0.0000000
17-03-27 07:45:55 [1] Step: 29300 Acc: 0.61531 0.84297 Cost: 1.37026 0.94401 0.36275 0.06349 Time: 0.00087
17-03-27 07:45:55 [1] Train Extra: lr=0.0001291 inv=0.4070313 sub=0.0000000
17-03-27 07:47:28 [1] Step: 29400 Acc: 0.61125 0.84433 Cost: 1.20842 0.90830 0.23661 0.06352 Time: 0.00085
17-03-27 07:47:28 [1] Train Extra: lr=0.0001288 inv=0.4023438 sub=0.0000000
17-03-27 07:49:13 [1] Step: 29500 Acc: 0.62344 0.85243 Cost: 1.21642 0.90816 0.24469 0.06357 Time: 0.00090
17-03-27 07:49:13 [1] Train Extra: lr=0.0001284 inv=0.3884375 sub=0.0000000
17-03-27 07:50:24 [1] Step: 29500 Eval acc: 0.64205 0.84935 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 07:50:24 [1] Eval Extra: inv=0.3527880
17-03-27 07:52:00 [1] Step: 29600 Acc: 0.62781 0.84685 Cost: 1.14068 0.86813 0.20895 0.06360 Time: 0.00085
17-03-27 07:52:00 [1] Train Extra: lr=0.0001280 inv=0.3857813 sub=0.0000000
17-03-27 07:53:33 [1] Step: 29700 Acc: 0.63500 0.84451 Cost: 1.13641 0.99826 0.07451 0.06365 Time: 0.00086
17-03-27 07:53:33 [1] Train Extra: lr=0.0001277 inv=0.3785938 sub=0.0000000
17-03-27 07:55:07 [1] Step: 29800 Acc: 0.62219 0.84463 Cost: 1.31843 1.05531 0.19938 0.06374 Time: 0.00084
17-03-27 07:55:07 [1] Train Extra: lr=0.0001273 inv=0.4223438 sub=0.0000000
17-03-27 07:56:38 [1] Step: 29900 Acc: 0.64125 0.85245 Cost: 1.18314 0.85712 0.26226 0.06376 Time: 0.00088
17-03-27 07:56:38 [1] Train Extra: lr=0.0001269 inv=0.3785938 sub=0.0000000
17-03-27 07:58:07 [1] Step: 30000 Acc: 0.63719 0.83709 Cost: 0.99520 0.81168 0.11964 0.06388 Time: 0.00080
17-03-27 07:58:07 [1] Train Extra: lr=0.0001266 inv=0.4115625 sub=0.0000000
17-03-27 07:59:19 [1] Step: 30000 Eval acc: 0.63920 0.85340 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 07:59:19 [1] Eval Extra: inv=0.3907271
17-03-27 07:59:19 [1] Checkpointing.
17-03-27 08:00:46 [1] Step: 30100 Acc: 0.64062 0.84550 Cost: 0.87913 0.69898 0.11624 0.06391 Time: 0.00083
17-03-27 08:00:46 [1] Train Extra: lr=0.0001262 inv=0.3632812 sub=0.0000000
17-03-27 08:02:11 [1] Step: 30200 Acc: 0.64031 0.85056 Cost: 1.05265 0.81212 0.17656 0.06397 Time: 0.00085
17-03-27 08:02:11 [1] Train Extra: lr=0.0001258 inv=0.3729688 sub=0.0000000
17-03-27 08:03:46 [1] Step: 30300 Acc: 0.62125 0.84650 Cost: 1.36497 1.02343 0.27752 0.06403 Time: 0.00085
17-03-27 08:03:46 [1] Train Extra: lr=0.0001255 inv=0.4109375 sub=0.0000000
17-03-27 08:05:13 [1] Step: 30400 Acc: 0.63156 0.85023 Cost: 1.00691 0.81158 0.13123 0.06410 Time: 0.00085
17-03-27 08:05:13 [1] Train Extra: lr=0.0001251 inv=0.3687500 sub=0.0000000
17-03-27 08:06:52 [1] Step: 30500 Acc: 0.63344 0.84181 Cost: 1.06449 0.74484 0.25553 0.06411 Time: 0.00088
17-03-27 08:06:52 [1] Train Extra: lr=0.0001248 inv=0.4176563 sub=0.0000000
17-03-27 08:08:02 [1] Step: 30500 Eval acc: 0.64154 0.85080 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 08:08:02 [1] Eval Extra: inv=0.4105392
17-03-27 08:09:28 [1] Step: 30600 Acc: 0.60938 0.84346 Cost: 1.04921 0.83293 0.15209 0.06419 Time: 0.00082
17-03-27 08:09:28 [1] Train Extra: lr=0.0001244 inv=0.4110937 sub=0.0000000
17-03-27 08:10:48 [1] Step: 30700 Acc: 0.61750 0.84729 Cost: 1.16347 0.89772 0.20146 0.06429 Time: 0.00083
17-03-27 08:10:48 [1] Train Extra: lr=0.0001240 inv=0.3345313 sub=0.0000000
17-03-27 08:12:14 [1] Step: 30800 Acc: 0.62969 0.84623 Cost: 0.95692 0.69976 0.19281 0.06435 Time: 0.00083
17-03-27 08:12:14 [1] Train Extra: lr=0.0001237 inv=0.3878125 sub=0.0000000
17-03-27 08:13:50 [1] Step: 30900 Acc: 0.61594 0.85128 Cost: 1.23605 0.96159 0.21004 0.06441 Time: 0.00087
17-03-27 08:13:50 [1] Train Extra: lr=0.0001233 inv=0.3825000 sub=0.0000000
17-03-27 08:15:37 [1] Step: 31000 Acc: 0.62687 0.85050 Cost: 1.01926 0.76657 0.18827 0.06443 Time: 0.00092
17-03-27 08:15:37 [1] Train Extra: lr=0.0001230 inv=0.3942188 sub=0.0000000
17-03-27 08:16:50 [1] Step: 31000 Eval acc: 0.64246 0.85414 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 08:16:50 [1] Eval Extra: inv=0.4008374
17-03-27 08:18:24 [1] Step: 31100 Acc: 0.62281 0.84768 Cost: 1.15886 0.82641 0.26798 0.06447 Time: 0.00087
17-03-27 08:18:24 [1] Train Extra: lr=0.0001226 inv=0.3762500 sub=0.0000000
17-03-27 08:19:59 [1] Step: 31200 Acc: 0.61938 0.84812 Cost: 1.07093 0.85091 0.15550 0.06452 Time: 0.00085
17-03-27 08:19:59 [1] Train Extra: lr=0.0001223 inv=0.4107812 sub=0.0000000
17-03-27 08:21:40 [1] Step: 31300 Acc: 0.63562 0.84805 Cost: 1.14599 0.84231 0.23912 0.06455 Time: 0.00088
17-03-27 08:21:40 [1] Train Extra: lr=0.0001219 inv=0.4042188 sub=0.0000000
17-03-27 08:23:06 [1] Step: 31400 Acc: 0.62969 0.83951 Cost: 1.24064 0.90061 0.27543 0.06460 Time: 0.00080
17-03-27 08:23:06 [1] Train Extra: lr=0.0001216 inv=0.4384375 sub=0.0000000
17-03-27 08:24:44 [1] Step: 31500 Acc: 0.61656 0.83160 Cost: 1.06399 0.87208 0.12719 0.06471 Time: 0.00080
17-03-27 08:24:44 [1] Train Extra: lr=0.0001212 inv=0.4632812 sub=0.0000000
17-03-27 08:25:56 [1] Step: 31500 Eval acc: 0.64389 0.85123 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 08:25:56 [1] Eval Extra: inv=0.3737745
17-03-27 08:27:37 [1] Step: 31600 Acc: 0.63031 0.84187 Cost: 1.52169 1.03962 0.41729 0.06478 Time: 0.00087
17-03-27 08:27:37 [1] Train Extra: lr=0.0001209 inv=0.4162500 sub=0.0000000
17-03-27 08:29:05 [1] Step: 31700 Acc: 0.63094 0.84539 Cost: 0.98700 0.84755 0.07459 0.06486 Time: 0.00082
17-03-27 08:29:05 [1] Train Extra: lr=0.0001205 inv=0.3901562 sub=0.0000000
17-03-27 08:30:26 [1] Step: 31800 Acc: 0.62750 0.84365 Cost: 1.05662 0.75588 0.23580 0.06494 Time: 0.00080
17-03-27 08:30:26 [1] Train Extra: lr=0.0001202 inv=0.3942188 sub=0.0000000
17-03-27 08:31:59 [1] Step: 31900 Acc: 0.64219 0.84553 Cost: 1.27232 0.96412 0.24314 0.06505 Time: 0.00088
17-03-27 08:31:59 [1] Train Extra: lr=0.0001198 inv=0.3839063 sub=0.0000000
17-03-27 08:33:26 [1] Step: 32000 Acc: 0.64656 0.84489 Cost: 0.89905 0.70831 0.12568 0.06506 Time: 0.00083
17-03-27 08:33:26 [1] Train Extra: lr=0.0001195 inv=0.3950000 sub=0.0000000
17-03-27 08:34:37 [1] Step: 32000 Eval acc: 0.63797 0.85211 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 08:34:37 [1] Eval Extra: inv=0.3561581
17-03-27 08:35:57 [1] Step: 32100 Acc: 0.63438 0.84521 Cost: 0.99476 0.71232 0.21732 0.06513 Time: 0.00080
17-03-27 08:35:57 [1] Train Extra: lr=0.0001191 inv=0.3746875 sub=0.0000000
17-03-27 08:37:24 [1] Step: 32200 Acc: 0.63844 0.85684 Cost: 0.95252 0.71268 0.17479 0.06506 Time: 0.00087
17-03-27 08:37:24 [1] Train Extra: lr=0.0001188 inv=0.3500000 sub=0.0000000
17-03-27 08:39:06 [1] Step: 32300 Acc: 0.62594 0.84433 Cost: 1.14264 0.88783 0.18969 0.06512 Time: 0.00089
17-03-27 08:39:06 [1] Train Extra: lr=0.0001185 inv=0.3806250 sub=0.0000000
17-03-27 08:40:32 [1] Step: 32400 Acc: 0.62687 0.84953 Cost: 1.24285 0.95781 0.21989 0.06515 Time: 0.00085
17-03-27 08:40:32 [1] Train Extra: lr=0.0001181 inv=0.3823437 sub=0.0000000
17-03-27 08:42:01 [1] Step: 32500 Acc: 0.62687 0.84898 Cost: 1.09383 0.73015 0.29851 0.06517 Time: 0.00085
17-03-27 08:42:01 [1] Train Extra: lr=0.0001178 inv=0.3953125 sub=0.0000000
17-03-27 08:43:13 [1] Step: 32500 Eval acc: 0.64726 0.85475 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 08:43:13 [1] Eval Extra: inv=0.3969567
17-03-27 08:44:41 [1] Step: 32600 Acc: 0.62906 0.84347 Cost: 1.26214 0.99522 0.20174 0.06518 Time: 0.00082
17-03-27 08:44:41 [1] Train Extra: lr=0.0001174 inv=0.4164062 sub=0.0000000
17-03-27 08:46:15 [1] Step: 32700 Acc: 0.63156 0.85029 Cost: 1.25692 0.97604 0.21560 0.06528 Time: 0.00089
17-03-27 08:46:15 [1] Train Extra: lr=0.0001171 inv=0.3873437 sub=0.0000000
17-03-27 08:47:48 [1] Step: 32800 Acc: 0.62969 0.85663 Cost: 1.12939 0.78738 0.27667 0.06534 Time: 0.00087
17-03-27 08:47:48 [1] Train Extra: lr=0.0001168 inv=0.4060937 sub=0.0000000
17-03-27 08:49:16 [1] Step: 32900 Acc: 0.62062 0.84277 Cost: 1.14478 0.86333 0.21610 0.06535 Time: 0.00081
17-03-27 08:49:16 [1] Train Extra: lr=0.0001164 inv=0.4217187 sub=0.0000000
17-03-27 08:50:49 [1] Step: 33000 Acc: 0.62187 0.84578 Cost: 1.16670 0.84554 0.25582 0.06534 Time: 0.00086
17-03-27 08:50:49 [1] Train Extra: lr=0.0001161 inv=0.4150000 sub=0.0000000
17-03-27 08:51:57 [1] Step: 33000 Eval acc: 0.64685 0.85358 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 08:51:57 [1] Eval Extra: inv=0.3778084
17-03-27 08:53:30 [1] Step: 33100 Acc: 0.61438 0.84532 Cost: 0.84725 0.63434 0.14755 0.06536 Time: 0.00085
17-03-27 08:53:30 [1] Train Extra: lr=0.0001158 inv=0.4054687 sub=0.0000000
17-03-27 08:54:58 [1] Step: 33200 Acc: 0.61156 0.84046 Cost: 1.25185 0.89753 0.28885 0.06546 Time: 0.00082
17-03-27 08:54:58 [1] Train Extra: lr=0.0001154 inv=0.4001562 sub=0.0000000
17-03-27 08:56:31 [1] Step: 33300 Acc: 0.62875 0.85228 Cost: 1.00550 0.71889 0.22117 0.06545 Time: 0.00087
17-03-27 08:56:31 [1] Train Extra: lr=0.0001151 inv=0.3998437 sub=0.0000000
17-03-27 08:58:02 [1] Step: 33400 Acc: 0.63469 0.84518 Cost: 1.09831 0.79408 0.23881 0.06543 Time: 0.00087
17-03-27 08:58:02 [1] Train Extra: lr=0.0001148 inv=0.3943750 sub=0.0000000
17-03-27 08:59:29 [1] Step: 33500 Acc: 0.63594 0.84417 Cost: 1.40715 1.13939 0.20225 0.06551 Time: 0.00083
17-03-27 08:59:29 [1] Train Extra: lr=0.0001144 inv=0.4085937 sub=0.0000000
17-03-27 09:00:35 [1] Step: 33500 Eval acc: 0.64747 0.85323 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 09:00:35 [1] Eval Extra: inv=0.4197815
17-03-27 09:00:35 [1] Checkpointing with new best dev accuracy of 0.647467
17-03-27 09:02:07 [1] Step: 33600 Acc: 0.61813 0.84830 Cost: 1.35255 0.95386 0.33308 0.06561 Time: 0.00087
17-03-27 09:02:07 [1] Train Extra: lr=0.0001141 inv=0.3718750 sub=0.0000000
17-03-27 09:03:43 [1] Step: 33700 Acc: 0.64031 0.84698 Cost: 1.21106 0.84296 0.30248 0.06563 Time: 0.00088
17-03-27 09:03:43 [1] Train Extra: lr=0.0001138 inv=0.3946875 sub=0.0000000
17-03-27 09:05:20 [1] Step: 33800 Acc: 0.64375 0.84555 Cost: 1.02731 0.68990 0.27176 0.06564 Time: 0.00086
17-03-27 09:05:20 [1] Train Extra: lr=0.0001135 inv=0.4085937 sub=0.0000000
17-03-27 09:06:47 [1] Step: 33900 Acc: 0.63219 0.85510 Cost: 1.17581 0.85575 0.25440 0.06566 Time: 0.00087
17-03-27 09:06:47 [1] Train Extra: lr=0.0001131 inv=0.3559375 sub=0.0000000
17-03-27 09:08:13 [1] Step: 34000 Acc: 0.64500 0.84456 Cost: 1.28104 0.94870 0.26664 0.06570 Time: 0.00084
17-03-27 09:08:13 [1] Train Extra: lr=0.0001128 inv=0.4029687 sub=0.0000000
17-03-27 09:09:25 [1] Step: 34000 Eval acc: 0.64471 0.85404 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 09:09:25 [1] Eval Extra: inv=0.4166156
17-03-27 09:10:44 [1] Step: 34100 Acc: 0.64406 0.85759 Cost: 1.09331 0.81882 0.20877 0.06572 Time: 0.00084
17-03-27 09:10:44 [1] Train Extra: lr=0.0001125 inv=0.3707813 sub=0.0000000
17-03-27 09:12:13 [1] Step: 34200 Acc: 0.62906 0.85233 Cost: 1.21074 0.92066 0.22424 0.06584 Time: 0.00085
17-03-27 09:12:13 [1] Train Extra: lr=0.0001122 inv=0.3829688 sub=0.0000000
17-03-27 09:13:47 [1] Step: 34300 Acc: 0.62687 0.84383 Cost: 1.24735 0.98829 0.19326 0.06581 Time: 0.00085
17-03-27 09:13:47 [1] Train Extra: lr=0.0001118 inv=0.4225000 sub=0.0000000
17-03-27 09:15:27 [1] Step: 34400 Acc: 0.62125 0.85060 Cost: 1.14733 0.89879 0.18282 0.06572 Time: 0.00089
17-03-27 09:15:27 [1] Train Extra: lr=0.0001115 inv=0.3939063 sub=0.0000000
17-03-27 09:16:53 [1] Step: 34500 Acc: 0.61500 0.84352 Cost: 1.31485 0.92644 0.32263 0.06578 Time: 0.00084
17-03-27 09:16:53 [1] Train Extra: lr=0.0001112 inv=0.3971875 sub=0.0000000
17-03-27 09:18:00 [1] Step: 34500 Eval acc: 0.64869 0.85698 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 09:18:00 [1] Eval Extra: inv=0.3842422
17-03-27 09:19:31 [1] Step: 34600 Acc: 0.62813 0.84194 Cost: 1.10229 0.80576 0.23079 0.06574 Time: 0.00086
17-03-27 09:19:31 [1] Train Extra: lr=0.0001109 inv=0.4020313 sub=0.0000000
17-03-27 09:21:07 [1] Step: 34700 Acc: 0.62000 0.84169 Cost: 1.24957 0.92526 0.25864 0.06568 Time: 0.00084
17-03-27 09:21:07 [1] Train Extra: lr=0.0001106 inv=0.4304688 sub=0.0000000
17-03-27 09:22:33 [1] Step: 34800 Acc: 0.64094 0.84936 Cost: 0.76154 0.58138 0.11438 0.06578 Time: 0.00083
17-03-27 09:22:33 [1] Train Extra: lr=0.0001102 inv=0.3873437 sub=0.0000000
17-03-27 09:24:00 [1] Step: 34900 Acc: 0.63187 0.84913 Cost: 1.04412 0.79369 0.18458 0.06584 Time: 0.00086
17-03-27 09:24:00 [1] Train Extra: lr=0.0001099 inv=0.3843750 sub=0.0000000
17-03-27 09:25:31 [1] Step: 35000 Acc: 0.63906 0.84413 Cost: 0.99064 0.75721 0.16757 0.06586 Time: 0.00083
17-03-27 09:25:31 [1] Train Extra: lr=0.0001096 inv=0.4009375 sub=0.0000000
17-03-27 09:26:43 [1] Step: 35000 Eval acc: 0.64451 0.85312 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 09:26:43 [1] Eval Extra: inv=0.3614685
17-03-27 09:26:43 [1] Checkpointing.
17-03-27 09:28:08 [1] Step: 35100 Acc: 0.63219 0.84444 Cost: 1.22236 0.83064 0.32584 0.06589 Time: 0.00084
17-03-27 09:28:08 [1] Train Extra: lr=0.0001093 inv=0.3685937 sub=0.0000000
17-03-27 09:29:35 [1] Step: 35200 Acc: 0.62344 0.85377 Cost: 1.16275 0.86597 0.23089 0.06589 Time: 0.00084
17-03-27 09:29:35 [1] Train Extra: lr=0.0001090 inv=0.3956250 sub=0.0000000
17-03-27 09:31:02 [1] Step: 35300 Acc: 0.63875 0.84794 Cost: 1.20527 0.98565 0.15371 0.06591 Time: 0.00084
17-03-27 09:31:02 [1] Train Extra: lr=0.0001087 inv=0.3945312 sub=0.0000000
17-03-27 09:32:36 [1] Step: 35400 Acc: 0.63031 0.84449 Cost: 1.38365 0.96279 0.35497 0.06589 Time: 0.00085
17-03-27 09:32:36 [1] Train Extra: lr=0.0001084 inv=0.4101562 sub=0.0000000
17-03-27 09:34:09 [1] Step: 35500 Acc: 0.63375 0.85340 Cost: 1.44378 1.06321 0.31469 0.06588 Time: 0.00088
17-03-27 09:34:09 [1] Train Extra: lr=0.0001080 inv=0.3942188 sub=0.0000000
17-03-27 09:35:21 [1] Step: 35500 Eval acc: 0.64747 0.85608 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 09:35:21 [1] Eval Extra: inv=0.3711193
17-03-27 09:36:46 [1] Step: 35600 Acc: 0.64906 0.84188 Cost: 1.19530 0.88955 0.23976 0.06599 Time: 0.00084
17-03-27 09:36:46 [1] Train Extra: lr=0.0001077 inv=0.3695312 sub=0.0000000
17-03-27 09:38:21 [1] Step: 35700 Acc: 0.63375 0.86005 Cost: 1.14693 0.87359 0.20729 0.06604 Time: 0.00089
17-03-27 09:38:21 [1] Train Extra: lr=0.0001074 inv=0.3789062 sub=0.0000000
17-03-27 09:39:48 [1] Step: 35800 Acc: 0.63656 0.84803 Cost: 0.76781 0.53806 0.16366 0.06610 Time: 0.00084
17-03-27 09:39:48 [1] Train Extra: lr=0.0001071 inv=0.3837500 sub=0.0000000
17-03-27 09:41:35 [1] Step: 35900 Acc: 0.62594 0.84563 Cost: 1.22864 0.89259 0.26997 0.06607 Time: 0.00090
17-03-27 09:41:35 [1] Train Extra: lr=0.0001068 inv=0.4042188 sub=0.0000000
17-03-27 09:43:01 [1] Step: 36000 Acc: 0.63250 0.85471 Cost: 1.04659 0.75037 0.23014 0.06608 Time: 0.00085
17-03-27 09:43:01 [1] Train Extra: lr=0.0001065 inv=0.3759375 sub=0.0000000
17-03-27 09:44:13 [1] Step: 36000 Eval acc: 0.65176 0.85426 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 09:44:13 [1] Eval Extra: inv=0.3800041
17-03-27 09:44:13 [1] Checkpointing with new best dev accuracy of 0.651757
17-03-27 09:45:44 [1] Step: 36100 Acc: 0.63219 0.84002 Cost: 1.44718 1.08476 0.29630 0.06612 Time: 0.00082
17-03-27 09:45:44 [1] Train Extra: lr=0.0001062 inv=0.4185937 sub=0.0000000
17-03-27 09:47:10 [1] Step: 36200 Acc: 0.65250 0.85387 Cost: 1.19148 0.92142 0.20392 0.06614 Time: 0.00087
17-03-27 09:47:10 [1] Train Extra: lr=0.0001059 inv=0.3517188 sub=0.0000000
17-03-27 09:48:43 [1] Step: 36300 Acc: 0.61594 0.84524 Cost: 1.13177 0.78860 0.27702 0.06614 Time: 0.00085
17-03-27 09:48:43 [1] Train Extra: lr=0.0001056 inv=0.4009375 sub=0.0000000
17-03-27 09:50:19 [1] Step: 36400 Acc: 0.62813 0.84730 Cost: 1.16400 0.83007 0.26775 0.06617 Time: 0.00087
17-03-27 09:50:19 [1] Train Extra: lr=0.0001053 inv=0.4012500 sub=0.0000000
17-03-27 09:51:52 [1] Step: 36500 Acc: 0.63750 0.84128 Cost: 1.13764 0.84825 0.22326 0.06613 Time: 0.00084
17-03-27 09:51:52 [1] Train Extra: lr=0.0001050 inv=0.4254688 sub=0.0000000
17-03-27 09:53:04 [1] Step: 36500 Eval acc: 0.64073 0.85136 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 09:53:04 [1] Eval Extra: inv=0.4127859
17-03-27 09:54:30 [1] Step: 36600 Acc: 0.63906 0.84107 Cost: 1.05512 0.78804 0.20089 0.06618 Time: 0.00082
17-03-27 09:54:30 [1] Train Extra: lr=0.0001047 inv=0.3959375 sub=0.0000000
17-03-27 09:56:06 [1] Step: 36700 Acc: 0.60750 0.84325 Cost: 1.09909 0.77935 0.25364 0.06611 Time: 0.00084
17-03-27 09:56:06 [1] Train Extra: lr=0.0001044 inv=0.4385938 sub=0.0000000
17-03-27 09:57:39 [1] Step: 36800 Acc: 0.64125 0.84971 Cost: 1.18756 1.01942 0.10198 0.06617 Time: 0.00088
17-03-27 09:57:39 [1] Train Extra: lr=0.0001041 inv=0.3751563 sub=0.0000000
17-03-27 09:59:15 [1] Step: 36900 Acc: 0.62969 0.84746 Cost: 1.42045 1.07741 0.27686 0.06618 Time: 0.00086
17-03-27 09:59:15 [1] Train Extra: lr=0.0001038 inv=0.3910938 sub=0.0000000
17-03-27 10:00:54 [1] Step: 37000 Acc: 0.62000 0.85293 Cost: 0.77491 0.64444 0.06438 0.06608 Time: 0.00090
17-03-27 10:00:54 [1] Train Extra: lr=0.0001035 inv=0.3935938 sub=0.0000000
17-03-27 10:02:02 [1] Step: 37000 Eval acc: 0.63817 0.85308 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 10:02:02 [1] Eval Extra: inv=0.3739277
17-03-27 10:03:28 [1] Step: 37100 Acc: 0.63594 0.85207 Cost: 1.19408 0.91098 0.21689 0.06621 Time: 0.00085
17-03-27 10:03:28 [1] Train Extra: lr=0.0001032 inv=0.3918750 sub=0.0000000
17-03-27 10:05:02 [1] Step: 37200 Acc: 0.62813 0.85173 Cost: 1.08555 0.90177 0.11758 0.06619 Time: 0.00087
17-03-27 10:05:02 [1] Train Extra: lr=0.0001029 inv=0.4057812 sub=0.0000000
17-03-27 10:06:29 [1] Step: 37300 Acc: 0.62750 0.84494 Cost: 1.17928 0.84824 0.26485 0.06619 Time: 0.00084
17-03-27 10:06:29 [1] Train Extra: lr=0.0001026 inv=0.3757813 sub=0.0000000
17-03-27 10:07:55 [1] Step: 37400 Acc: 0.62844 0.84935 Cost: 1.37070 0.98740 0.31707 0.06624 Time: 0.00085
17-03-27 10:07:55 [1] Train Extra: lr=0.0001023 inv=0.3670312 sub=0.0000000
17-03-27 10:09:27 [1] Step: 37500 Acc: 0.62281 0.84378 Cost: 1.12516 0.96739 0.09153 0.06625 Time: 0.00085
17-03-27 10:09:27 [1] Train Extra: lr=0.0001020 inv=0.4073438 sub=0.0000000
17-03-27 10:10:35 [1] Step: 37500 Eval acc: 0.64042 0.85099 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 10:10:35 [1] Eval Extra: inv=0.4351511
17-03-27 10:12:01 [1] Step: 37600 Acc: 0.62469 0.84247 Cost: 1.52074 1.15826 0.29623 0.06625 Time: 0.00083
17-03-27 10:12:01 [1] Train Extra: lr=0.0001017 inv=0.3826562 sub=0.0000000
17-03-27 10:13:27 [1] Step: 37700 Acc: 0.62969 0.84755 Cost: 0.72561 0.51573 0.14364 0.06624 Time: 0.00085
17-03-27 10:13:27 [1] Train Extra: lr=0.0001014 inv=0.3696875 sub=0.0000000
17-03-27 10:14:59 [1] Step: 37800 Acc: 0.63250 0.84849 Cost: 1.10774 0.81986 0.22163 0.06625 Time: 0.00086
17-03-27 10:14:59 [1] Train Extra: lr=0.0001011 inv=0.3890625 sub=0.0000000
17-03-27 10:16:42 [1] Step: 37900 Acc: 0.62281 0.85033 Cost: 1.01730 0.75332 0.19775 0.06622 Time: 0.00088
17-03-27 10:16:42 [1] Train Extra: lr=0.0001008 inv=0.4254688 sub=0.0000000
17-03-27 10:18:12 [1] Step: 38000 Acc: 0.63875 0.84263 Cost: 1.32390 0.98952 0.26820 0.06618 Time: 0.00082
17-03-27 10:18:12 [1] Train Extra: lr=0.0001005 inv=0.4093750 sub=0.0000000
17-03-27 10:19:21 [1] Step: 38000 Eval acc: 0.64828 0.85377 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 10:19:21 [1] Eval Extra: inv=0.3944036
17-03-27 10:20:44 [1] Step: 38100 Acc: 0.63344 0.84322 Cost: 1.24405 0.90345 0.27436 0.06623 Time: 0.00079
17-03-27 10:20:44 [1] Train Extra: lr=0.0001003 inv=0.4031250 sub=0.0000000
17-03-27 10:22:17 [1] Step: 38200 Acc: 0.61687 0.84820 Cost: 1.01993 0.80591 0.14772 0.06630 Time: 0.00084
17-03-27 10:22:17 [1] Train Extra: lr=0.0001000 inv=0.4071875 sub=0.0000000
17-03-27 10:23:39 [1] Step: 38300 Acc: 0.63375 0.85208 Cost: 1.05838 0.74974 0.24235 0.06630 Time: 0.00081
17-03-27 10:23:39 [1] Train Extra: lr=0.0000997 inv=0.3893750 sub=0.0000000
17-03-27 10:25:04 [1] Step: 38400 Acc: 0.62469 0.84101 Cost: 1.16635 0.90199 0.19811 0.06626 Time: 0.00079
17-03-27 10:25:04 [1] Train Extra: lr=0.0000994 inv=0.4123438 sub=0.0000000
17-03-27 10:26:33 [1] Step: 38500 Acc: 0.62594 0.84537 Cost: 1.05627 0.78421 0.20576 0.06629 Time: 0.00084
17-03-27 10:26:33 [1] Train Extra: lr=0.0000991 inv=0.4118750 sub=0.0000000
17-03-27 10:27:42 [1] Step: 38500 Eval acc: 0.65176 0.85328 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 10:27:42 [1] Eval Extra: inv=0.4226920
17-03-27 10:29:02 [1] Step: 38600 Acc: 0.61344 0.84418 Cost: 1.09512 0.78856 0.24035 0.06622 Time: 0.00079
17-03-27 10:29:02 [1] Train Extra: lr=0.0000988 inv=0.4017188 sub=0.0000000
17-03-27 10:30:23 [1] Step: 38700 Acc: 0.62562 0.85451 Cost: 0.77512 0.62495 0.08399 0.06619 Time: 0.00083
17-03-27 10:30:23 [1] Train Extra: lr=0.0000985 inv=0.3789062 sub=0.0000000
17-03-27 10:31:48 [1] Step: 38800 Acc: 0.63000 0.84792 Cost: 1.34500 0.97330 0.30549 0.06621 Time: 0.00082
17-03-27 10:31:48 [1] Train Extra: lr=0.0000983 inv=0.3828125 sub=0.0000000
17-03-27 10:33:19 [1] Step: 38900 Acc: 0.61844 0.84444 Cost: 1.07377 0.85720 0.15029 0.06628 Time: 0.00083
17-03-27 10:33:19 [1] Train Extra: lr=0.0000980 inv=0.4067188 sub=0.0000000
17-03-27 10:34:45 [1] Step: 39000 Acc: 0.63781 0.84698 Cost: 1.26756 0.93346 0.26780 0.06630 Time: 0.00080
17-03-27 10:34:45 [1] Train Extra: lr=0.0000977 inv=0.4029687 sub=0.0000000
17-03-27 10:35:56 [1] Step: 39000 Eval acc: 0.65114 0.85526 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 10:35:56 [1] Eval Extra: inv=0.3956291
17-03-27 10:37:30 [1] Step: 39100 Acc: 0.63469 0.84652 Cost: 1.16076 0.86750 0.22691 0.06635 Time: 0.00087
17-03-27 10:37:30 [1] Train Extra: lr=0.0000974 inv=0.4107812 sub=0.0000000
17-03-27 10:38:50 [1] Step: 39200 Acc: 0.66594 0.84936 Cost: 1.14200 0.84235 0.23305 0.06659 Time: 0.00080
17-03-27 10:38:50 [1] Train Extra: lr=0.0000971 inv=0.3784375 sub=0.0000000
17-03-27 10:40:25 [1] Step: 39300 Acc: 0.64125 0.85107 Cost: 1.51625 1.08123 0.36836 0.06666 Time: 0.00087
17-03-27 10:40:25 [1] Train Extra: lr=0.0000969 inv=0.4020313 sub=0.0000000
17-03-27 10:41:51 [1] Step: 39400 Acc: 0.63625 0.84738 Cost: 1.23910 0.86654 0.30578 0.06678 Time: 0.00081
17-03-27 10:41:51 [1] Train Extra: lr=0.0000966 inv=0.3760938 sub=0.0000000
17-03-27 10:43:21 [1] Step: 39500 Acc: 0.64906 0.85336 Cost: 0.91819 0.71077 0.14051 0.06691 Time: 0.00084
17-03-27 10:43:21 [1] Train Extra: lr=0.0000963 inv=0.4073438 sub=0.0000000
17-03-27 10:44:27 [1] Step: 39500 Eval acc: 0.65247 0.85603 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 10:44:27 [1] Eval Extra: inv=0.3651961
17-03-27 10:45:58 [1] Step: 39600 Acc: 0.65187 0.84813 Cost: 1.06311 0.75061 0.24552 0.06699 Time: 0.00083
17-03-27 10:45:58 [1] Train Extra: lr=0.0000960 inv=0.4145313 sub=0.0000000
17-03-27 10:47:36 [1] Step: 39700 Acc: 0.65125 0.84697 Cost: 1.15215 0.78513 0.29991 0.06711 Time: 0.00085
17-03-27 10:47:36 [1] Train Extra: lr=0.0000957 inv=0.4207812 sub=0.0000000
17-03-27 10:49:13 [1] Step: 39800 Acc: 0.65063 0.85177 Cost: 1.10211 0.82737 0.20759 0.06714 Time: 0.00087
17-03-27 10:49:13 [1] Train Extra: lr=0.0000955 inv=0.3928125 sub=0.0000000
17-03-27 10:50:38 [1] Step: 39900 Acc: 0.64187 0.84611 Cost: 1.04280 0.73919 0.23635 0.06726 Time: 0.00080
17-03-27 10:50:38 [1] Train Extra: lr=0.0000952 inv=0.3926562 sub=0.0000000
17-03-27 10:52:00 [1] Step: 40000 Acc: 0.64687 0.84646 Cost: 0.90123 0.72462 0.10921 0.06740 Time: 0.00081
17-03-27 10:52:00 [1] Train Extra: lr=0.0000949 inv=0.3807813 sub=0.0000000
17-03-27 10:53:10 [1] Step: 40000 Eval acc: 0.64951 0.85471 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 10:53:10 [1] Eval Extra: inv=0.4116115
17-03-27 10:53:10 [1] Checkpointing.
17-03-27 10:54:36 [1] Step: 40100 Acc: 0.64656 0.85326 Cost: 0.98186 0.66451 0.24979 0.06756 Time: 0.00081
17-03-27 10:54:36 [1] Train Extra: lr=0.0000946 inv=0.3981250 sub=0.0000000
17-03-27 10:56:00 [1] Step: 40200 Acc: 0.63781 0.84864 Cost: 1.13680 0.90820 0.16101 0.06758 Time: 0.00084
17-03-27 10:56:00 [1] Train Extra: lr=0.0000944 inv=0.3750000 sub=0.0000000
17-03-27 10:57:32 [1] Step: 40300 Acc: 0.64750 0.84433 Cost: 1.10779 0.81048 0.22966 0.06765 Time: 0.00084
17-03-27 10:57:32 [1] Train Extra: lr=0.0000941 inv=0.3993750 sub=0.0000000
17-03-27 10:58:51 [1] Step: 40400 Acc: 0.64656 0.85115 Cost: 1.17378 0.89398 0.21209 0.06771 Time: 0.00080
17-03-27 10:58:51 [1] Train Extra: lr=0.0000938 inv=0.3826562 sub=0.0000000
17-03-27 11:00:14 [1] Step: 40500 Acc: 0.64656 0.85485 Cost: 0.96152 0.75022 0.14349 0.06781 Time: 0.00083
17-03-27 11:00:14 [1] Train Extra: lr=0.0000936 inv=0.3842187 sub=0.0000000
17-03-27 11:01:23 [1] Step: 40500 Eval acc: 0.65431 0.85455 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 11:01:23 [1] Eval Extra: inv=0.3754085
17-03-27 11:02:48 [1] Step: 40600 Acc: 0.64469 0.85406 Cost: 1.18631 0.80837 0.31003 0.06791 Time: 0.00083
17-03-27 11:02:48 [1] Train Extra: lr=0.0000933 inv=0.3846875 sub=0.0000000
17-03-27 11:04:08 [1] Step: 40700 Acc: 0.65281 0.85673 Cost: 1.04233 0.83063 0.14370 0.06800 Time: 0.00082
17-03-27 11:04:08 [1] Train Extra: lr=0.0000930 inv=0.3706250 sub=0.0000000
17-03-27 11:05:37 [1] Step: 40800 Acc: 0.65625 0.84254 Cost: 1.00278 0.69355 0.24120 0.06802 Time: 0.00084
17-03-27 11:05:37 [1] Train Extra: lr=0.0000928 inv=0.3884375 sub=0.0000000
17-03-27 11:06:55 [1] Step: 40900 Acc: 0.63844 0.84440 Cost: 0.93047 0.74287 0.11953 0.06806 Time: 0.00077
17-03-27 11:06:55 [1] Train Extra: lr=0.0000925 inv=0.4106250 sub=0.0000000
17-03-27 11:08:17 [1] Step: 41000 Acc: 0.65031 0.84690 Cost: 0.97856 0.70749 0.20302 0.06806 Time: 0.00083
17-03-27 11:08:17 [1] Train Extra: lr=0.0000922 inv=0.3660937 sub=0.0000000
17-03-27 11:09:22 [1] Step: 41000 Eval acc: 0.65176 0.85346 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-27 11:09:22 [1] Eval Extra: inv=0.3865400
17-03-27 11:10:52 [1] Step: 41100 Acc: 0.64719 0.85503 Cost: 1.02671 0.78601 0.17262 0.06808 Time: 0.00084
17-03-27 11:10:52 [1] Train Extra: lr=0.0000920 inv=0.3826562 sub=0.0000000
17-03-27 11:12:22 [1] Step: 41200 Acc: 0.64375 0.83956 Cost: 0.99038 0.78178 0.14039 0.06820 Time: 0.00083
17-03-27 11:12:22 [1] Train Extra: lr=0.0000917 inv=0.4245313 sub=0.0000000
17-03-27 11:13:45 [1] Step: 41300 Acc: 0.64500 0.84825 Cost: 1.29427 1.03789 0.18814 0.06823 Time: 0.00084
17-03-27 11:13:45 [1] Train Extra: lr=0.0000914 inv=0.3662500 sub=0.0000000
17-03-27 11:15:20 [1] Step: 41400 Acc: 0.63156 0.85819 Cost: 0.96134 0.71535 0.17767 0.06832 Time: 0.00085
17-03-27 11:15:20 [1] Train Extra: lr=0.0000912 inv=0.4062500 sub=0.0000000
17-03-27 11:16:49 [1] Step: 41500 Acc: 0.63906 0.85279 Cost: 1.60210 1.21775 0.31597 0.06838 Time: 0.00086
17-03-27 11:16:49 [1] Train Extra: lr=0.0000909 inv=0.3868750 sub=0.0000000
17-03-27 11:17:56 [1] Step: 41500 Eval acc: 0.64737 0.85473 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 11:17:56 [1] Eval Extra: inv=0.4385212
17-03-27 11:19:35 [1] Step: 41600 Acc: 0.63375 0.85122 Cost: 0.88512 0.80429 0.01241 0.06843 Time: 0.00088
17-03-27 11:19:35 [1] Train Extra: lr=0.0000907 inv=0.4031250 sub=0.0000000
17-03-27 11:21:01 [1] Step: 41700 Acc: 0.65969 0.85124 Cost: 1.04251 0.77749 0.19652 0.06851 Time: 0.00082
17-03-27 11:21:01 [1] Train Extra: lr=0.0000904 inv=0.3690625 sub=0.0000000
17-03-27 11:22:39 [1] Step: 41800 Acc: 0.63594 0.85078 Cost: 1.26287 0.91781 0.27652 0.06853 Time: 0.00086
17-03-27 11:22:39 [1] Train Extra: lr=0.0000901 inv=0.4254688 sub=0.0000000
17-03-27 11:23:58 [1] Step: 41900 Acc: 0.65344 0.85046 Cost: 1.23928 0.87310 0.29756 0.06862 Time: 0.00080
17-03-27 11:23:58 [1] Train Extra: lr=0.0000899 inv=0.3840625 sub=0.0000000
17-03-27 11:25:20 [1] Step: 42000 Acc: 0.64469 0.84575 Cost: 1.07024 0.76332 0.23825 0.06867 Time: 0.00082
17-03-27 11:25:20 [1] Train Extra: lr=0.0000896 inv=0.4037500 sub=0.0000000
17-03-27 11:26:28 [1] Step: 42000 Eval acc: 0.64992 0.85623 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 11:26:28 [1] Eval Extra: inv=0.4107945
17-03-27 11:28:01 [1] Step: 42100 Acc: 0.63687 0.85732 Cost: 1.11589 0.84856 0.19862 0.06871 Time: 0.00085
17-03-27 11:28:01 [1] Train Extra: lr=0.0000894 inv=0.3745312 sub=0.0000000
17-03-27 11:29:41 [1] Step: 42200 Acc: 0.63844 0.84780 Cost: 1.12465 0.83300 0.22287 0.06878 Time: 0.00084
17-03-27 11:29:41 [1] Train Extra: lr=0.0000891 inv=0.4120313 sub=0.0000000
17-03-27 11:31:18 [1] Step: 42300 Acc: 0.62469 0.84841 Cost: 1.09765 0.86729 0.16154 0.06881 Time: 0.00085
17-03-27 11:31:18 [1] Train Extra: lr=0.0000888 inv=0.4176563 sub=0.0000000
17-03-27 11:32:43 [1] Step: 42400 Acc: 0.64250 0.85180 Cost: 1.39858 1.05269 0.27699 0.06891 Time: 0.00083
17-03-27 11:32:43 [1] Train Extra: lr=0.0000886 inv=0.3823437 sub=0.0000000
17-03-27 11:34:08 [1] Step: 42500 Acc: 0.64500 0.85487 Cost: 0.99737 0.71009 0.21836 0.06892 Time: 0.00084
17-03-27 11:34:08 [1] Train Extra: lr=0.0000883 inv=0.3715625 sub=0.0000000
17-03-27 11:35:13 [1] Step: 42500 Eval acc: 0.65891 0.85367 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 11:35:13 [1] Eval Extra: inv=0.3781148
17-03-27 11:35:13 [1] Checkpointing with new best dev accuracy of 0.658905
17-03-27 11:36:49 [1] Step: 42600 Acc: 0.64375 0.85074 Cost: 1.23031 0.85296 0.30842 0.06893 Time: 0.00086
17-03-27 11:36:49 [1] Train Extra: lr=0.0000881 inv=0.3985938 sub=0.0000000
17-03-27 11:38:13 [1] Step: 42700 Acc: 0.62687 0.85411 Cost: 1.04535 0.73539 0.24092 0.06903 Time: 0.00081
17-03-27 11:38:13 [1] Train Extra: lr=0.0000878 inv=0.3989063 sub=0.0000000
17-03-27 11:39:43 [1] Step: 42800 Acc: 0.63594 0.84626 Cost: 1.09268 0.78375 0.23988 0.06904 Time: 0.00082
17-03-27 11:39:43 [1] Train Extra: lr=0.0000876 inv=0.4104687 sub=0.0000000
17-03-27 11:41:28 [1] Step: 42900 Acc: 0.64250 0.84971 Cost: 1.21924 0.82058 0.32955 0.06910 Time: 0.00087
17-03-27 11:41:28 [1] Train Extra: lr=0.0000873 inv=0.4125000 sub=0.0000000
17-03-27 11:42:49 [1] Step: 43000 Acc: 0.64906 0.85107 Cost: 1.17008 0.88041 0.22057 0.06911 Time: 0.00083
17-03-27 11:42:49 [1] Train Extra: lr=0.0000871 inv=0.3868750 sub=0.0000000
17-03-27 11:43:54 [1] Step: 43000 Eval acc: 0.65513 0.84931 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 11:43:54 [1] Eval Extra: inv=0.3831189
17-03-27 11:45:19 [1] Step: 43100 Acc: 0.65281 0.84916 Cost: 1.12066 0.80300 0.24850 0.06915 Time: 0.00082
17-03-27 11:45:19 [1] Train Extra: lr=0.0000868 inv=0.3935938 sub=0.0000000
17-03-27 11:46:49 [1] Step: 43200 Acc: 0.64531 0.84729 Cost: 0.80764 0.64165 0.09680 0.06920 Time: 0.00083
17-03-27 11:46:49 [1] Train Extra: lr=0.0000866 inv=0.3828125 sub=0.0000000
17-03-27 11:48:20 [1] Step: 43300 Acc: 0.63594 0.84588 Cost: 0.92449 0.78929 0.06602 0.06917 Time: 0.00083
17-03-27 11:48:20 [1] Train Extra: lr=0.0000863 inv=0.4287500 sub=0.0000000
17-03-27 11:49:40 [1] Step: 43400 Acc: 0.64406 0.84550 Cost: 1.00411 0.75908 0.17581 0.06921 Time: 0.00080
17-03-27 11:49:40 [1] Train Extra: lr=0.0000861 inv=0.3781250 sub=0.0000000
17-03-27 11:51:05 [1] Step: 43500 Acc: 0.65594 0.85138 Cost: 1.21494 0.94659 0.19911 0.06924 Time: 0.00083
17-03-27 11:51:05 [1] Train Extra: lr=0.0000858 inv=0.3875000 sub=0.0000000
17-03-27 11:52:13 [1] Step: 43500 Eval acc: 0.65247 0.84927 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 11:52:13 [1] Eval Extra: inv=0.3650940
17-03-27 11:53:37 [1] Step: 43600 Acc: 0.65094 0.85115 Cost: 0.86083 0.67954 0.11195 0.06934 Time: 0.00081
17-03-27 11:53:37 [1] Train Extra: lr=0.0000856 inv=0.3901562 sub=0.0000000
17-03-27 11:55:07 [1] Step: 43700 Acc: 0.62187 0.84424 Cost: 1.37293 0.96043 0.34323 0.06927 Time: 0.00083
17-03-27 11:55:07 [1] Train Extra: lr=0.0000853 inv=0.3979687 sub=0.0000000
17-03-27 11:56:33 [1] Step: 43800 Acc: 0.64094 0.85129 Cost: 1.22818 0.91080 0.24801 0.06937 Time: 0.00081
17-03-27 11:56:33 [1] Train Extra: lr=0.0000851 inv=0.3737500 sub=0.0000000
17-03-27 11:58:10 [1] Step: 43900 Acc: 0.61250 0.84417 Cost: 1.19423 0.81753 0.30732 0.06938 Time: 0.00083
17-03-27 11:58:10 [1] Train Extra: lr=0.0000848 inv=0.4231250 sub=0.0000000
17-03-27 11:59:27 [1] Step: 44000 Acc: 0.63875 0.84596 Cost: 0.92092 0.66751 0.18407 0.06934 Time: 0.00078
17-03-27 11:59:27 [1] Train Extra: lr=0.0000846 inv=0.3912500 sub=0.0000000
17-03-27 12:00:37 [1] Step: 44000 Eval acc: 0.65431 0.85341 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 12:00:37 [1] Eval Extra: inv=0.3877145
17-03-27 12:02:09 [1] Step: 44100 Acc: 0.63469 0.85931 Cost: 0.89482 0.69389 0.13163 0.06929 Time: 0.00085
17-03-27 12:02:09 [1] Train Extra: lr=0.0000844 inv=0.3954687 sub=0.0000000
17-03-27 12:03:31 [1] Step: 44200 Acc: 0.62281 0.84568 Cost: 0.99970 0.65903 0.27141 0.06925 Time: 0.00080
17-03-27 12:03:31 [1] Train Extra: lr=0.0000841 inv=0.4065625 sub=0.0000000
17-03-27 12:04:55 [1] Step: 44300 Acc: 0.65156 0.84619 Cost: 1.15929 0.78767 0.30225 0.06938 Time: 0.00079
17-03-27 12:04:55 [1] Train Extra: lr=0.0000839 inv=0.3818750 sub=0.0000000
17-03-27 12:06:15 [1] Step: 44400 Acc: 0.65187 0.84690 Cost: 0.92661 0.58854 0.26869 0.06939 Time: 0.00080
17-03-27 12:06:15 [1] Train Extra: lr=0.0000836 inv=0.3907813 sub=0.0000000
17-03-27 12:07:40 [1] Step: 44500 Acc: 0.63687 0.85218 Cost: 1.18768 0.84693 0.27135 0.06940 Time: 0.00083
17-03-27 12:07:40 [1] Train Extra: lr=0.0000834 inv=0.3803125 sub=0.0000000
17-03-27 12:08:50 [1] Step: 44500 Eval acc: 0.65063 0.86049 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 12:08:50 [1] Eval Extra: inv=0.3280739
17-03-27 12:10:16 [1] Step: 44600 Acc: 0.63438 0.85046 Cost: 0.95975 0.65766 0.23268 0.06942 Time: 0.00080
17-03-27 12:10:16 [1] Train Extra: lr=0.0000832 inv=0.3937500 sub=0.0000000
17-03-27 12:11:46 [1] Step: 44700 Acc: 0.64094 0.85135 Cost: 1.26078 0.99620 0.19509 0.06948 Time: 0.00086
17-03-27 12:11:46 [1] Train Extra: lr=0.0000829 inv=0.4021875 sub=0.0000000
17-03-27 12:13:16 [1] Step: 44800 Acc: 0.63531 0.85161 Cost: 1.13304 0.82960 0.23395 0.06949 Time: 0.00083
17-03-27 12:13:16 [1] Train Extra: lr=0.0000827 inv=0.4070313 sub=0.0000000
17-03-27 12:14:33 [1] Step: 44900 Acc: 0.65719 0.84736 Cost: 0.91034 0.77698 0.06388 0.06948 Time: 0.00077
17-03-27 12:14:33 [1] Train Extra: lr=0.0000824 inv=0.3814063 sub=0.0000000
17-03-27 12:16:11 [1] Step: 45000 Acc: 0.64750 0.85101 Cost: 0.94356 0.71840 0.15561 0.06955 Time: 0.00088
17-03-27 12:16:11 [1] Train Extra: lr=0.0000822 inv=0.3898437 sub=0.0000000
17-03-27 12:17:16 [1] Step: 45000 Eval acc: 0.64849 0.85155 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 12:17:16 [1] Eval Extra: inv=0.3775531
17-03-27 12:17:16 [1] Checkpointing.
17-03-27 12:18:47 [1] Step: 45100 Acc: 0.63313 0.85142 Cost: 1.26895 0.88759 0.31179 0.06958 Time: 0.00083
17-03-27 12:18:47 [1] Train Extra: lr=0.0000820 inv=0.4053125 sub=0.0000000
17-03-27 12:20:18 [1] Step: 45200 Acc: 0.63531 0.84457 Cost: 0.93517 0.67442 0.19123 0.06952 Time: 0.00083
17-03-27 12:20:18 [1] Train Extra: lr=0.0000817 inv=0.3903125 sub=0.0000000
17-03-27 12:21:49 [1] Step: 45300 Acc: 0.65250 0.84442 Cost: 1.21872 0.90510 0.24403 0.06958 Time: 0.00083
17-03-27 12:21:49 [1] Train Extra: lr=0.0000815 inv=0.4200000 sub=0.0000000
17-03-27 12:23:20 [1] Step: 45400 Acc: 0.64844 0.84739 Cost: 0.89840 0.67641 0.15236 0.06963 Time: 0.00083
17-03-27 12:23:20 [1] Train Extra: lr=0.0000813 inv=0.4007812 sub=0.0000000
17-03-27 12:24:45 [1] Step: 45500 Acc: 0.63375 0.83947 Cost: 1.24986 0.98679 0.19347 0.06960 Time: 0.00080
17-03-27 12:24:45 [1] Train Extra: lr=0.0000810 inv=0.4120313 sub=0.0000000
17-03-27 12:25:55 [1] Step: 45500 Eval acc: 0.64971 0.85327 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 12:25:55 [1] Eval Extra: inv=0.4027778
17-03-27 12:27:06 [1] Step: 45600 Acc: 0.64187 0.85433 Cost: 1.07241 0.81700 0.18577 0.06964 Time: 0.00078
17-03-27 12:27:06 [1] Train Extra: lr=0.0000808 inv=0.3657813 sub=0.0000000
17-03-27 12:28:37 [1] Step: 45700 Acc: 0.64406 0.84293 Cost: 1.06288 0.86438 0.12881 0.06970 Time: 0.00080
17-03-27 12:28:37 [1] Train Extra: lr=0.0000806 inv=0.4145313 sub=0.0000000
17-03-27 12:30:00 [1] Step: 45800 Acc: 0.64312 0.84362 Cost: 1.04416 0.74875 0.22577 0.06964 Time: 0.00081
17-03-27 12:30:00 [1] Train Extra: lr=0.0000803 inv=0.3928125 sub=0.0000000
17-03-27 12:31:22 [1] Step: 45900 Acc: 0.65625 0.84712 Cost: 1.01695 0.83155 0.11576 0.06965 Time: 0.00081
17-03-27 12:31:22 [1] Train Extra: lr=0.0000801 inv=0.3890625 sub=0.0000000
17-03-27 12:32:47 [1] Step: 46000 Acc: 0.62813 0.84964 Cost: 1.36790 1.12009 0.17811 0.06970 Time: 0.00081
17-03-27 12:32:47 [1] Train Extra: lr=0.0000799 inv=0.3870312 sub=0.0000000
17-03-27 12:33:57 [1] Step: 46000 Eval acc: 0.65645 0.84973 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 12:33:57 [1] Eval Extra: inv=0.4102328
17-03-27 12:35:30 [1] Step: 46100 Acc: 0.63500 0.85227 Cost: 1.30776 0.94941 0.28871 0.06964 Time: 0.00082
17-03-27 12:35:30 [1] Train Extra: lr=0.0000796 inv=0.4185937 sub=0.0000000
17-03-27 12:36:59 [1] Step: 46200 Acc: 0.65156 0.84536 Cost: 0.99572 0.73155 0.19440 0.06976 Time: 0.00083
17-03-27 12:36:59 [1] Train Extra: lr=0.0000794 inv=0.4098438 sub=0.0000000
17-03-27 12:38:22 [1] Step: 46300 Acc: 0.64406 0.84744 Cost: 1.05153 0.77041 0.21132 0.06980 Time: 0.00082
17-03-27 12:38:22 [1] Train Extra: lr=0.0000792 inv=0.3775000 sub=0.0000000
17-03-27 12:39:48 [1] Step: 46400 Acc: 0.63656 0.84861 Cost: 1.01040 0.80769 0.13285 0.06985 Time: 0.00080
17-03-27 12:39:48 [1] Train Extra: lr=0.0000790 inv=0.4146875 sub=0.0000000
17-03-27 12:41:13 [1] Step: 46500 Acc: 0.64563 0.84666 Cost: 0.95765 0.76452 0.12329 0.06984 Time: 0.00081
17-03-27 12:41:13 [1] Train Extra: lr=0.0000787 inv=0.3959375 sub=0.0000000
17-03-27 12:42:23 [1] Step: 46500 Eval acc: 0.65850 0.84946 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 12:42:23 [1] Eval Extra: inv=0.4112030
17-03-27 12:43:59 [1] Step: 46600 Acc: 0.63562 0.85120 Cost: 0.94382 0.74616 0.12782 0.06984 Time: 0.00085
17-03-27 12:43:59 [1] Train Extra: lr=0.0000785 inv=0.3750000 sub=0.0000000
17-03-27 12:45:30 [1] Step: 46700 Acc: 0.63844 0.84475 Cost: 0.99545 0.79556 0.12997 0.06993 Time: 0.00084
17-03-27 12:45:30 [1] Train Extra: lr=0.0000783 inv=0.3832813 sub=0.0000000
17-03-27 12:46:56 [1] Step: 46800 Acc: 0.62687 0.84464 Cost: 1.24470 0.92420 0.25050 0.06999 Time: 0.00082
17-03-27 12:46:56 [1] Train Extra: lr=0.0000781 inv=0.3928125 sub=0.0000000
17-03-27 12:48:28 [1] Step: 46900 Acc: 0.64906 0.83932 Cost: 1.31911 1.05190 0.19724 0.06997 Time: 0.00084
17-03-27 12:48:28 [1] Train Extra: lr=0.0000778 inv=0.3862500 sub=0.0000000
17-03-27 12:49:54 [1] Step: 47000 Acc: 0.63313 0.84918 Cost: 1.27933 0.91016 0.29917 0.07000 Time: 0.00080
17-03-27 12:49:54 [1] Train Extra: lr=0.0000776 inv=0.4082812 sub=0.0000000
17-03-27 12:51:03 [1] Step: 47000 Eval acc: 0.66391 0.86003 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 12:51:03 [1] Eval Extra: inv=0.3357843
17-03-27 12:51:03 [1] Checkpointing with new best dev accuracy of 0.663909
17-03-27 12:52:35 [1] Step: 47100 Acc: 0.64531 0.85846 Cost: 1.11178 0.83441 0.20728 0.07010 Time: 0.00085
17-03-27 12:52:35 [1] Train Extra: lr=0.0000774 inv=0.3987500 sub=0.0000000
17-03-27 12:54:11 [1] Step: 47200 Acc: 0.63750 0.84099 Cost: 1.04208 0.74250 0.22942 0.07016 Time: 0.00083
17-03-27 12:54:11 [1] Train Extra: lr=0.0000772 inv=0.4146875 sub=0.0000000
17-03-27 12:55:29 [1] Step: 47300 Acc: 0.64938 0.85097 Cost: 0.70195 0.51700 0.11478 0.07017 Time: 0.00079
17-03-27 12:55:29 [1] Train Extra: lr=0.0000769 inv=0.3593750 sub=0.0000000
17-03-27 12:56:50 [1] Step: 47400 Acc: 0.64062 0.84854 Cost: 0.92392 0.65804 0.19576 0.07012 Time: 0.00081
17-03-27 12:56:50 [1] Train Extra: lr=0.0000767 inv=0.3835938 sub=0.0000000
17-03-27 12:58:22 [1] Step: 47500 Acc: 0.64125 0.85200 Cost: 0.99163 0.73562 0.18594 0.07007 Time: 0.00085
17-03-27 12:58:22 [1] Train Extra: lr=0.0000765 inv=0.4160937 sub=0.0000000
17-03-27 12:59:32 [1] Step: 47500 Eval acc: 0.65962 0.85525 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 12:59:32 [1] Eval Extra: inv=0.3616728
17-03-27 13:01:03 [1] Step: 47600 Acc: 0.63375 0.85142 Cost: 1.58461 1.24032 0.27421 0.07008 Time: 0.00084
17-03-27 13:01:03 [1] Train Extra: lr=0.0000763 inv=0.4042188 sub=0.0000000
17-03-27 13:02:29 [1] Step: 47700 Acc: 0.63781 0.84808 Cost: 1.15676 0.87511 0.21151 0.07014 Time: 0.00080
17-03-27 13:02:29 [1] Train Extra: lr=0.0000761 inv=0.4040625 sub=0.0000000
17-03-27 13:03:57 [1] Step: 47800 Acc: 0.63938 0.84842 Cost: 1.03080 0.74183 0.21887 0.07010 Time: 0.00084
17-03-27 13:03:57 [1] Train Extra: lr=0.0000758 inv=0.3715625 sub=0.0000000
17-03-27 13:05:23 [1] Step: 47900 Acc: 0.63187 0.84273 Cost: 0.93143 0.71093 0.15038 0.07012 Time: 0.00079
17-03-27 13:05:23 [1] Train Extra: lr=0.0000756 inv=0.4043750 sub=0.0000000
17-03-27 13:06:53 [1] Step: 48000 Acc: 0.63750 0.85229 Cost: 1.42332 1.16431 0.18889 0.07011 Time: 0.00084
17-03-27 13:06:53 [1] Train Extra: lr=0.0000754 inv=0.3851562 sub=0.0000000
17-03-27 13:08:02 [1] Step: 48000 Eval acc: 0.66238 0.85557 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 13:08:02 [1] Eval Extra: inv=0.3904718
17-03-27 13:09:39 [1] Step: 48100 Acc: 0.64344 0.86174 Cost: 1.17541 0.98908 0.11624 0.07010 Time: 0.00088
17-03-27 13:09:39 [1] Train Extra: lr=0.0000752 inv=0.3801562 sub=0.0000000
17-03-27 13:11:24 [1] Step: 48200 Acc: 0.64750 0.85313 Cost: 1.42694 1.08247 0.27429 0.07019 Time: 0.00086
17-03-27 13:11:24 [1] Train Extra: lr=0.0000750 inv=0.4478125 sub=0.0000000
17-03-27 13:12:56 [1] Step: 48300 Acc: 0.63813 0.85169 Cost: 1.00658 0.75077 0.18561 0.07020 Time: 0.00085
17-03-27 13:12:56 [1] Train Extra: lr=0.0000748 inv=0.3985938 sub=0.0000000
17-03-27 13:14:30 [1] Step: 48400 Acc: 0.64000 0.84463 Cost: 1.06671 0.78257 0.21392 0.07022 Time: 0.00082
17-03-27 13:14:30 [1] Train Extra: lr=0.0000745 inv=0.4248438 sub=0.0000000
17-03-27 13:16:07 [1] Step: 48500 Acc: 0.63313 0.84945 Cost: 1.30901 0.99136 0.24751 0.07014 Time: 0.00086
17-03-27 13:16:07 [1] Train Extra: lr=0.0000743 inv=0.4165625 sub=0.0000000
17-03-27 13:17:16 [1] Step: 48500 Eval acc: 0.65880 0.85420 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 13:17:16 [1] Eval Extra: inv=0.3779616
17-03-27 13:18:48 [1] Step: 48600 Acc: 0.64531 0.84546 Cost: 0.96243 0.66925 0.22305 0.07013 Time: 0.00082
17-03-27 13:18:48 [1] Train Extra: lr=0.0000741 inv=0.4150000 sub=0.0000000
17-03-27 13:20:26 [1] Step: 48700 Acc: 0.63594 0.84768 Cost: 1.02450 0.78063 0.17367 0.07020 Time: 0.00083
17-03-27 13:20:26 [1] Train Extra: lr=0.0000739 inv=0.4445312 sub=0.0000000
17-03-27 13:21:57 [1] Step: 48800 Acc: 0.64719 0.85143 Cost: 1.07372 0.88438 0.11914 0.07019 Time: 0.00083
17-03-27 13:21:57 [1] Train Extra: lr=0.0000737 inv=0.3923437 sub=0.0000000
17-03-27 13:23:17 [1] Step: 48900 Acc: 0.65500 0.84642 Cost: 1.08399 0.85263 0.16110 0.07026 Time: 0.00082
17-03-27 13:23:17 [1] Train Extra: lr=0.0000735 inv=0.3675000 sub=0.0000000
17-03-27 13:24:42 [1] Step: 49000 Acc: 0.66063 0.85355 Cost: 1.04263 0.78371 0.18855 0.07037 Time: 0.00083
17-03-27 13:24:42 [1] Train Extra: lr=0.0000733 inv=0.3728125 sub=0.0000000
17-03-27 13:25:51 [1] Step: 49000 Eval acc: 0.65676 0.85486 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 13:25:51 [1] Eval Extra: inv=0.3593239
17-03-27 13:27:16 [1] Step: 49100 Acc: 0.65687 0.84738 Cost: 0.79016 0.59657 0.12312 0.07047 Time: 0.00081
17-03-27 13:27:16 [1] Train Extra: lr=0.0000731 inv=0.3906250 sub=0.0000000
17-03-27 13:28:46 [1] Step: 49200 Acc: 0.65812 0.84982 Cost: 0.93693 0.76233 0.10404 0.07056 Time: 0.00085
17-03-27 13:28:46 [1] Train Extra: lr=0.0000728 inv=0.3990625 sub=0.0000000
17-03-27 13:30:03 [1] Step: 49300 Acc: 0.66063 0.84703 Cost: 0.80607 0.65252 0.08289 0.07067 Time: 0.00079
17-03-27 13:30:03 [1] Train Extra: lr=0.0000726 inv=0.3851562 sub=0.0000000
17-03-27 13:31:44 [1] Step: 49400 Acc: 0.64594 0.84957 Cost: 1.12228 0.82006 0.23149 0.07073 Time: 0.00085
17-03-27 13:31:44 [1] Train Extra: lr=0.0000724 inv=0.4228125 sub=0.0000000
17-03-27 13:33:15 [1] Step: 49500 Acc: 0.66031 0.84654 Cost: 1.16454 0.83000 0.26364 0.07090 Time: 0.00082
17-03-27 13:33:15 [1] Train Extra: lr=0.0000722 inv=0.4309375 sub=0.0000000
17-03-27 13:34:23 [1] Step: 49500 Eval acc: 0.65462 0.85508 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 13:34:23 [1] Eval Extra: inv=0.4298407
17-03-27 13:35:56 [1] Step: 49600 Acc: 0.64438 0.84947 Cost: 1.24291 0.88236 0.28958 0.07097 Time: 0.00082
17-03-27 13:35:56 [1] Train Extra: lr=0.0000720 inv=0.4176563 sub=0.0000000
17-03-27 13:37:22 [1] Step: 49700 Acc: 0.67375 0.84444 Cost: 1.11466 0.81691 0.22671 0.07104 Time: 0.00081
17-03-27 13:37:22 [1] Train Extra: lr=0.0000718 inv=0.4068750 sub=0.0000000
17-03-27 13:38:54 [1] Step: 49800 Acc: 0.64656 0.84975 Cost: 1.09941 0.81066 0.21762 0.07112 Time: 0.00084
17-03-27 13:38:54 [1] Train Extra: lr=0.0000716 inv=0.4120313 sub=0.0000000
17-03-27 13:40:24 [1] Step: 49900 Acc: 0.64469 0.84813 Cost: 1.18348 0.85940 0.25289 0.07119 Time: 0.00083
17-03-27 13:40:24 [1] Train Extra: lr=0.0000714 inv=0.4114062 sub=0.0000000
17-03-27 13:41:56 [1] Step: 50000 Acc: 0.64000 0.84421 Cost: 1.18561 0.80947 0.30496 0.07119 Time: 0.00082
17-03-27 13:41:56 [1] Train Extra: lr=0.0000712 inv=0.4257812 sub=0.0000000
17-03-27 13:43:05 [1] Step: 50000 Eval acc: 0.65952 0.85791 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 13:43:05 [1] Eval Extra: inv=0.4427083
17-03-27 13:43:05 [1] Checkpointing.
17-03-27 13:44:25 [1] Step: 50100 Acc: 0.65094 0.84822 Cost: 1.06589 0.77929 0.21530 0.07130 Time: 0.00078
17-03-27 13:44:25 [1] Train Extra: lr=0.0000710 inv=0.4042188 sub=0.0000000
17-03-27 13:45:48 [1] Step: 50200 Acc: 0.66406 0.85490 Cost: 1.21483 0.86490 0.27856 0.07137 Time: 0.00083
17-03-27 13:45:48 [1] Train Extra: lr=0.0000708 inv=0.3878125 sub=0.0000000
17-03-27 13:47:18 [1] Step: 50300 Acc: 0.66156 0.85064 Cost: 1.16104 0.76234 0.32724 0.07146 Time: 0.00084
17-03-27 13:47:18 [1] Train Extra: lr=0.0000706 inv=0.3995313 sub=0.0000000
17-03-27 13:48:49 [1] Step: 50400 Acc: 0.66312 0.85434 Cost: 1.03045 0.71154 0.24743 0.07148 Time: 0.00083
17-03-27 13:48:49 [1] Train Extra: lr=0.0000704 inv=0.3971875 sub=0.0000000
17-03-27 13:50:18 [1] Step: 50500 Acc: 0.63406 0.85418 Cost: 1.07699 0.84695 0.15844 0.07160 Time: 0.00085
17-03-27 13:50:18 [1] Train Extra: lr=0.0000702 inv=0.4034375 sub=0.0000000
17-03-27 13:51:23 [1] Step: 50500 Eval acc: 0.65359 0.85319 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-27 13:51:23 [1] Eval Extra: inv=0.4181475
17-03-27 13:52:45 [1] Step: 50600 Acc: 0.64687 0.84301 Cost: 0.86400 0.62893 0.16339 0.07167 Time: 0.00080
17-03-27 13:52:45 [1] Train Extra: lr=0.0000700 inv=0.3914063 sub=0.0000000
17-03-27 13:54:16 [1] Step: 50700 Acc: 0.65281 0.85780 Cost: 1.13021 0.87529 0.18318 0.07174 Time: 0.00085
17-03-27 13:54:16 [1] Train Extra: lr=0.0000698 inv=0.3935938 sub=0.0000000
17-03-27 13:55:40 [1] Step: 50800 Acc: 0.65656 0.84353 Cost: 1.01828 0.77069 0.17579 0.07180 Time: 0.00080
17-03-27 13:55:40 [1] Train Extra: lr=0.0000696 inv=0.3956250 sub=0.0000000
17-03-27 13:57:09 [1] Step: 50900 Acc: 0.64687 0.85003 Cost: 1.38478 1.04802 0.26483 0.07192 Time: 0.00083
17-03-27 13:57:09 [1] Train Extra: lr=0.0000694 inv=0.4176563 sub=0.0000000
17-03-27 13:58:36 [1] Step: 51000 Acc: 0.65656 0.85152 Cost: 0.98563 0.72069 0.19294 0.07200 Time: 0.00081
17-03-27 13:58:36 [1] Train Extra: lr=0.0000692 inv=0.3950000 sub=0.0000000
17-03-27 13:59:47 [1] Step: 51000 Eval acc: 0.65839 0.85457 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 13:59:47 [1] Eval Extra: inv=0.3536050
17-03-27 14:01:11 [1] Step: 51100 Acc: 0.66031 0.85031 Cost: 0.92373 0.72990 0.12172 0.07210 Time: 0.00085
17-03-27 14:01:11 [1] Train Extra: lr=0.0000690 inv=0.3584375 sub=0.0000000
17-03-27 14:02:35 [1] Step: 51200 Acc: 0.64375 0.85079 Cost: 1.22967 0.87927 0.27819 0.07221 Time: 0.00080
17-03-27 14:02:35 [1] Train Extra: lr=0.0000688 inv=0.3984375 sub=0.0000000
17-03-27 14:04:09 [1] Step: 51300 Acc: 0.66594 0.83676 Cost: 1.29015 0.95562 0.26232 0.07222 Time: 0.00080
17-03-27 14:04:09 [1] Train Extra: lr=0.0000686 inv=0.4314062 sub=0.0000000
17-03-27 14:05:39 [1] Step: 51400 Acc: 0.65625 0.84410 Cost: 0.92078 0.65528 0.19319 0.07231 Time: 0.00083
17-03-27 14:05:39 [1] Train Extra: lr=0.0000684 inv=0.4217187 sub=0.0000000
17-03-27 14:07:07 [1] Step: 51500 Acc: 0.64219 0.85409 Cost: 1.40475 1.06706 0.26534 0.07235 Time: 0.00082
17-03-27 14:07:07 [1] Train Extra: lr=0.0000682 inv=0.3835938 sub=0.0000000
17-03-27 14:08:18 [1] Step: 51500 Eval acc: 0.65645 0.85408 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 14:08:18 [1] Eval Extra: inv=0.3718852
17-03-27 14:09:54 [1] Step: 51600 Acc: 0.65687 0.85097 Cost: 0.98177 0.65201 0.25735 0.07241 Time: 0.00086
17-03-27 14:09:54 [1] Train Extra: lr=0.0000680 inv=0.4179688 sub=0.0000000
17-03-27 14:11:21 [1] Step: 51700 Acc: 0.64094 0.85242 Cost: 1.04394 0.64457 0.32689 0.07248 Time: 0.00082
17-03-27 14:11:21 [1] Train Extra: lr=0.0000678 inv=0.3920312 sub=0.0000000
17-03-27 14:12:45 [1] Step: 51800 Acc: 0.64500 0.84710 Cost: 0.90931 0.70289 0.13397 0.07244 Time: 0.00079
17-03-27 14:12:45 [1] Train Extra: lr=0.0000676 inv=0.3903125 sub=0.0000000
17-03-27 14:14:00 [1] Step: 51900 Acc: 0.66250 0.84935 Cost: 0.85697 0.62852 0.15601 0.07244 Time: 0.00079
17-03-27 14:14:00 [1] Train Extra: lr=0.0000674 inv=0.3737500 sub=0.0000000
17-03-27 14:15:22 [1] Step: 52000 Acc: 0.64781 0.85475 Cost: 0.90552 0.68619 0.14681 0.07252 Time: 0.00083
17-03-27 14:15:22 [1] Train Extra: lr=0.0000672 inv=0.3815625 sub=0.0000000
17-03-27 14:16:27 [1] Step: 52000 Eval acc: 0.66217 0.85428 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-27 14:16:27 [1] Eval Extra: inv=0.3522774
17-03-27 14:17:51 [1] Step: 52100 Acc: 0.64187 0.84647 Cost: 1.09054 0.78128 0.23664 0.07262 Time: 0.00081
17-03-27 14:17:51 [1] Train Extra: lr=0.0000670 inv=0.4029687 sub=0.0000000
17-03-27 14:19:32 [1] Step: 52200 Acc: 0.65406 0.85228 Cost: 0.92641 0.75624 0.09749 0.07268 Time: 0.00088
17-03-27 14:19:32 [1] Train Extra: lr=0.0000668 inv=0.4168750 sub=0.0000000
17-03-27 14:21:04 [1] Step: 52300 Acc: 0.65469 0.85045 Cost: 1.19867 0.84672 0.27917 0.07278 Time: 0.00084
17-03-27 14:21:04 [1] Train Extra: lr=0.0000666 inv=0.4096875 sub=0.0000000
17-03-27 14:22:32 [1] Step: 52400 Acc: 0.64687 0.86170 Cost: 1.12935 0.93672 0.11984 0.07279 Time: 0.00083
17-03-27 14:22:32 [1] Train Extra: lr=0.0000664 inv=0.4109375 sub=0.0000000
17-03-27 14:24:01 [1] Step: 52500 Acc: 0.65312 0.85293 Cost: 0.84198 0.61207 0.15705 0.07286 Time: 0.00087
17-03-27 14:24:01 [1] Train Extra: lr=0.0000663 inv=0.3812500 sub=0.0000000
17-03-27 14:25:09 [1] Step: 52500 Eval acc: 0.65574 0.85438 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 14:25:09 [1] Eval Extra: inv=0.4034416
17-03-27 14:26:34 [1] Step: 52600 Acc: 0.65531 0.84807 Cost: 1.08733 0.78250 0.23189 0.07294 Time: 0.00079
17-03-27 14:26:34 [1] Train Extra: lr=0.0000661 inv=0.4201563 sub=0.0000000
17-03-27 14:28:03 [1] Step: 52700 Acc: 0.65375 0.85659 Cost: 1.23411 0.86794 0.29323 0.07295 Time: 0.00086
17-03-27 14:28:03 [1] Train Extra: lr=0.0000659 inv=0.3779688 sub=0.0000000
17-03-27 14:29:21 [1] Step: 52800 Acc: 0.67688 0.84703 Cost: 0.82899 0.62690 0.12913 0.07295 Time: 0.00079
17-03-27 14:29:21 [1] Train Extra: lr=0.0000657 inv=0.3973437 sub=0.0000000
17-03-27 14:30:44 [1] Step: 52900 Acc: 0.63594 0.85222 Cost: 1.16632 0.81847 0.27485 0.07300 Time: 0.00082
17-03-27 14:30:44 [1] Train Extra: lr=0.0000655 inv=0.3725000 sub=0.0000000
17-03-27 14:32:12 [1] Step: 53000 Acc: 0.65438 0.84162 Cost: 1.00565 0.70251 0.23000 0.07315 Time: 0.00083
17-03-27 14:32:12 [1] Train Extra: lr=0.0000653 inv=0.4018750 sub=0.0000000
17-03-27 14:33:17 [1] Step: 53000 Eval acc: 0.66207 0.85947 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-27 14:33:17 [1] Eval Extra: inv=0.3829657
17-03-27 14:34:53 [1] Step: 53100 Acc: 0.63875 0.85788 Cost: 1.16443 0.87068 0.22058 0.07318 Time: 0.00087
17-03-27 14:34:53 [1] Train Extra: lr=0.0000651 inv=0.4162500 sub=0.0000000
17-03-27 14:36:19 [1] Step: 53200 Acc: 0.65438 0.85098 Cost: 0.80033 0.60667 0.12042 0.07324 Time: 0.00081
17-03-27 14:36:19 [1] Train Extra: lr=0.0000649 inv=0.3867188 sub=0.0000000
17-03-27 14:37:43 [1] Step: 53300 Acc: 0.64687 0.85320 Cost: 1.11516 0.78383 0.25808 0.07325 Time: 0.00083
17-03-27 14:37:43 [1] Train Extra: lr=0.0000647 inv=0.3756250 sub=0.0000000
17-03-27 14:38:58 [1] Step: 53400 Acc: 0.67844 0.85875 Cost: 1.00267 0.69624 0.23317 0.07325 Time: 0.00081
17-03-27 14:38:58 [1] Train Extra: lr=0.0000646 inv=0.3687500 sub=0.0000000
17-03-27 14:40:30 [1] Step: 53500 Acc: 0.65031 0.85232 Cost: 1.00204 0.74236 0.18635 0.07333 Time: 0.00086
17-03-27 14:40:30 [1] Train Extra: lr=0.0000644 inv=0.3828125 sub=0.0000000
17-03-27 14:41:37 [1] Step: 53500 Eval acc: 0.66074 0.85783 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 14:41:37 [1] Eval Extra: inv=0.3596303
17-03-27 14:43:06 [1] Step: 53600 Acc: 0.65438 0.85401 Cost: 1.26495 0.92603 0.26560 0.07332 Time: 0.00085
17-03-27 14:43:06 [1] Train Extra: lr=0.0000642 inv=0.3854688 sub=0.0000000
17-03-27 14:44:35 [1] Step: 53700 Acc: 0.65687 0.85044 Cost: 0.78143 0.58543 0.12264 0.07337 Time: 0.00084
17-03-27 14:44:35 [1] Train Extra: lr=0.0000640 inv=0.3967188 sub=0.0000000
17-03-27 14:45:54 [1] Step: 53800 Acc: 0.64812 0.84647 Cost: 0.83601 0.55840 0.20419 0.07341 Time: 0.00077
17-03-27 14:45:54 [1] Train Extra: lr=0.0000638 inv=0.4125000 sub=0.0000000
17-03-27 14:47:33 [1] Step: 53900 Acc: 0.66312 0.84601 Cost: 1.05645 0.83481 0.14823 0.07341 Time: 0.00085
17-03-27 14:47:33 [1] Train Extra: lr=0.0000636 inv=0.4028125 sub=0.0000000
17-03-27 14:49:04 [1] Step: 54000 Acc: 0.64687 0.84290 Cost: 1.10690 0.75444 0.27895 0.07351 Time: 0.00082
17-03-27 14:49:04 [1] Train Extra: lr=0.0000635 inv=0.4131250 sub=0.0000000
17-03-27 14:50:13 [1] Step: 54000 Eval acc: 0.66085 0.85984 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 14:50:13 [1] Eval Extra: inv=0.2589359
17-03-27 14:51:42 [1] Step: 54100 Acc: 0.67750 0.85009 Cost: 1.11173 0.87669 0.16148 0.07356 Time: 0.00083
17-03-27 14:51:42 [1] Train Extra: lr=0.0000633 inv=0.3939063 sub=0.0000000
17-03-27 14:53:07 [1] Step: 54200 Acc: 0.66375 0.86352 Cost: 0.93746 0.78387 0.08001 0.07358 Time: 0.00087
17-03-27 14:53:07 [1] Train Extra: lr=0.0000631 inv=0.3442188 sub=0.0000000
17-03-27 14:54:38 [1] Step: 54300 Acc: 0.63375 0.85267 Cost: 0.82291 0.68143 0.06789 0.07358 Time: 0.00084
17-03-27 14:54:38 [1] Train Extra: lr=0.0000629 inv=0.4137500 sub=0.0000000
17-03-27 14:56:07 [1] Step: 54400 Acc: 0.65500 0.84998 Cost: 1.15307 0.80050 0.27897 0.07360 Time: 0.00084
17-03-27 14:56:07 [1] Train Extra: lr=0.0000627 inv=0.3970313 sub=0.0000000
17-03-27 14:57:30 [1] Step: 54500 Acc: 0.64125 0.84683 Cost: 1.20767 0.87430 0.25970 0.07367 Time: 0.00080
17-03-27 14:57:30 [1] Train Extra: lr=0.0000625 inv=0.3975000 sub=0.0000000
17-03-27 14:58:35 [1] Step: 54500 Eval acc: 0.65605 0.85657 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 14:58:35 [1] Eval Extra: inv=0.3979779
17-03-27 15:00:07 [1] Step: 54600 Acc: 0.64125 0.84611 Cost: 1.23851 0.96600 0.19880 0.07371 Time: 0.00081
17-03-27 15:00:07 [1] Train Extra: lr=0.0000624 inv=0.4268750 sub=0.0000000
17-03-27 15:01:36 [1] Step: 54700 Acc: 0.67563 0.84636 Cost: 1.10750 0.84198 0.19179 0.07373 Time: 0.00082
17-03-27 15:01:36 [1] Train Extra: lr=0.0000622 inv=0.4189062 sub=0.0000000
17-03-27 15:03:06 [1] Step: 54800 Acc: 0.67219 0.86234 Cost: 1.07260 0.79146 0.20732 0.07382 Time: 0.00088
17-03-27 15:03:06 [1] Train Extra: lr=0.0000620 inv=0.3656250 sub=0.0000000
17-03-27 15:04:31 [1] Step: 54900 Acc: 0.64406 0.84465 Cost: 1.12888 0.82793 0.22712 0.07382 Time: 0.00080
17-03-27 15:04:31 [1] Train Extra: lr=0.0000618 inv=0.4142188 sub=0.0000000
17-03-27 15:06:02 [1] Step: 55000 Acc: 0.64563 0.84959 Cost: 1.00079 0.74459 0.18233 0.07386 Time: 0.00082
17-03-27 15:06:02 [1] Train Extra: lr=0.0000617 inv=0.4235937 sub=0.0000000
17-03-27 15:07:09 [1] Step: 55000 Eval acc: 0.66932 0.85254 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 15:07:09 [1] Eval Extra: inv=0.4078329
17-03-27 15:07:09 [1] Checkpointing with new best dev accuracy of 0.669322
17-03-27 15:07:09 [1] Checkpointing.
17-03-27 15:08:38 [1] Step: 55100 Acc: 0.64406 0.84985 Cost: 0.94041 0.70717 0.15936 0.07388 Time: 0.00084
17-03-27 15:08:38 [1] Train Extra: lr=0.0000615 inv=0.4251563 sub=0.0000000
17-03-27 15:10:09 [1] Step: 55200 Acc: 0.63656 0.84956 Cost: 1.08203 0.89187 0.11622 0.07394 Time: 0.00082
17-03-27 15:10:09 [1] Train Extra: lr=0.0000613 inv=0.4037500 sub=0.0000000
17-03-27 15:11:39 [1] Step: 55300 Acc: 0.65469 0.84436 Cost: 1.35303 1.01064 0.26851 0.07388 Time: 0.00082
17-03-27 15:11:39 [1] Train Extra: lr=0.0000611 inv=0.4110937 sub=0.0000000
17-03-27 15:13:11 [1] Step: 55400 Acc: 0.64812 0.84883 Cost: 1.03108 0.73662 0.22062 0.07384 Time: 0.00083
17-03-27 15:13:11 [1] Train Extra: lr=0.0000609 inv=0.3989063 sub=0.0000000
17-03-27 15:14:29 [1] Step: 55500 Acc: 0.65625 0.85737 Cost: 0.91556 0.72823 0.11344 0.07389 Time: 0.00081
17-03-27 15:14:29 [1] Train Extra: lr=0.0000608 inv=0.3721875 sub=0.0000000
17-03-27 15:15:38 [1] Step: 55500 Eval acc: 0.66350 0.85670 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 15:15:38 [1] Eval Extra: inv=0.4193730
17-03-27 15:17:07 [1] Step: 55600 Acc: 0.65031 0.84539 Cost: 1.39826 1.00412 0.32021 0.07393 Time: 0.00081
17-03-27 15:17:07 [1] Train Extra: lr=0.0000606 inv=0.4287500 sub=0.0000000
17-03-27 15:18:31 [1] Step: 55700 Acc: 0.67625 0.85931 Cost: 1.41709 1.06153 0.28160 0.07396 Time: 0.00085
17-03-27 15:18:31 [1] Train Extra: lr=0.0000604 inv=0.3570313 sub=0.0000000
17-03-27 15:19:56 [1] Step: 55800 Acc: 0.64031 0.84915 Cost: 0.99447 0.82301 0.09738 0.07408 Time: 0.00081
17-03-27 15:19:56 [1] Train Extra: lr=0.0000603 inv=0.4231250 sub=0.0000000
17-03-27 15:21:27 [1] Step: 55900 Acc: 0.65125 0.85743 Cost: 0.83937 0.57176 0.19355 0.07406 Time: 0.00088
17-03-27 15:21:27 [1] Train Extra: lr=0.0000601 inv=0.3862500 sub=0.0000000
17-03-27 15:22:57 [1] Step: 56000 Acc: 0.65719 0.85153 Cost: 0.99664 0.65501 0.26757 0.07406 Time: 0.00084
17-03-27 15:22:57 [1] Train Extra: lr=0.0000599 inv=0.4221875 sub=0.0000000
17-03-27 15:24:03 [1] Step: 56000 Eval acc: 0.66319 0.85628 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 15:24:03 [1] Eval Extra: inv=0.3490094
17-03-27 15:25:33 [1] Step: 56100 Acc: 0.66219 0.85896 Cost: 0.84204 0.66451 0.10339 0.07415 Time: 0.00086
17-03-27 15:25:33 [1] Train Extra: lr=0.0000597 inv=0.4017188 sub=0.0000000
17-03-27 15:26:55 [1] Step: 56200 Acc: 0.66125 0.84929 Cost: 1.20191 0.94095 0.18684 0.07412 Time: 0.00081
17-03-27 15:26:55 [1] Train Extra: lr=0.0000596 inv=0.4292187 sub=0.0000000
17-03-27 15:28:29 [1] Step: 56300 Acc: 0.63406 0.85274 Cost: 1.28492 0.98830 0.22249 0.07413 Time: 0.00084
17-03-27 15:28:29 [1] Train Extra: lr=0.0000594 inv=0.3984375 sub=0.0000000
17-03-27 15:29:59 [1] Step: 56400 Acc: 0.64781 0.84332 Cost: 1.17477 0.85611 0.24452 0.07415 Time: 0.00080
17-03-27 15:29:59 [1] Train Extra: lr=0.0000592 inv=0.4337500 sub=0.0000000
17-03-27 15:31:20 [1] Step: 56500 Acc: 0.65406 0.84878 Cost: 1.32578 0.95923 0.29229 0.07427 Time: 0.00082
17-03-27 15:31:20 [1] Train Extra: lr=0.0000590 inv=0.3662500 sub=0.0000000
17-03-27 15:32:25 [1] Step: 56500 Eval acc: 0.66309 0.85894 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 15:32:25 [1] Eval Extra: inv=0.3470690
17-03-27 15:33:54 [1] Step: 56600 Acc: 0.65938 0.84941 Cost: 1.18166 0.82213 0.28530 0.07423 Time: 0.00083
17-03-27 15:33:54 [1] Train Extra: lr=0.0000589 inv=0.4109375 sub=0.0000000
17-03-27 15:35:26 [1] Step: 56700 Acc: 0.64406 0.84669 Cost: 1.12693 0.77926 0.27342 0.07425 Time: 0.00081
17-03-27 15:35:26 [1] Train Extra: lr=0.0000587 inv=0.4220313 sub=0.0000000
17-03-27 15:36:50 [1] Step: 56800 Acc: 0.64375 0.84346 Cost: 0.95025 0.70956 0.16646 0.07424 Time: 0.00079
17-03-27 15:36:50 [1] Train Extra: lr=0.0000585 inv=0.4096875 sub=0.0000000
17-03-27 15:38:22 [1] Step: 56900 Acc: 0.64406 0.85610 Cost: 1.18655 0.90009 0.21222 0.07424 Time: 0.00084
17-03-27 15:38:22 [1] Train Extra: lr=0.0000584 inv=0.3832813 sub=0.0000000
17-03-27 15:40:05 [1] Step: 57000 Acc: 0.65969 0.86268 Cost: 0.92170 0.74673 0.10066 0.07431 Time: 0.00091
17-03-27 15:40:05 [1] Train Extra: lr=0.0000582 inv=0.3901562 sub=0.0000000
17-03-27 15:41:11 [1] Step: 57000 Eval acc: 0.66371 0.85718 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 15:41:11 [1] Eval Extra: inv=0.3920547
17-03-27 15:42:40 [1] Step: 57100 Acc: 0.67031 0.84400 Cost: 0.90234 0.68075 0.14728 0.07431 Time: 0.00082
17-03-27 15:42:40 [1] Train Extra: lr=0.0000580 inv=0.4073438 sub=0.0000000
17-03-27 15:44:11 [1] Step: 57200 Acc: 0.64625 0.84808 Cost: 0.90916 0.59564 0.23911 0.07441 Time: 0.00084
17-03-27 15:44:11 [1] Train Extra: lr=0.0000579 inv=0.3929687 sub=0.0000000
17-03-27 15:45:43 [1] Step: 57300 Acc: 0.63344 0.84353 Cost: 1.00544 0.73475 0.19625 0.07444 Time: 0.00081
17-03-27 15:45:43 [1] Train Extra: lr=0.0000577 inv=0.4276563 sub=0.0000000
17-03-27 15:47:23 [1] Step: 57400 Acc: 0.64062 0.84989 Cost: 1.03918 0.77267 0.19210 0.07441 Time: 0.00085
17-03-27 15:47:23 [1] Train Extra: lr=0.0000575 inv=0.4317187 sub=0.0000000
17-03-27 15:49:07 [1] Step: 57500 Acc: 0.64625 0.84957 Cost: 1.02864 0.75137 0.20281 0.07445 Time: 0.00089
17-03-27 15:49:07 [1] Train Extra: lr=0.0000574 inv=0.4214062 sub=0.0000000
17-03-27 15:50:12 [1] Step: 57500 Eval acc: 0.66279 0.85636 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 15:50:12 [1] Eval Extra: inv=0.4067606
17-03-27 15:51:42 [1] Step: 57600 Acc: 0.65906 0.85168 Cost: 0.99175 0.69683 0.22043 0.07449 Time: 0.00083
17-03-27 15:51:42 [1] Train Extra: lr=0.0000572 inv=0.4178125 sub=0.0000000
17-03-27 15:53:07 [1] Step: 57700 Acc: 0.65344 0.84262 Cost: 0.91891 0.62837 0.21606 0.07448 Time: 0.00080
17-03-27 15:53:07 [1] Train Extra: lr=0.0000570 inv=0.4043750 sub=0.0000000
17-03-27 15:54:47 [1] Step: 57800 Acc: 0.64594 0.84883 Cost: 1.14636 0.87458 0.19722 0.07456 Time: 0.00083
17-03-27 15:54:47 [1] Train Extra: lr=0.0000569 inv=0.4434375 sub=0.0000000
17-03-27 15:56:15 [1] Step: 57900 Acc: 0.64000 0.85253 Cost: 1.08053 0.80080 0.20514 0.07459 Time: 0.00084
17-03-27 15:56:15 [1] Train Extra: lr=0.0000567 inv=0.3784375 sub=0.0000000
17-03-27 15:57:47 [1] Step: 58000 Acc: 0.65469 0.85190 Cost: 1.15511 0.82125 0.25921 0.07464 Time: 0.00084
17-03-27 15:57:47 [1] Train Extra: lr=0.0000566 inv=0.3764062 sub=0.0000000
17-03-27 15:58:56 [1] Step: 58000 Eval acc: 0.65380 0.86134 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 15:58:56 [1] Eval Extra: inv=0.3854677
17-03-27 16:00:20 [1] Step: 58100 Acc: 0.64094 0.84895 Cost: 0.77174 0.56616 0.13095 0.07463 Time: 0.00082
17-03-27 16:00:20 [1] Train Extra: lr=0.0000564 inv=0.3879687 sub=0.0000000
17-03-27 16:01:39 [1] Step: 58200 Acc: 0.65625 0.85232 Cost: 1.23062 0.94023 0.21575 0.07464 Time: 0.00081
17-03-27 16:01:39 [1] Train Extra: lr=0.0000562 inv=0.3804688 sub=0.0000000
17-03-27 16:03:10 [1] Step: 58300 Acc: 0.64687 0.85365 Cost: 1.23141 0.87992 0.27691 0.07458 Time: 0.00086
17-03-27 16:03:10 [1] Train Extra: lr=0.0000561 inv=0.4018750 sub=0.0000000
17-03-27 16:04:36 [1] Step: 58400 Acc: 0.65250 0.84435 Cost: 1.11035 0.89992 0.13587 0.07456 Time: 0.00080
17-03-27 16:04:36 [1] Train Extra: lr=0.0000559 inv=0.4329688 sub=0.0000000
17-03-27 16:06:08 [1] Step: 58500 Acc: 0.64938 0.84997 Cost: 0.74964 0.59880 0.07624 0.07460 Time: 0.00083
17-03-27 16:06:08 [1] Train Extra: lr=0.0000557 inv=0.4110937 sub=0.0000000
17-03-27 16:07:16 [1] Step: 58500 Eval acc: 0.66912 0.85948 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 16:07:16 [1] Eval Extra: inv=0.4069649
17-03-27 16:08:46 [1] Step: 58600 Acc: 0.62500 0.84628 Cost: 1.04235 0.80923 0.15863 0.07449 Time: 0.00082
17-03-27 16:08:46 [1] Train Extra: lr=0.0000556 inv=0.4159375 sub=0.0000000
17-03-27 16:10:25 [1] Step: 58700 Acc: 0.66969 0.84662 Cost: 1.18153 0.82047 0.28645 0.07461 Time: 0.00086
17-03-27 16:10:25 [1] Train Extra: lr=0.0000554 inv=0.4326563 sub=0.0000000
17-03-27 16:11:55 [1] Step: 58800 Acc: 0.67500 0.84518 Cost: 1.13956 0.79243 0.27246 0.07467 Time: 0.00082
17-03-27 16:11:55 [1] Train Extra: lr=0.0000553 inv=0.4118750 sub=0.0000000
17-03-27 16:13:21 [1] Step: 58900 Acc: 0.66312 0.84397 Cost: 1.02048 0.76079 0.18491 0.07478 Time: 0.00079
17-03-27 16:13:21 [1] Train Extra: lr=0.0000551 inv=0.4240625 sub=0.0000000
17-03-27 16:14:46 [1] Step: 59000 Acc: 0.66687 0.85213 Cost: 0.83557 0.66276 0.09784 0.07497 Time: 0.00081
17-03-27 16:14:46 [1] Train Extra: lr=0.0000550 inv=0.4053125 sub=0.0000000
17-03-27 16:15:55 [1] Step: 59000 Eval acc: 0.66605 0.85292 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 16:15:55 [1] Eval Extra: inv=0.4380617
17-03-27 16:17:27 [1] Step: 59100 Acc: 0.66875 0.84724 Cost: 1.24266 0.92216 0.24537 0.07513 Time: 0.00082
17-03-27 16:17:27 [1] Train Extra: lr=0.0000548 inv=0.4348437 sub=0.0000000
17-03-27 16:18:56 [1] Step: 59200 Acc: 0.68031 0.84869 Cost: 0.71398 0.57258 0.06621 0.07519 Time: 0.00083
17-03-27 16:18:56 [1] Train Extra: lr=0.0000546 inv=0.4121875 sub=0.0000000
17-03-27 16:20:21 [1] Step: 59300 Acc: 0.67969 0.84711 Cost: 1.17626 0.77622 0.32472 0.07533 Time: 0.00081
17-03-27 16:20:21 [1] Train Extra: lr=0.0000545 inv=0.3826562 sub=0.0000000
17-03-27 16:21:44 [1] Step: 59400 Acc: 0.66875 0.85780 Cost: 1.12055 0.76183 0.28321 0.07551 Time: 0.00083
17-03-27 16:21:44 [1] Train Extra: lr=0.0000543 inv=0.3704688 sub=0.0000000
17-03-27 16:23:17 [1] Step: 59500 Acc: 0.64781 0.85747 Cost: 1.28092 0.94700 0.25834 0.07558 Time: 0.00086
17-03-27 16:23:17 [1] Train Extra: lr=0.0000542 inv=0.4128125 sub=0.0000000
17-03-27 16:24:23 [1] Step: 59500 Eval acc: 0.66565 0.85942 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 16:24:23 [1] Eval Extra: inv=0.3635110
17-03-27 16:25:51 [1] Step: 59600 Acc: 0.67656 0.85307 Cost: 1.23035 0.91310 0.24160 0.07565 Time: 0.00085
17-03-27 16:25:51 [1] Train Extra: lr=0.0000540 inv=0.3828125 sub=0.0000000
17-03-27 16:27:16 [1] Step: 59700 Acc: 0.66000 0.85267 Cost: 1.06580 0.72615 0.26393 0.07572 Time: 0.00081
17-03-27 16:27:16 [1] Train Extra: lr=0.0000539 inv=0.3845312 sub=0.0000000
17-03-27 16:28:49 [1] Step: 59800 Acc: 0.66375 0.85560 Cost: 1.01691 0.72574 0.21541 0.07577 Time: 0.00085
17-03-27 16:28:49 [1] Train Extra: lr=0.0000537 inv=0.4040625 sub=0.0000000
17-03-27 16:30:19 [1] Step: 59900 Acc: 0.67500 0.85610 Cost: 1.09913 0.88578 0.13746 0.07589 Time: 0.00084
17-03-27 16:30:19 [1] Train Extra: lr=0.0000535 inv=0.4029687 sub=0.0000000
17-03-27 16:31:52 [1] Step: 60000 Acc: 0.65094 0.84996 Cost: 0.91455 0.64428 0.19433 0.07595 Time: 0.00084
17-03-27 16:31:52 [1] Train Extra: lr=0.0000534 inv=0.4029687 sub=0.0000000
17-03-27 16:33:01 [1] Step: 60000 Eval acc: 0.66483 0.85515 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 16:33:01 [1] Eval Extra: inv=0.3961908
17-03-27 16:33:01 [1] Checkpointing.
17-03-27 16:34:29 [1] Step: 60100 Acc: 0.67375 0.85257 Cost: 1.33873 1.03349 0.22921 0.07604 Time: 0.00084
17-03-27 16:34:29 [1] Train Extra: lr=0.0000532 inv=0.3957812 sub=0.0000000
17-03-27 16:35:49 [1] Step: 60200 Acc: 0.67281 0.84890 Cost: 0.83322 0.57842 0.17872 0.07608 Time: 0.00077
17-03-27 16:35:49 [1] Train Extra: lr=0.0000531 inv=0.3937500 sub=0.0000000
17-03-27 16:37:19 [1] Step: 60300 Acc: 0.65563 0.85660 Cost: 0.84825 0.64887 0.12313 0.07624 Time: 0.00084
17-03-27 16:37:19 [1] Train Extra: lr=0.0000529 inv=0.3939063 sub=0.0000000
17-03-27 16:38:56 [1] Step: 60400 Acc: 0.66094 0.85595 Cost: 0.84034 0.60775 0.15627 0.07632 Time: 0.00088
17-03-27 16:38:56 [1] Train Extra: lr=0.0000528 inv=0.3914063 sub=0.0000000
17-03-27 16:40:22 [1] Step: 60500 Acc: 0.66781 0.84210 Cost: 0.93290 0.70956 0.14700 0.07634 Time: 0.00080
17-03-27 16:40:22 [1] Train Extra: lr=0.0000526 inv=0.3989063 sub=0.0000000
17-03-27 16:41:32 [1] Step: 60500 Eval acc: 0.66493 0.84944 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 16:41:32 [1] Eval Extra: inv=0.3797488
17-03-27 16:43:03 [1] Step: 60600 Acc: 0.67219 0.85587 Cost: 1.15996 0.91922 0.16425 0.07648 Time: 0.00087
17-03-27 16:43:03 [1] Train Extra: lr=0.0000525 inv=0.3945312 sub=0.0000000
17-03-27 16:44:28 [1] Step: 60700 Acc: 0.65750 0.85343 Cost: 0.91883 0.64724 0.19500 0.07659 Time: 0.00084
17-03-27 16:44:28 [1] Train Extra: lr=0.0000523 inv=0.3750000 sub=0.0000000
17-03-27 16:46:09 [1] Step: 60800 Acc: 0.66375 0.85058 Cost: 1.18598 0.80737 0.30190 0.07672 Time: 0.00083
17-03-27 16:46:09 [1] Train Extra: lr=0.0000522 inv=0.4406250 sub=0.0000000
17-03-27 16:47:37 [1] Step: 60900 Acc: 0.66531 0.85592 Cost: 0.99422 0.81386 0.10357 0.07680 Time: 0.00086
17-03-27 16:47:37 [1] Train Extra: lr=0.0000520 inv=0.3576563 sub=0.0000000
17-03-27 16:49:11 [1] Step: 61000 Acc: 0.64875 0.84985 Cost: 1.03405 0.80886 0.14841 0.07678 Time: 0.00083
17-03-27 16:49:11 [1] Train Extra: lr=0.0000519 inv=0.4035937 sub=0.0000000
17-03-27 16:50:22 [1] Step: 61000 Eval acc: 0.66156 0.85821 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 16:50:22 [1] Eval Extra: inv=0.3792892
17-03-27 16:51:46 [1] Step: 61100 Acc: 0.67188 0.85158 Cost: 1.18680 0.87055 0.23943 0.07683 Time: 0.00082
17-03-27 16:51:46 [1] Train Extra: lr=0.0000517 inv=0.3946875 sub=0.0000000
17-03-27 16:53:12 [1] Step: 61200 Acc: 0.66875 0.85214 Cost: 1.00584 0.71593 0.21300 0.07690 Time: 0.00082
17-03-27 16:53:12 [1] Train Extra: lr=0.0000516 inv=0.3768750 sub=0.0000000
17-03-27 16:54:49 [1] Step: 61300 Acc: 0.67312 0.85480 Cost: 1.09125 0.90268 0.11156 0.07701 Time: 0.00088
17-03-27 16:54:49 [1] Train Extra: lr=0.0000514 inv=0.4018750 sub=0.0000000
17-03-27 16:56:12 [1] Step: 61400 Acc: 0.66625 0.84899 Cost: 0.84984 0.62317 0.14961 0.07706 Time: 0.00082
17-03-27 16:56:12 [1] Train Extra: lr=0.0000513 inv=0.3868750 sub=0.0000000
17-03-27 16:57:42 [1] Step: 61500 Acc: 0.65844 0.84652 Cost: 1.26016 0.95669 0.22634 0.07713 Time: 0.00083
17-03-27 16:57:42 [1] Train Extra: lr=0.0000511 inv=0.4050000 sub=0.0000000
17-03-27 16:58:48 [1] Step: 61500 Eval acc: 0.66064 0.85787 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 16:58:48 [1] Eval Extra: inv=0.3844975
17-03-27 17:00:16 [1] Step: 61600 Acc: 0.66250 0.85313 Cost: 1.16090 0.80747 0.27626 0.07717 Time: 0.00085
17-03-27 17:00:16 [1] Train Extra: lr=0.0000510 inv=0.4043750 sub=0.0000000
17-03-27 17:01:35 [1] Step: 61700 Acc: 0.65594 0.85043 Cost: 1.17559 0.83905 0.25932 0.07721 Time: 0.00077
17-03-27 17:01:35 [1] Train Extra: lr=0.0000508 inv=0.4010938 sub=0.0000000
17-03-27 17:03:03 [1] Step: 61800 Acc: 0.66281 0.84961 Cost: 1.32134 1.02517 0.21893 0.07723 Time: 0.00085
17-03-27 17:03:03 [1] Train Extra: lr=0.0000507 inv=0.3964063 sub=0.0000000
17-03-27 17:04:22 [1] Step: 61900 Acc: 0.67437 0.84887 Cost: 1.01155 0.74422 0.19006 0.07727 Time: 0.00077
17-03-27 17:04:22 [1] Train Extra: lr=0.0000506 inv=0.4068750 sub=0.0000000
17-03-27 17:06:05 [1] Step: 62000 Acc: 0.67469 0.84949 Cost: 1.05738 0.81770 0.16235 0.07732 Time: 0.00087
17-03-27 17:06:05 [1] Train Extra: lr=0.0000504 inv=0.4120313 sub=0.0000000
17-03-27 17:07:12 [1] Step: 62000 Eval acc: 0.66136 0.85379 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 17:07:12 [1] Eval Extra: inv=0.4044118
17-03-27 17:08:42 [1] Step: 62100 Acc: 0.65531 0.84774 Cost: 0.90259 0.59300 0.23224 0.07735 Time: 0.00084
17-03-27 17:08:42 [1] Train Extra: lr=0.0000503 inv=0.3817187 sub=0.0000000
17-03-27 17:09:59 [1] Step: 62200 Acc: 0.67188 0.85615 Cost: 0.80844 0.57241 0.15862 0.07741 Time: 0.00081
17-03-27 17:09:59 [1] Train Extra: lr=0.0000501 inv=0.3590625 sub=0.0000000
17-03-27 17:11:28 [1] Step: 62300 Acc: 0.67375 0.84971 Cost: 0.97095 0.69576 0.19766 0.07752 Time: 0.00083
17-03-27 17:11:28 [1] Train Extra: lr=0.0000500 inv=0.4157812 sub=0.0000000
17-03-27 17:12:55 [1] Step: 62400 Acc: 0.64812 0.84893 Cost: 1.30839 0.98879 0.24206 0.07754 Time: 0.00080
17-03-27 17:12:55 [1] Train Extra: lr=0.0000498 inv=0.4001562 sub=0.0000000
17-03-27 17:14:24 [1] Step: 62500 Acc: 0.65156 0.84478 Cost: 1.16629 0.87161 0.21707 0.07761 Time: 0.00083
17-03-27 17:14:24 [1] Train Extra: lr=0.0000497 inv=0.4171875 sub=0.0000000
17-03-27 17:15:34 [1] Step: 62500 Eval acc: 0.66718 0.85268 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 17:15:34 [1] Eval Extra: inv=0.3722937
17-03-27 17:17:08 [1] Step: 62600 Acc: 0.66187 0.84873 Cost: 1.02235 0.73308 0.21164 0.07764 Time: 0.00084
17-03-27 17:17:08 [1] Train Extra: lr=0.0000495 inv=0.4201563 sub=0.0000000
17-03-27 17:18:43 [1] Step: 62700 Acc: 0.65656 0.85547 Cost: 0.86842 0.65173 0.13905 0.07764 Time: 0.00085
17-03-27 17:18:43 [1] Train Extra: lr=0.0000494 inv=0.4151563 sub=0.0000000
17-03-27 17:20:13 [1] Step: 62800 Acc: 0.65219 0.85187 Cost: 0.84854 0.62668 0.14420 0.07767 Time: 0.00085
17-03-27 17:20:13 [1] Train Extra: lr=0.0000493 inv=0.4165625 sub=0.0000000
17-03-27 17:21:50 [1] Step: 62900 Acc: 0.66563 0.85227 Cost: 1.15678 0.87103 0.20803 0.07773 Time: 0.00086
17-03-27 17:21:50 [1] Train Extra: lr=0.0000491 inv=0.4092188 sub=0.0000000
17-03-27 17:23:21 [1] Step: 63000 Acc: 0.65000 0.85809 Cost: 1.09756 0.79689 0.22293 0.07774 Time: 0.00086
17-03-27 17:23:21 [1] Train Extra: lr=0.0000490 inv=0.3757813 sub=0.0000000
17-03-27 17:24:26 [1] Step: 63000 Eval acc: 0.66166 0.85882 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-27 17:24:26 [1] Eval Extra: inv=0.3621834
17-03-27 17:26:01 [1] Step: 63100 Acc: 0.65969 0.84777 Cost: 1.34414 1.01120 0.25517 0.07777 Time: 0.00085
17-03-27 17:26:01 [1] Train Extra: lr=0.0000488 inv=0.4214062 sub=0.0000000
17-03-27 17:27:31 [1] Step: 63200 Acc: 0.66156 0.85921 Cost: 1.10546 0.79346 0.23421 0.07780 Time: 0.00085
17-03-27 17:27:31 [1] Train Extra: lr=0.0000487 inv=0.4015625 sub=0.0000000
17-03-27 17:29:03 [1] Step: 63300 Acc: 0.66563 0.85414 Cost: 1.00797 0.83163 0.09850 0.07784 Time: 0.00084
17-03-27 17:29:03 [1] Train Extra: lr=0.0000486 inv=0.4115625 sub=0.0000000
17-03-27 17:30:35 [1] Step: 63400 Acc: 0.66375 0.84696 Cost: 1.11469 0.83559 0.20125 0.07785 Time: 0.00081
17-03-27 17:30:35 [1] Train Extra: lr=0.0000484 inv=0.4214062 sub=0.0000000
17-03-27 17:32:05 [1] Step: 63500 Acc: 0.64938 0.84634 Cost: 1.41661 1.04958 0.28912 0.07792 Time: 0.00081
17-03-27 17:32:05 [1] Train Extra: lr=0.0000483 inv=0.4259375 sub=0.0000000
17-03-27 17:33:15 [1] Step: 63500 Eval acc: 0.66595 0.85812 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 17:33:15 [1] Eval Extra: inv=0.4186070
17-03-27 17:34:34 [1] Step: 63600 Acc: 0.64875 0.85469 Cost: 1.14031 0.89415 0.16821 0.07795 Time: 0.00078
17-03-27 17:34:34 [1] Train Extra: lr=0.0000481 inv=0.3793750 sub=0.0000000
17-03-27 17:35:57 [1] Step: 63700 Acc: 0.67063 0.84834 Cost: 1.02547 0.79523 0.15224 0.07800 Time: 0.00081
17-03-27 17:35:57 [1] Train Extra: lr=0.0000480 inv=0.3859375 sub=0.0000000
17-03-27 17:37:19 [1] Step: 63800 Acc: 0.65500 0.84655 Cost: 1.02417 0.82405 0.12211 0.07801 Time: 0.00080
17-03-27 17:37:19 [1] Train Extra: lr=0.0000479 inv=0.3931250 sub=0.0000000
17-03-27 17:38:59 [1] Step: 63900 Acc: 0.64219 0.84786 Cost: 1.22979 0.87631 0.27541 0.07807 Time: 0.00083
17-03-27 17:38:59 [1] Train Extra: lr=0.0000477 inv=0.4521875 sub=0.0000000
17-03-27 17:40:30 [1] Step: 64000 Acc: 0.66656 0.85627 Cost: 0.84068 0.65455 0.10804 0.07809 Time: 0.00085
17-03-27 17:40:30 [1] Train Extra: lr=0.0000476 inv=0.3906250 sub=0.0000000
17-03-27 17:41:39 [1] Step: 64000 Eval acc: 0.66738 0.84989 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 17:41:39 [1] Eval Extra: inv=0.4186070
17-03-27 17:43:09 [1] Step: 64100 Acc: 0.65750 0.85333 Cost: 1.04206 0.73869 0.22519 0.07818 Time: 0.00085
17-03-27 17:43:09 [1] Train Extra: lr=0.0000475 inv=0.4007812 sub=0.0000000
17-03-27 17:44:33 [1] Step: 64200 Acc: 0.63813 0.84765 Cost: 0.95288 0.68858 0.18611 0.07819 Time: 0.00080
17-03-27 17:44:33 [1] Train Extra: lr=0.0000473 inv=0.4165625 sub=0.0000000
17-03-27 17:46:06 [1] Step: 64300 Acc: 0.65812 0.85075 Cost: 1.26929 0.89987 0.29123 0.07820 Time: 0.00083
17-03-27 17:46:06 [1] Train Extra: lr=0.0000472 inv=0.4265625 sub=0.0000000
17-03-27 17:47:38 [1] Step: 64400 Acc: 0.65812 0.84876 Cost: 1.16796 0.78547 0.30424 0.07826 Time: 0.00083
17-03-27 17:47:38 [1] Train Extra: lr=0.0000470 inv=0.4110937 sub=0.0000000
17-03-27 17:49:02 [1] Step: 64500 Acc: 0.65438 0.85104 Cost: 1.28841 0.97194 0.23813 0.07834 Time: 0.00082
17-03-27 17:49:02 [1] Train Extra: lr=0.0000469 inv=0.3875000 sub=0.0000000
17-03-27 17:50:10 [1] Step: 64500 Eval acc: 0.66248 0.84988 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 17:50:10 [1] Eval Extra: inv=0.4388276
17-03-27 17:51:42 [1] Step: 64600 Acc: 0.65750 0.85006 Cost: 1.13711 0.71781 0.34094 0.07836 Time: 0.00083
17-03-27 17:51:42 [1] Train Extra: lr=0.0000468 inv=0.4139062 sub=0.0000000
17-03-27 17:53:07 [1] Step: 64700 Acc: 0.64875 0.86258 Cost: 1.12703 0.86128 0.18740 0.07835 Time: 0.00084
17-03-27 17:53:07 [1] Train Extra: lr=0.0000466 inv=0.3926562 sub=0.0000000
17-03-27 17:54:46 [1] Step: 64800 Acc: 0.65031 0.84929 Cost: 1.11612 0.77347 0.26435 0.07831 Time: 0.00084
17-03-27 17:54:46 [1] Train Extra: lr=0.0000465 inv=0.4457813 sub=0.0000000
17-03-27 17:56:16 [1] Step: 64900 Acc: 0.65000 0.84487 Cost: 1.12391 0.88909 0.15653 0.07829 Time: 0.00083
17-03-27 17:56:16 [1] Train Extra: lr=0.0000464 inv=0.4115625 sub=0.0000000
17-03-27 17:57:39 [1] Step: 65000 Acc: 0.67250 0.84667 Cost: 1.05073 0.88508 0.08726 0.07839 Time: 0.00081
17-03-27 17:57:39 [1] Train Extra: lr=0.0000462 inv=0.4176563 sub=0.0000000
17-03-27 17:58:45 [1] Step: 65000 Eval acc: 0.67055 0.85536 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 17:58:45 [1] Eval Extra: inv=0.4565972
17-03-27 17:58:45 [1] Checkpointing.
17-03-27 18:00:13 [1] Step: 65100 Acc: 0.66094 0.84572 Cost: 1.09245 0.82513 0.18891 0.07841 Time: 0.00082
17-03-27 18:00:13 [1] Train Extra: lr=0.0000461 inv=0.4225000 sub=0.0000000
17-03-27 18:01:48 [1] Step: 65200 Acc: 0.65250 0.84800 Cost: 0.95068 0.69186 0.18033 0.07849 Time: 0.00082
17-03-27 18:01:48 [1] Train Extra: lr=0.0000460 inv=0.4418750 sub=0.0000000
17-03-27 18:03:11 [1] Step: 65300 Acc: 0.67063 0.84875 Cost: 0.99121 0.66128 0.25143 0.07850 Time: 0.00081
17-03-27 18:03:11 [1] Train Extra: lr=0.0000458 inv=0.3932813 sub=0.0000000
17-03-27 18:04:39 [1] Step: 65400 Acc: 0.66438 0.85170 Cost: 1.01726 0.72019 0.21858 0.07849 Time: 0.00086
17-03-27 18:04:39 [1] Train Extra: lr=0.0000457 inv=0.3817187 sub=0.0000000
17-03-27 18:06:05 [1] Step: 65500 Acc: 0.65500 0.84856 Cost: 1.08815 0.79628 0.21340 0.07846 Time: 0.00080
17-03-27 18:06:05 [1] Train Extra: lr=0.0000456 inv=0.4304688 sub=0.0000000
17-03-27 18:07:15 [1] Step: 65500 Eval acc: 0.66493 0.85314 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 18:07:15 [1] Eval Extra: inv=0.4049224
17-03-27 18:08:41 [1] Step: 65600 Acc: 0.65094 0.85683 Cost: 1.30349 1.09703 0.12800 0.07846 Time: 0.00084
17-03-27 18:08:41 [1] Train Extra: lr=0.0000454 inv=0.3709375 sub=0.0000000
17-03-27 18:10:06 [1] Step: 65700 Acc: 0.64906 0.84554 Cost: 1.08761 0.85243 0.15668 0.07850 Time: 0.00081
17-03-27 18:10:06 [1] Train Extra: lr=0.0000453 inv=0.4018750 sub=0.0000000
17-03-27 18:11:35 [1] Step: 65800 Acc: 0.65781 0.86534 Cost: 1.10251 0.79601 0.22806 0.07844 Time: 0.00089
17-03-27 18:11:35 [1] Train Extra: lr=0.0000452 inv=0.3612500 sub=0.0000000
17-03-27 18:12:53 [1] Step: 65900 Acc: 0.66750 0.84970 Cost: 1.08837 0.87833 0.13156 0.07848 Time: 0.00079
17-03-27 18:12:53 [1] Train Extra: lr=0.0000451 inv=0.3803125 sub=0.0000000
17-03-27 18:14:18 [1] Step: 66000 Acc: 0.65187 0.85099 Cost: 1.27880 0.95858 0.24166 0.07856 Time: 0.00081
17-03-27 18:14:18 [1] Train Extra: lr=0.0000449 inv=0.4056250 sub=0.0000000
17-03-27 18:15:26 [1] Step: 66000 Eval acc: 0.67014 0.85697 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 18:15:26 [1] Eval Extra: inv=0.4275940
17-03-27 18:16:50 [1] Step: 66100 Acc: 0.65438 0.84758 Cost: 0.98031 0.77131 0.13040 0.07860 Time: 0.00082
17-03-27 18:16:50 [1] Train Extra: lr=0.0000448 inv=0.3839063 sub=0.0000000
17-03-27 18:18:09 [1] Step: 66200 Acc: 0.65281 0.84973 Cost: 1.08485 0.72582 0.28048 0.07855 Time: 0.00079
17-03-27 18:18:09 [1] Train Extra: lr=0.0000447 inv=0.3884375 sub=0.0000000
17-03-27 18:19:39 [1] Step: 66300 Acc: 0.64563 0.85020 Cost: 0.80663 0.55958 0.16845 0.07859 Time: 0.00085
17-03-27 18:19:39 [1] Train Extra: lr=0.0000445 inv=0.4087500 sub=0.0000000
17-03-27 18:21:11 [1] Step: 66400 Acc: 0.65781 0.84607 Cost: 0.86301 0.66006 0.12438 0.07857 Time: 0.00084
17-03-27 18:21:11 [1] Train Extra: lr=0.0000444 inv=0.4143750 sub=0.0000000
17-03-27 18:22:38 [1] Step: 66500 Acc: 0.66156 0.85504 Cost: 0.77564 0.60168 0.09531 0.07864 Time: 0.00082
17-03-27 18:22:38 [1] Train Extra: lr=0.0000443 inv=0.4010938 sub=0.0000000
17-03-27 18:23:47 [1] Step: 66500 Eval acc: 0.67188 0.85933 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 18:23:47 [1] Eval Extra: inv=0.3919526
17-03-27 18:25:06 [1] Step: 66600 Acc: 0.65563 0.84880 Cost: 0.84301 0.66187 0.10251 0.07863 Time: 0.00078
17-03-27 18:25:06 [1] Train Extra: lr=0.0000442 inv=0.3960938 sub=0.0000000
17-03-27 18:26:38 [1] Step: 66700 Acc: 0.65500 0.84865 Cost: 1.17526 0.82336 0.27329 0.07861 Time: 0.00084
17-03-27 18:26:38 [1] Train Extra: lr=0.0000440 inv=0.4078125 sub=0.0000000
17-03-27 18:28:13 [1] Step: 66800 Acc: 0.65844 0.85407 Cost: 1.16122 0.85179 0.23084 0.07860 Time: 0.00086
17-03-27 18:28:13 [1] Train Extra: lr=0.0000439 inv=0.4143750 sub=0.0000000
17-03-27 18:29:36 [1] Step: 66900 Acc: 0.65812 0.84983 Cost: 1.27250 1.01584 0.17803 0.07863 Time: 0.00079
17-03-27 18:29:36 [1] Train Extra: lr=0.0000438 inv=0.4031250 sub=0.0000000
17-03-27 18:31:01 [1] Step: 67000 Acc: 0.65625 0.85014 Cost: 1.03803 0.71813 0.24123 0.07867 Time: 0.00081
17-03-27 18:31:01 [1] Train Extra: lr=0.0000437 inv=0.4248438 sub=0.0000000
17-03-27 18:32:11 [1] Step: 67000 Eval acc: 0.67106 0.85369 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 18:32:11 [1] Eval Extra: inv=0.4223856
17-03-27 18:33:42 [1] Step: 67100 Acc: 0.66750 0.84288 Cost: 0.87003 0.67148 0.11987 0.07868 Time: 0.00082
17-03-27 18:33:42 [1] Train Extra: lr=0.0000435 inv=0.4365625 sub=0.0000000
17-03-27 18:35:10 [1] Step: 67200 Acc: 0.66312 0.84502 Cost: 1.42436 1.17707 0.16854 0.07876 Time: 0.00082
17-03-27 18:35:10 [1] Train Extra: lr=0.0000434 inv=0.4335938 sub=0.0000000
17-03-27 18:36:48 [1] Step: 67300 Acc: 0.65969 0.85309 Cost: 1.02981 0.68603 0.26502 0.07876 Time: 0.00088
17-03-27 18:36:48 [1] Train Extra: lr=0.0000433 inv=0.4081250 sub=0.0000000
17-03-27 18:38:15 [1] Step: 67400 Acc: 0.65750 0.84455 Cost: 1.30060 0.94499 0.27683 0.07878 Time: 0.00081
17-03-27 18:38:15 [1] Train Extra: lr=0.0000432 inv=0.4042188 sub=0.0000000
17-03-27 18:39:37 [1] Step: 67500 Acc: 0.65469 0.84423 Cost: 1.31904 0.98058 0.25970 0.07876 Time: 0.00081
17-03-27 18:39:37 [1] Train Extra: lr=0.0000430 inv=0.4120313 sub=0.0000000
17-03-27 18:40:41 [1] Step: 67500 Eval acc: 0.66830 0.85430 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-27 18:40:41 [1] Eval Extra: inv=0.3971099
17-03-27 18:41:55 [1] Step: 67600 Acc: 0.64563 0.85738 Cost: 1.32538 1.05556 0.19102 0.07880 Time: 0.00080
17-03-27 18:41:55 [1] Train Extra: lr=0.0000429 inv=0.3600000 sub=0.0000000
17-03-27 18:43:28 [1] Step: 67700 Acc: 0.66156 0.85050 Cost: 1.14013 0.83474 0.22649 0.07891 Time: 0.00085
17-03-27 18:43:28 [1] Train Extra: lr=0.0000428 inv=0.3967188 sub=0.0000000
17-03-27 18:45:01 [1] Step: 67800 Acc: 0.66281 0.85553 Cost: 1.07602 0.79527 0.20182 0.07893 Time: 0.00087
17-03-27 18:45:01 [1] Train Extra: lr=0.0000427 inv=0.4114062 sub=0.0000000
17-03-27 18:46:35 [1] Step: 67900 Acc: 0.65656 0.85026 Cost: 1.35801 1.01994 0.25917 0.07890 Time: 0.00086
17-03-27 18:46:35 [1] Train Extra: lr=0.0000425 inv=0.4232812 sub=0.0000000
17-03-27 18:47:55 [1] Step: 68000 Acc: 0.65844 0.85881 Cost: 1.21223 0.80725 0.32607 0.07891 Time: 0.00082
17-03-27 18:47:55 [1] Train Extra: lr=0.0000424 inv=0.3909375 sub=0.0000000
17-03-27 18:49:06 [1] Step: 68000 Eval acc: 0.66626 0.85877 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 18:49:06 [1] Eval Extra: inv=0.4478145
17-03-27 18:50:29 [1] Step: 68100 Acc: 0.66344 0.84746 Cost: 1.09311 0.74350 0.27071 0.07890 Time: 0.00081
17-03-27 18:50:29 [1] Train Extra: lr=0.0000423 inv=0.4167188 sub=0.0000000
17-03-27 18:51:54 [1] Step: 68200 Acc: 0.66719 0.84714 Cost: 0.92948 0.65797 0.19252 0.07899 Time: 0.00080
17-03-27 18:51:54 [1] Train Extra: lr=0.0000422 inv=0.4406250 sub=0.0000000
17-03-27 18:53:27 [1] Step: 68300 Acc: 0.64625 0.85545 Cost: 1.30789 0.95401 0.27489 0.07899 Time: 0.00085
17-03-27 18:53:27 [1] Train Extra: lr=0.0000421 inv=0.4048438 sub=0.0000000
17-03-27 18:54:58 [1] Step: 68400 Acc: 0.63969 0.84942 Cost: 0.91711 0.72551 0.11257 0.07903 Time: 0.00085
17-03-27 18:54:58 [1] Train Extra: lr=0.0000419 inv=0.4179688 sub=0.0000000
17-03-27 18:56:21 [1] Step: 68500 Acc: 0.68219 0.84707 Cost: 1.13249 0.79045 0.26295 0.07909 Time: 0.00080
17-03-27 18:56:21 [1] Train Extra: lr=0.0000418 inv=0.3950000 sub=0.0000000
17-03-27 18:57:27 [1] Step: 68500 Eval acc: 0.66207 0.85550 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 18:57:27 [1] Eval Extra: inv=0.4251430
17-03-27 18:58:56 [1] Step: 68600 Acc: 0.66500 0.84414 Cost: 1.29002 1.00191 0.20891 0.07921 Time: 0.00082
17-03-27 18:58:56 [1] Train Extra: lr=0.0000417 inv=0.4196875 sub=0.0000000
17-03-27 19:00:43 [1] Step: 68700 Acc: 0.66531 0.86039 Cost: 1.01547 0.78087 0.15534 0.07926 Time: 0.00088
17-03-27 19:00:43 [1] Train Extra: lr=0.0000416 inv=0.4353125 sub=0.0000000
17-03-27 19:02:13 [1] Step: 68800 Acc: 0.68125 0.84437 Cost: 0.86943 0.66402 0.12603 0.07938 Time: 0.00081
17-03-27 19:02:13 [1] Train Extra: lr=0.0000415 inv=0.4223438 sub=0.0000000
17-03-27 19:03:30 [1] Step: 68900 Acc: 0.68063 0.85451 Cost: 0.80927 0.59153 0.13828 0.07947 Time: 0.00080
17-03-27 19:03:30 [1] Train Extra: lr=0.0000413 inv=0.3564062 sub=0.0000000
17-03-27 19:04:52 [1] Step: 69000 Acc: 0.67875 0.85011 Cost: 1.14754 0.80222 0.26571 0.07962 Time: 0.00081
17-03-27 19:04:52 [1] Train Extra: lr=0.0000412 inv=0.4157812 sub=0.0000000
17-03-27 19:05:58 [1] Step: 69000 Eval acc: 0.67177 0.85830 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 19:05:58 [1] Eval Extra: inv=0.3815359
17-03-27 19:07:27 [1] Step: 69100 Acc: 0.67250 0.84679 Cost: 0.88874 0.63695 0.17207 0.07971 Time: 0.00082
17-03-27 19:07:27 [1] Train Extra: lr=0.0000411 inv=0.4140625 sub=0.0000000
17-03-27 19:08:48 [1] Step: 69200 Acc: 0.67531 0.84939 Cost: 1.24080 0.99097 0.17005 0.07979 Time: 0.00081
17-03-27 19:08:48 [1] Train Extra: lr=0.0000410 inv=0.3809375 sub=0.0000000
17-03-27 19:10:06 [1] Step: 69300 Acc: 0.68156 0.84795 Cost: 0.77434 0.56213 0.13234 0.07987 Time: 0.00078
17-03-27 19:10:06 [1] Train Extra: lr=0.0000409 inv=0.4154687 sub=0.0000000
17-03-27 19:11:46 [1] Step: 69400 Acc: 0.66750 0.84667 Cost: 0.86628 0.66219 0.12414 0.07994 Time: 0.00084
17-03-27 19:11:46 [1] Train Extra: lr=0.0000407 inv=0.4479688 sub=0.0000000
17-03-27 19:13:22 [1] Step: 69500 Acc: 0.67094 0.84701 Cost: 0.97450 0.68093 0.21355 0.08002 Time: 0.00085
17-03-27 19:13:22 [1] Train Extra: lr=0.0000406 inv=0.4451562 sub=0.0000000
17-03-27 19:14:29 [1] Step: 69500 Eval acc: 0.66708 0.85930 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 19:14:29 [1] Eval Extra: inv=0.3781148
17-03-27 19:16:13 [1] Step: 69600 Acc: 0.68281 0.84490 Cost: 0.96209 0.67688 0.20510 0.08011 Time: 0.00087
17-03-27 19:16:13 [1] Train Extra: lr=0.0000405 inv=0.4514063 sub=0.0000000
17-03-27 19:17:48 [1] Step: 69700 Acc: 0.67219 0.85256 Cost: 0.86684 0.61251 0.17407 0.08026 Time: 0.00084
17-03-27 19:17:48 [1] Train Extra: lr=0.0000404 inv=0.4065625 sub=0.0000000
17-03-27 19:19:24 [1] Step: 69800 Acc: 0.67063 0.84533 Cost: 0.83423 0.65124 0.10267 0.08032 Time: 0.00084
17-03-27 19:19:24 [1] Train Extra: lr=0.0000403 inv=0.4218750 sub=0.0000000
17-03-27 19:20:47 [1] Step: 69900 Acc: 0.67094 0.85810 Cost: 0.83940 0.67857 0.08044 0.08038 Time: 0.00083
17-03-27 19:20:47 [1] Train Extra: lr=0.0000402 inv=0.3771875 sub=0.0000000
17-03-27 19:22:05 [1] Step: 70000 Acc: 0.67500 0.85498 Cost: 0.95065 0.72731 0.14286 0.08048 Time: 0.00080
17-03-27 19:22:05 [1] Train Extra: lr=0.0000400 inv=0.3793750 sub=0.0000000
17-03-27 19:23:14 [1] Step: 70000 Eval acc: 0.66646 0.85639 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 19:23:14 [1] Eval Extra: inv=0.3539624
17-03-27 19:23:14 [1] Checkpointing.
17-03-27 19:24:33 [1] Step: 70100 Acc: 0.68219 0.85676 Cost: 1.13495 0.84009 0.21433 0.08053 Time: 0.00082
17-03-27 19:24:33 [1] Train Extra: lr=0.0000399 inv=0.3637500 sub=0.0000000
17-03-27 19:25:58 [1] Step: 70200 Acc: 0.68531 0.85746 Cost: 0.96234 0.67507 0.20658 0.08068 Time: 0.00086
17-03-27 19:25:58 [1] Train Extra: lr=0.0000398 inv=0.3814063 sub=0.0000000
17-03-27 19:27:26 [1] Step: 70300 Acc: 0.67063 0.84955 Cost: 0.75507 0.56250 0.11186 0.08071 Time: 0.00082
17-03-27 19:27:26 [1] Train Extra: lr=0.0000397 inv=0.4265625 sub=0.0000000
17-03-27 19:29:06 [1] Step: 70400 Acc: 0.66656 0.85305 Cost: 0.71588 0.47641 0.15866 0.08081 Time: 0.00087
17-03-27 19:29:06 [1] Train Extra: lr=0.0000396 inv=0.4260937 sub=0.0000000
17-03-27 19:30:31 [1] Step: 70500 Acc: 0.66969 0.84677 Cost: 1.14834 0.83814 0.22932 0.08088 Time: 0.00080
17-03-27 19:30:31 [1] Train Extra: lr=0.0000395 inv=0.4150000 sub=0.0000000
17-03-27 19:31:41 [1] Step: 70500 Eval acc: 0.66207 0.84908 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 19:31:41 [1] Eval Extra: inv=0.4475592
17-03-27 19:33:06 [1] Step: 70600 Acc: 0.66563 0.84854 Cost: 1.08490 0.79727 0.20664 0.08100 Time: 0.00081
17-03-27 19:33:06 [1] Train Extra: lr=0.0000394 inv=0.3965625 sub=0.0000000
17-03-27 19:34:36 [1] Step: 70700 Acc: 0.66531 0.84974 Cost: 0.97651 0.70268 0.19280 0.08103 Time: 0.00084
17-03-27 19:34:36 [1] Train Extra: lr=0.0000392 inv=0.4112500 sub=0.0000000
17-03-27 19:36:11 [1] Step: 70800 Acc: 0.67125 0.85762 Cost: 0.99184 0.65321 0.25757 0.08105 Time: 0.00084
17-03-27 19:36:11 [1] Train Extra: lr=0.0000391 inv=0.4100000 sub=0.0000000
17-03-27 19:37:36 [1] Step: 70900 Acc: 0.66875 0.85144 Cost: 1.09247 0.86449 0.14686 0.08113 Time: 0.00082
17-03-27 19:37:36 [1] Train Extra: lr=0.0000390 inv=0.4131250 sub=0.0000000
17-03-27 19:39:15 [1] Step: 71000 Acc: 0.66531 0.84361 Cost: 0.88287 0.71148 0.09021 0.08119 Time: 0.00083
17-03-27 19:39:15 [1] Train Extra: lr=0.0000389 inv=0.4332813 sub=0.0000000
17-03-27 19:40:24 [1] Step: 71000 Eval acc: 0.66207 0.85828 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 19:40:24 [1] Eval Extra: inv=0.4187092
17-03-27 19:41:49 [1] Step: 71100 Acc: 0.67719 0.84848 Cost: 1.04604 0.78490 0.17988 0.08126 Time: 0.00082
17-03-27 19:41:49 [1] Train Extra: lr=0.0000388 inv=0.4126563 sub=0.0000000
17-03-27 19:43:22 [1] Step: 71200 Acc: 0.66563 0.85643 Cost: 1.04209 0.76945 0.19135 0.08128 Time: 0.00086
17-03-27 19:43:22 [1] Train Extra: lr=0.0000387 inv=0.4081250 sub=0.0000000
17-03-27 19:44:55 [1] Step: 71300 Acc: 0.65875 0.85325 Cost: 1.04575 0.75274 0.21174 0.08127 Time: 0.00085
17-03-27 19:44:55 [1] Train Extra: lr=0.0000386 inv=0.4287500 sub=0.0000000
17-03-27 19:46:26 [1] Step: 71400 Acc: 0.67063 0.84735 Cost: 1.42339 1.00872 0.33336 0.08131 Time: 0.00083
17-03-27 19:46:26 [1] Train Extra: lr=0.0000385 inv=0.4239062 sub=0.0000000
17-03-27 19:47:52 [1] Step: 71500 Acc: 0.66906 0.85338 Cost: 1.30528 1.04712 0.17675 0.08141 Time: 0.00081
17-03-27 19:47:52 [1] Train Extra: lr=0.0000384 inv=0.4045313 sub=0.0000000
17-03-27 19:49:02 [1] Step: 71500 Eval acc: 0.66197 0.85614 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 19:49:02 [1] Eval Extra: inv=0.3883783
17-03-27 19:50:32 [1] Step: 71600 Acc: 0.66187 0.84964 Cost: 1.06848 0.75202 0.23502 0.08143 Time: 0.00084
17-03-27 19:50:32 [1] Train Extra: lr=0.0000382 inv=0.4245313 sub=0.0000000
17-03-27 19:51:50 [1] Step: 71700 Acc: 0.67219 0.84612 Cost: 0.93461 0.68051 0.17254 0.08156 Time: 0.00077
17-03-27 19:51:50 [1] Train Extra: lr=0.0000381 inv=0.4229688 sub=0.0000000
17-03-27 19:53:19 [1] Step: 71800 Acc: 0.67219 0.86460 Cost: 1.15699 0.75356 0.32185 0.08158 Time: 0.00088
17-03-27 19:53:19 [1] Train Extra: lr=0.0000380 inv=0.3676563 sub=0.0000000
17-03-27 19:54:45 [1] Step: 71900 Acc: 0.67656 0.85359 Cost: 1.07438 0.74631 0.24641 0.08165 Time: 0.00082
17-03-27 19:54:45 [1] Train Extra: lr=0.0000379 inv=0.4340625 sub=0.0000000
17-03-27 19:56:10 [1] Step: 72000 Acc: 0.66750 0.85271 Cost: 1.27837 1.01040 0.18628 0.08169 Time: 0.00083
17-03-27 19:56:10 [1] Train Extra: lr=0.0000378 inv=0.4120313 sub=0.0000000
17-03-27 19:57:16 [1] Step: 72000 Eval acc: 0.67096 0.85611 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 19:57:16 [1] Eval Extra: inv=0.3853145
17-03-27 19:58:38 [1] Step: 72100 Acc: 0.67875 0.85143 Cost: 1.14197 0.85054 0.20971 0.08172 Time: 0.00081
17-03-27 19:58:38 [1] Train Extra: lr=0.0000377 inv=0.3885938 sub=0.0000000
17-03-27 20:00:08 [1] Step: 72200 Acc: 0.67156 0.84621 Cost: 1.17144 0.87301 0.21666 0.08178 Time: 0.00084
17-03-27 20:00:08 [1] Train Extra: lr=0.0000376 inv=0.4035937 sub=0.0000000
17-03-27 20:01:34 [1] Step: 72300 Acc: 0.66781 0.85793 Cost: 1.05122 0.74586 0.22349 0.08186 Time: 0.00085
17-03-27 20:01:34 [1] Train Extra: lr=0.0000375 inv=0.3782813 sub=0.0000000
17-03-27 20:03:07 [1] Step: 72400 Acc: 0.66344 0.85496 Cost: 0.91244 0.66207 0.16854 0.08183 Time: 0.00086
17-03-27 20:03:07 [1] Train Extra: lr=0.0000374 inv=0.4204688 sub=0.0000000
17-03-27 20:04:33 [1] Step: 72500 Acc: 0.67531 0.84511 Cost: 1.28411 0.98073 0.22149 0.08190 Time: 0.00081
17-03-27 20:04:33 [1] Train Extra: lr=0.0000373 inv=0.4225000 sub=0.0000000
17-03-27 20:05:43 [1] Step: 72500 Eval acc: 0.66299 0.85441 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 20:05:43 [1] Eval Extra: inv=0.4186581
17-03-27 20:07:09 [1] Step: 72600 Acc: 0.66750 0.85079 Cost: 1.04042 0.72372 0.23462 0.08209 Time: 0.00081
17-03-27 20:07:09 [1] Train Extra: lr=0.0000372 inv=0.3842187 sub=0.0000000
17-03-27 20:08:33 [1] Step: 72700 Acc: 0.66438 0.84562 Cost: 1.05509 0.76297 0.20998 0.08215 Time: 0.00079
17-03-27 20:08:33 [1] Train Extra: lr=0.0000371 inv=0.4057812 sub=0.0000000
17-03-27 20:09:58 [1] Step: 72800 Acc: 0.68500 0.85319 Cost: 0.94706 0.72649 0.13836 0.08222 Time: 0.00084
17-03-27 20:09:58 [1] Train Extra: lr=0.0000369 inv=0.3773437 sub=0.0000000
17-03-27 20:11:38 [1] Step: 72900 Acc: 0.67563 0.84516 Cost: 1.50281 1.20886 0.21169 0.08226 Time: 0.00085
17-03-27 20:11:38 [1] Train Extra: lr=0.0000368 inv=0.4432813 sub=0.0000000
17-03-27 20:13:04 [1] Step: 73000 Acc: 0.67656 0.85477 Cost: 1.04207 0.74692 0.21284 0.08231 Time: 0.00084
17-03-27 20:13:04 [1] Train Extra: lr=0.0000367 inv=0.3939063 sub=0.0000000
17-03-27 20:14:14 [1] Step: 73000 Eval acc: 0.67188 0.85487 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 20:14:14 [1] Eval Extra: inv=0.4333640
17-03-27 20:15:40 [1] Step: 73100 Acc: 0.67281 0.84295 Cost: 1.04410 0.76621 0.19554 0.08235 Time: 0.00080
17-03-27 20:15:40 [1] Train Extra: lr=0.0000366 inv=0.4096875 sub=0.0000000
17-03-27 20:17:13 [1] Step: 73200 Acc: 0.69063 0.85084 Cost: 1.18615 0.84790 0.25589 0.08236 Time: 0.00086
17-03-27 20:17:13 [1] Train Extra: lr=0.0000365 inv=0.4168750 sub=0.0000000
17-03-27 20:18:31 [1] Step: 73300 Acc: 0.66344 0.85116 Cost: 1.30095 0.97284 0.24570 0.08242 Time: 0.00081
17-03-27 20:18:31 [1] Train Extra: lr=0.0000364 inv=0.3818750 sub=0.0000000
17-03-27 20:20:03 [1] Step: 73400 Acc: 0.66625 0.85010 Cost: 1.22980 0.99115 0.15619 0.08246 Time: 0.00083
17-03-27 20:20:03 [1] Train Extra: lr=0.0000363 inv=0.4200000 sub=0.0000000
17-03-27 20:21:32 [1] Step: 73500 Acc: 0.66031 0.85775 Cost: 1.02648 0.78753 0.15648 0.08247 Time: 0.00088
17-03-27 20:21:32 [1] Train Extra: lr=0.0000362 inv=0.3896875 sub=0.0000000
17-03-27 20:22:38 [1] Step: 73500 Eval acc: 0.67034 0.85739 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 20:22:38 [1] Eval Extra: inv=0.4294833
17-03-27 20:24:00 [1] Step: 73600 Acc: 0.67906 0.84941 Cost: 0.87927 0.71180 0.08496 0.08251 Time: 0.00083
17-03-27 20:24:00 [1] Train Extra: lr=0.0000361 inv=0.3939063 sub=0.0000000
17-03-27 20:25:38 [1] Step: 73700 Acc: 0.65469 0.85329 Cost: 1.23353 0.92171 0.22927 0.08255 Time: 0.00087
17-03-27 20:25:38 [1] Train Extra: lr=0.0000360 inv=0.4446875 sub=0.0000000
17-03-27 20:27:07 [1] Step: 73800 Acc: 0.67469 0.85285 Cost: 0.91235 0.65462 0.17517 0.08256 Time: 0.00082
17-03-27 20:27:07 [1] Train Extra: lr=0.0000359 inv=0.4212500 sub=0.0000000
17-03-27 20:28:39 [1] Step: 73900 Acc: 0.66906 0.85779 Cost: 1.02257 0.59971 0.34024 0.08262 Time: 0.00086
17-03-27 20:28:39 [1] Train Extra: lr=0.0000358 inv=0.4064063 sub=0.0000000
17-03-27 20:30:11 [1] Step: 74000 Acc: 0.67000 0.85060 Cost: 1.59345 1.20227 0.30850 0.08268 Time: 0.00084
17-03-27 20:30:11 [1] Train Extra: lr=0.0000357 inv=0.4075000 sub=0.0000000
17-03-27 20:31:21 [1] Step: 74000 Eval acc: 0.66646 0.85823 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 20:31:21 [1] Eval Extra: inv=0.4242749
17-03-27 20:32:52 [1] Step: 74100 Acc: 0.66187 0.84147 Cost: 1.45742 1.05783 0.31692 0.08267 Time: 0.00083
17-03-27 20:32:52 [1] Train Extra: lr=0.0000356 inv=0.4453125 sub=0.0000000
17-03-27 20:34:19 [1] Step: 74200 Acc: 0.66250 0.84751 Cost: 1.17011 0.84239 0.24503 0.08270 Time: 0.00080
17-03-27 20:34:19 [1] Train Extra: lr=0.0000355 inv=0.4171875 sub=0.0000000
17-03-27 20:35:44 [1] Step: 74300 Acc: 0.67375 0.85169 Cost: 1.21164 0.80743 0.32148 0.08273 Time: 0.00082
17-03-27 20:35:44 [1] Train Extra: lr=0.0000354 inv=0.4273438 sub=0.0000000
17-03-27 20:37:05 [1] Step: 74400 Acc: 0.67656 0.85178 Cost: 1.03745 0.80169 0.15301 0.08276 Time: 0.00083
17-03-27 20:37:05 [1] Train Extra: lr=0.0000353 inv=0.4020313 sub=0.0000000
17-03-27 20:38:30 [1] Step: 74500 Acc: 0.65844 0.85691 Cost: 0.93009 0.58855 0.25875 0.08279 Time: 0.00081
17-03-27 20:38:30 [1] Train Extra: lr=0.0000352 inv=0.3950000 sub=0.0000000
17-03-27 20:39:35 [1] Step: 74500 Eval acc: 0.66473 0.85871 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 20:39:35 [1] Eval Extra: inv=0.3984886
17-03-27 20:41:00 [1] Step: 74600 Acc: 0.67406 0.84869 Cost: 1.26597 1.03974 0.14333 0.08290 Time: 0.00082
17-03-27 20:41:00 [1] Train Extra: lr=0.0000351 inv=0.4010938 sub=0.0000000
17-03-27 20:42:35 [1] Step: 74700 Acc: 0.66219 0.84822 Cost: 1.00457 0.76285 0.15883 0.08290 Time: 0.00085
17-03-27 20:42:35 [1] Train Extra: lr=0.0000350 inv=0.4237500 sub=0.0000000
17-03-27 20:44:00 [1] Step: 74800 Acc: 0.66812 0.85006 Cost: 1.07105 0.78128 0.20686 0.08291 Time: 0.00081
17-03-27 20:44:00 [1] Train Extra: lr=0.0000349 inv=0.4123438 sub=0.0000000
17-03-27 20:45:23 [1] Step: 74900 Acc: 0.67219 0.85624 Cost: 1.31313 0.97248 0.25776 0.08290 Time: 0.00084
17-03-27 20:45:23 [1] Train Extra: lr=0.0000348 inv=0.4032812 sub=0.0000000
17-03-27 20:46:50 [1] Step: 75000 Acc: 0.67031 0.85569 Cost: 1.01140 0.73811 0.19038 0.08291 Time: 0.00081
17-03-27 20:46:50 [1] Train Extra: lr=0.0000347 inv=0.4209375 sub=0.0000000
17-03-27 20:48:00 [1] Step: 75000 Eval acc: 0.66748 0.85390 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 20:48:00 [1] Eval Extra: inv=0.4144199
17-03-27 20:48:00 [1] Checkpointing.
17-03-27 20:49:26 [1] Step: 75100 Acc: 0.66719 0.85411 Cost: 1.03705 0.73446 0.21963 0.08296 Time: 0.00082
17-03-27 20:49:26 [1] Train Extra: lr=0.0000346 inv=0.4134375 sub=0.0000000
17-03-27 20:50:59 [1] Step: 75200 Acc: 0.66687 0.86010 Cost: 1.11862 0.85674 0.17890 0.08297 Time: 0.00088
17-03-27 20:50:59 [1] Train Extra: lr=0.0000345 inv=0.4235937 sub=0.0000000
17-03-27 20:52:37 [1] Step: 75300 Acc: 0.66969 0.85167 Cost: 1.20645 0.95570 0.16779 0.08295 Time: 0.00090
17-03-27 20:52:37 [1] Train Extra: lr=0.0000344 inv=0.4195313 sub=0.0000000
17-03-27 20:54:02 [1] Step: 75400 Acc: 0.67000 0.85686 Cost: 1.27419 0.91083 0.28030 0.08306 Time: 0.00084
17-03-27 20:54:02 [1] Train Extra: lr=0.0000343 inv=0.3823437 sub=0.0000000
17-03-27 20:55:29 [1] Step: 75500 Acc: 0.65438 0.85459 Cost: 1.01053 0.67970 0.24777 0.08306 Time: 0.00082
17-03-27 20:55:29 [1] Train Extra: lr=0.0000342 inv=0.4251563 sub=0.0000000
17-03-27 20:56:38 [1] Step: 75500 Eval acc: 0.66789 0.85692 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 20:56:38 [1] Eval Extra: inv=0.4146242
17-03-27 20:58:03 [1] Step: 75600 Acc: 0.66594 0.85077 Cost: 0.90611 0.63524 0.18774 0.08312 Time: 0.00081
17-03-27 20:58:03 [1] Train Extra: lr=0.0000341 inv=0.4320312 sub=0.0000000
17-03-27 20:59:27 [1] Step: 75700 Acc: 0.66719 0.84841 Cost: 1.16031 0.84935 0.22787 0.08309 Time: 0.00082
17-03-27 20:59:27 [1] Train Extra: lr=0.0000340 inv=0.4207812 sub=0.0000000
17-03-27 21:00:51 [1] Step: 75800 Acc: 0.65844 0.84786 Cost: 1.15767 0.87113 0.20347 0.08306 Time: 0.00081
17-03-27 21:00:51 [1] Train Extra: lr=0.0000339 inv=0.3957812 sub=0.0000000
17-03-27 21:02:15 [1] Step: 75900 Acc: 0.66500 0.85276 Cost: 1.12500 0.85447 0.18751 0.08302 Time: 0.00082
17-03-27 21:02:15 [1] Train Extra: lr=0.0000338 inv=0.3948437 sub=0.0000000
17-03-27 21:03:40 [1] Step: 76000 Acc: 0.65781 0.85452 Cost: 0.71540 0.52222 0.11006 0.08312 Time: 0.00082
17-03-27 21:03:40 [1] Train Extra: lr=0.0000337 inv=0.3920312 sub=0.0000000
17-03-27 21:04:49 [1] Step: 76000 Eval acc: 0.66810 0.85430 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 21:04:49 [1] Eval Extra: inv=0.4116626
17-03-27 21:06:30 [1] Step: 76100 Acc: 0.65531 0.85331 Cost: 0.84140 0.61418 0.14408 0.08314 Time: 0.00085
17-03-27 21:06:30 [1] Train Extra: lr=0.0000336 inv=0.4495312 sub=0.0000000
17-03-27 21:08:03 [1] Step: 76200 Acc: 0.65750 0.83926 Cost: 0.87209 0.66980 0.11912 0.08316 Time: 0.00081
17-03-27 21:08:03 [1] Train Extra: lr=0.0000335 inv=0.4348437 sub=0.0000000
17-03-27 21:09:42 [1] Step: 76300 Acc: 0.66094 0.85140 Cost: 1.19567 0.87705 0.23542 0.08320 Time: 0.00084
17-03-27 21:09:42 [1] Train Extra: lr=0.0000334 inv=0.4453125 sub=0.0000000
17-03-27 21:11:12 [1] Step: 76400 Acc: 0.67125 0.85244 Cost: 1.24519 0.85789 0.30410 0.08319 Time: 0.00083
17-03-27 21:11:12 [1] Train Extra: lr=0.0000333 inv=0.4268750 sub=0.0000000
17-03-27 21:12:44 [1] Step: 76500 Acc: 0.65281 0.85132 Cost: 0.86929 0.59452 0.19157 0.08320 Time: 0.00082
17-03-27 21:12:44 [1] Train Extra: lr=0.0000332 inv=0.4512500 sub=0.0000000
17-03-27 21:13:53 [1] Step: 76500 Eval acc: 0.67085 0.86359 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 21:13:53 [1] Eval Extra: inv=0.3834252
17-03-27 21:15:16 [1] Step: 76600 Acc: 0.67875 0.85004 Cost: 1.23601 0.86972 0.28302 0.08327 Time: 0.00080
17-03-27 21:15:16 [1] Train Extra: lr=0.0000331 inv=0.3970313 sub=0.0000000
17-03-27 21:16:47 [1] Step: 76700 Acc: 0.66000 0.85011 Cost: 1.09306 0.82848 0.18136 0.08322 Time: 0.00081
17-03-27 21:16:47 [1] Train Extra: lr=0.0000330 inv=0.4334375 sub=0.0000000
17-03-27 21:18:23 [1] Step: 76800 Acc: 0.67437 0.85785 Cost: 1.04208 0.75604 0.20276 0.08328 Time: 0.00088
17-03-27 21:18:23 [1] Train Extra: lr=0.0000329 inv=0.3918750 sub=0.0000000
17-03-27 21:20:03 [1] Step: 76900 Acc: 0.65375 0.86377 Cost: 1.00264 0.64810 0.27123 0.08331 Time: 0.00090
17-03-27 21:20:03 [1] Train Extra: lr=0.0000328 inv=0.3960938 sub=0.0000000
17-03-27 21:21:28 [1] Step: 77000 Acc: 0.67219 0.84975 Cost: 0.95724 0.64757 0.22637 0.08331 Time: 0.00082
17-03-27 21:21:28 [1] Train Extra: lr=0.0000327 inv=0.4118750 sub=0.0000000
17-03-27 21:22:33 [1] Step: 77000 Eval acc: 0.67525 0.85334 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 21:22:33 [1] Eval Extra: inv=0.4243770
17-03-27 21:22:33 [1] Checkpointing with new best dev accuracy of 0.675245
17-03-27 21:24:03 [1] Step: 77100 Acc: 0.65969 0.85346 Cost: 1.28622 0.92327 0.27964 0.08331 Time: 0.00086
17-03-27 21:24:03 [1] Train Extra: lr=0.0000326 inv=0.3996875 sub=0.0000000
17-03-27 21:25:31 [1] Step: 77200 Acc: 0.65469 0.85401 Cost: 1.64540 1.17915 0.38291 0.08334 Time: 0.00086
17-03-27 21:25:31 [1] Train Extra: lr=0.0000326 inv=0.3954687 sub=0.0000000
17-03-27 21:26:59 [1] Step: 77300 Acc: 0.67937 0.86067 Cost: 0.97850 0.68670 0.20840 0.08339 Time: 0.00086
17-03-27 21:26:59 [1] Train Extra: lr=0.0000325 inv=0.3948437 sub=0.0000000
17-03-27 21:28:39 [1] Step: 77400 Acc: 0.68156 0.85002 Cost: 1.01756 0.80501 0.12919 0.08336 Time: 0.00085
17-03-27 21:28:39 [1] Train Extra: lr=0.0000324 inv=0.4406250 sub=0.0000000
17-03-27 21:30:04 [1] Step: 77500 Acc: 0.66094 0.85171 Cost: 1.22925 0.90715 0.23870 0.08339 Time: 0.00083
17-03-27 21:30:04 [1] Train Extra: lr=0.0000323 inv=0.4068750 sub=0.0000000
17-03-27 21:31:15 [1] Step: 77500 Eval acc: 0.66891 0.85894 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 21:31:15 [1] Eval Extra: inv=0.4055862
17-03-27 21:32:54 [1] Step: 77600 Acc: 0.66719 0.85776 Cost: 1.17786 0.84442 0.24999 0.08346 Time: 0.00086
17-03-27 21:32:54 [1] Train Extra: lr=0.0000322 inv=0.4182812 sub=0.0000000
17-03-27 21:34:23 [1] Step: 77700 Acc: 0.65875 0.85276 Cost: 1.09805 0.71751 0.29708 0.08347 Time: 0.00083
17-03-27 21:34:23 [1] Train Extra: lr=0.0000321 inv=0.4151563 sub=0.0000000
17-03-27 21:35:47 [1] Step: 77800 Acc: 0.65469 0.84812 Cost: 0.93393 0.69069 0.15974 0.08350 Time: 0.00081
17-03-27 21:35:47 [1] Train Extra: lr=0.0000320 inv=0.4135937 sub=0.0000000
17-03-27 21:37:14 [1] Step: 77900 Acc: 0.65375 0.85553 Cost: 0.70197 0.51248 0.10598 0.08350 Time: 0.00084
17-03-27 21:37:14 [1] Train Extra: lr=0.0000319 inv=0.3979687 sub=0.0000000
17-03-27 21:38:48 [1] Step: 78000 Acc: 0.66281 0.85452 Cost: 0.95702 0.70486 0.16865 0.08351 Time: 0.00085
17-03-27 21:38:48 [1] Train Extra: lr=0.0000318 inv=0.4320312 sub=0.0000000
17-03-27 21:39:58 [1] Step: 78000 Eval acc: 0.66983 0.85543 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 21:39:58 [1] Eval Extra: inv=0.4056883
17-03-27 21:41:31 [1] Step: 78100 Acc: 0.67156 0.85144 Cost: 0.87342 0.78156 0.00838 0.08349 Time: 0.00085
17-03-27 21:41:31 [1] Train Extra: lr=0.0000317 inv=0.4232812 sub=0.0000000
17-03-27 21:43:00 [1] Step: 78200 Acc: 0.65969 0.86034 Cost: 1.20530 0.78829 0.33344 0.08357 Time: 0.00087
17-03-27 21:43:00 [1] Train Extra: lr=0.0000316 inv=0.3843750 sub=0.0000000
17-03-27 21:44:33 [1] Step: 78300 Acc: 0.68531 0.85632 Cost: 1.00587 0.76249 0.15968 0.08370 Time: 0.00085
17-03-27 21:44:33 [1] Train Extra: lr=0.0000315 inv=0.4353125 sub=0.0000000
17-03-27 21:45:57 [1] Step: 78400 Acc: 0.69281 0.85001 Cost: 1.08664 0.80997 0.19287 0.08380 Time: 0.00081
17-03-27 21:45:57 [1] Train Extra: lr=0.0000314 inv=0.4126563 sub=0.0000000
17-03-27 21:47:31 [1] Step: 78500 Acc: 0.67219 0.84724 Cost: 1.05558 0.75002 0.22167 0.08389 Time: 0.00082
17-03-27 21:47:31 [1] Train Extra: lr=0.0000314 inv=0.4375000 sub=0.0000000
17-03-27 21:48:40 [1] Step: 78500 Eval acc: 0.67371 0.86040 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 21:48:40 [1] Eval Extra: inv=0.3946589
17-03-27 21:50:19 [1] Step: 78600 Acc: 0.67781 0.85165 Cost: 1.01168 0.79132 0.13633 0.08403 Time: 0.00086
17-03-27 21:50:19 [1] Train Extra: lr=0.0000313 inv=0.4314062 sub=0.0000000
17-03-27 21:51:44 [1] Step: 78700 Acc: 0.69281 0.85709 Cost: 0.80224 0.48303 0.23506 0.08414 Time: 0.00083
17-03-27 21:51:44 [1] Train Extra: lr=0.0000312 inv=0.3965625 sub=0.0000000
17-03-27 21:53:23 [1] Step: 78800 Acc: 0.67937 0.85350 Cost: 0.87367 0.53928 0.25017 0.08422 Time: 0.00088
17-03-27 21:53:23 [1] Train Extra: lr=0.0000311 inv=0.4356250 sub=0.0000000
17-03-27 21:54:57 [1] Step: 78900 Acc: 0.67500 0.85079 Cost: 0.83759 0.68596 0.06733 0.08429 Time: 0.00082
17-03-27 21:54:57 [1] Train Extra: lr=0.0000310 inv=0.4390625 sub=0.0000000
17-03-27 21:56:23 [1] Step: 79000 Acc: 0.69156 0.85032 Cost: 0.99295 0.75726 0.15139 0.08429 Time: 0.00082
17-03-27 21:56:23 [1] Train Extra: lr=0.0000309 inv=0.4173438 sub=0.0000000
17-03-27 21:57:36 [1] Step: 79000 Eval acc: 0.66902 0.85701 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 21:57:36 [1] Eval Extra: inv=0.3742341
17-03-27 21:59:08 [1] Step: 79100 Acc: 0.68906 0.84926 Cost: 1.05683 0.81700 0.15546 0.08437 Time: 0.00085
17-03-27 21:59:08 [1] Train Extra: lr=0.0000308 inv=0.3928125 sub=0.0000000
17-03-27 22:00:35 [1] Step: 79200 Acc: 0.69312 0.85221 Cost: 0.72153 0.53526 0.10178 0.08449 Time: 0.00088
17-03-27 22:00:35 [1] Train Extra: lr=0.0000307 inv=0.3707813 sub=0.0000000
17-03-27 22:01:59 [1] Step: 79300 Acc: 0.68625 0.84612 Cost: 1.10276 0.75033 0.26778 0.08465 Time: 0.00081
17-03-27 22:01:59 [1] Train Extra: lr=0.0000306 inv=0.4093750 sub=0.0000000
17-03-27 22:03:20 [1] Step: 79400 Acc: 0.67625 0.85455 Cost: 1.14133 0.84376 0.21280 0.08477 Time: 0.00081
17-03-27 22:03:20 [1] Train Extra: lr=0.0000306 inv=0.3734375 sub=0.0000000
17-03-27 22:04:45 [1] Step: 79500 Acc: 0.67437 0.85083 Cost: 1.07599 0.75471 0.23645 0.08483 Time: 0.00083
17-03-27 22:04:45 [1] Train Extra: lr=0.0000305 inv=0.3959375 sub=0.0000000
17-03-27 22:05:54 [1] Step: 79500 Eval acc: 0.67239 0.85762 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 22:05:54 [1] Eval Extra: inv=0.3998672
17-03-27 22:07:14 [1] Step: 79600 Acc: 0.68344 0.86150 Cost: 1.00148 0.69597 0.22069 0.08482 Time: 0.00085
17-03-27 22:07:14 [1] Train Extra: lr=0.0000304 inv=0.3746875 sub=0.0000000
17-03-27 22:08:45 [1] Step: 79700 Acc: 0.66719 0.84483 Cost: 1.22681 0.90033 0.24157 0.08491 Time: 0.00081
17-03-27 22:08:45 [1] Train Extra: lr=0.0000303 inv=0.4360938 sub=0.0000000
17-03-27 22:10:12 [1] Step: 79800 Acc: 0.68437 0.85458 Cost: 0.97379 0.67805 0.21076 0.08497 Time: 0.00083
17-03-27 22:10:12 [1] Train Extra: lr=0.0000302 inv=0.3895312 sub=0.0000000
17-03-27 22:11:51 [1] Step: 79900 Acc: 0.66875 0.84380 Cost: 1.13910 0.78416 0.26989 0.08505 Time: 0.00085
17-03-27 22:11:51 [1] Train Extra: lr=0.0000301 inv=0.4381250 sub=0.0000000
17-03-27 22:13:17 [1] Step: 80000 Acc: 0.68250 0.84876 Cost: 1.00364 0.76908 0.14946 0.08510 Time: 0.00081
17-03-27 22:13:17 [1] Train Extra: lr=0.0000300 inv=0.3957812 sub=0.0000000
17-03-27 22:14:27 [1] Step: 80000 Eval acc: 0.66258 0.85503 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 22:14:27 [1] Eval Extra: inv=0.3720384
17-03-27 22:14:27 [1] Checkpointing.
17-03-27 22:15:59 [1] Step: 80100 Acc: 0.67031 0.85609 Cost: 1.07749 0.78478 0.20753 0.08518 Time: 0.00085
17-03-27 22:15:59 [1] Train Extra: lr=0.0000299 inv=0.4246875 sub=0.0000000
17-03-27 22:17:36 [1] Step: 80200 Acc: 0.68156 0.85038 Cost: 0.91787 0.66062 0.17201 0.08524 Time: 0.00085
17-03-27 22:17:36 [1] Train Extra: lr=0.0000299 inv=0.4378125 sub=0.0000000
17-03-27 22:19:09 [1] Step: 80300 Acc: 0.67625 0.85305 Cost: 0.74487 0.51705 0.14247 0.08535 Time: 0.00086
17-03-27 22:19:09 [1] Train Extra: lr=0.0000298 inv=0.4354688 sub=0.0000000
17-03-27 22:20:48 [1] Step: 80400 Acc: 0.70469 0.85689 Cost: 0.92693 0.67876 0.16275 0.08541 Time: 0.00088
17-03-27 22:20:48 [1] Train Extra: lr=0.0000297 inv=0.4018750 sub=0.0000000
17-03-27 22:22:14 [1] Step: 80500 Acc: 0.68375 0.85589 Cost: 0.85336 0.66256 0.10534 0.08546 Time: 0.00083
17-03-27 22:22:14 [1] Train Extra: lr=0.0000296 inv=0.3764062 sub=0.0000000
17-03-27 22:23:25 [1] Step: 80500 Eval acc: 0.67106 0.85988 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 22:23:25 [1] Eval Extra: inv=0.3988971
17-03-27 22:24:48 [1] Step: 80600 Acc: 0.68563 0.85028 Cost: 0.80025 0.49246 0.22224 0.08555 Time: 0.00083
17-03-27 22:24:48 [1] Train Extra: lr=0.0000295 inv=0.3825000 sub=0.0000000
17-03-27 22:26:13 [1] Step: 80700 Acc: 0.67594 0.85080 Cost: 0.78595 0.56104 0.13929 0.08562 Time: 0.00083
17-03-27 22:26:13 [1] Train Extra: lr=0.0000294 inv=0.4259375 sub=0.0000000
17-03-27 22:27:49 [1] Step: 80800 Acc: 0.66563 0.84697 Cost: 1.02096 0.72930 0.20603 0.08563 Time: 0.00081
17-03-27 22:27:49 [1] Train Extra: lr=0.0000294 inv=0.4564063 sub=0.0000000
17-03-27 22:29:13 [1] Step: 80900 Acc: 0.67656 0.85027 Cost: 0.95683 0.67269 0.19845 0.08569 Time: 0.00081
17-03-27 22:29:13 [1] Train Extra: lr=0.0000293 inv=0.3960938 sub=0.0000000
17-03-27 22:30:43 [1] Step: 81000 Acc: 0.68375 0.85195 Cost: 1.04794 0.73736 0.22483 0.08576 Time: 0.00085
17-03-27 22:30:43 [1] Train Extra: lr=0.0000292 inv=0.4045313 sub=0.0000000
17-03-27 22:31:53 [1] Step: 81000 Eval acc: 0.66942 0.85520 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 22:31:53 [1] Eval Extra: inv=0.4242749
17-03-27 22:33:34 [1] Step: 81100 Acc: 0.68312 0.84853 Cost: 0.97573 0.60892 0.28102 0.08579 Time: 0.00085
17-03-27 22:33:34 [1] Train Extra: lr=0.0000291 inv=0.4415625 sub=0.0000000
17-03-27 22:35:05 [1] Step: 81200 Acc: 0.66906 0.84811 Cost: 1.32963 0.93367 0.31013 0.08584 Time: 0.00083
17-03-27 22:35:05 [1] Train Extra: lr=0.0000290 inv=0.4203125 sub=0.0000000
17-03-27 22:36:29 [1] Step: 81300 Acc: 0.67000 0.85850 Cost: 1.13357 0.79507 0.25258 0.08593 Time: 0.00084
17-03-27 22:36:29 [1] Train Extra: lr=0.0000289 inv=0.3909375 sub=0.0000000
17-03-27 22:38:03 [1] Step: 81400 Acc: 0.66000 0.85434 Cost: 1.04600 0.75947 0.20054 0.08600 Time: 0.00086
17-03-27 22:38:03 [1] Train Extra: lr=0.0000288 inv=0.4142188 sub=0.0000000
17-03-27 22:39:35 [1] Step: 81500 Acc: 0.67094 0.84761 Cost: 0.91113 0.74218 0.08289 0.08606 Time: 0.00083
17-03-27 22:39:35 [1] Train Extra: lr=0.0000288 inv=0.4303125 sub=0.0000000
17-03-27 22:40:45 [1] Step: 81500 Eval acc: 0.66667 0.85267 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 22:40:45 [1] Eval Extra: inv=0.3849060
17-03-27 22:42:11 [1] Step: 81600 Acc: 0.67125 0.84529 Cost: 1.08814 0.85125 0.15081 0.08608 Time: 0.00081
17-03-27 22:42:11 [1] Train Extra: lr=0.0000287 inv=0.4137500 sub=0.0000000
17-03-27 22:43:48 [1] Step: 81700 Acc: 0.67188 0.85141 Cost: 0.94642 0.64652 0.21377 0.08613 Time: 0.00086
17-03-27 22:43:48 [1] Train Extra: lr=0.0000286 inv=0.4329688 sub=0.0000000
17-03-27 22:45:14 [1] Step: 81800 Acc: 0.68219 0.85376 Cost: 1.03207 0.67724 0.26871 0.08612 Time: 0.00082
17-03-27 22:45:14 [1] Train Extra: lr=0.0000285 inv=0.4104687 sub=0.0000000
17-03-27 22:46:34 [1] Step: 81900 Acc: 0.68344 0.85343 Cost: 0.81894 0.54383 0.18891 0.08620 Time: 0.00080
17-03-27 22:46:34 [1] Train Extra: lr=0.0000284 inv=0.4085937 sub=0.0000000
17-03-27 22:48:08 [1] Step: 82000 Acc: 0.67406 0.85955 Cost: 0.92211 0.66929 0.16656 0.08626 Time: 0.00088
17-03-27 22:48:08 [1] Train Extra: lr=0.0000284 inv=0.3932813 sub=0.0000000
17-03-27 22:49:19 [1] Step: 82000 Eval acc: 0.66953 0.85425 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-27 22:49:19 [1] Eval Extra: inv=0.3890931
17-03-27 22:50:57 [1] Step: 82100 Acc: 0.67219 0.84795 Cost: 0.92377 0.67103 0.16648 0.08627 Time: 0.00086
17-03-27 22:50:57 [1] Train Extra: lr=0.0000283 inv=0.4512500 sub=0.0000000
17-03-27 22:52:29 [1] Step: 82200 Acc: 0.67563 0.84891 Cost: 1.12006 0.76975 0.26397 0.08634 Time: 0.00085
17-03-27 22:52:29 [1] Train Extra: lr=0.0000282 inv=0.4375000 sub=0.0000000
17-03-27 22:53:48 [1] Step: 82300 Acc: 0.67812 0.85415 Cost: 1.29134 1.02086 0.18405 0.08644 Time: 0.00082
17-03-27 22:53:48 [1] Train Extra: lr=0.0000281 inv=0.3810938 sub=0.0000000
17-03-27 22:55:17 [1] Step: 82400 Acc: 0.66219 0.84981 Cost: 1.03591 0.71747 0.23193 0.08652 Time: 0.00083
17-03-27 22:55:17 [1] Train Extra: lr=0.0000280 inv=0.4132812 sub=0.0000000
17-03-27 22:56:49 [1] Step: 82500 Acc: 0.67312 0.85067 Cost: 1.45336 1.09677 0.27001 0.08658 Time: 0.00085
17-03-27 22:56:49 [1] Train Extra: lr=0.0000279 inv=0.4331250 sub=0.0000000
17-03-27 22:57:59 [1] Step: 82500 Eval acc: 0.66309 0.86095 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 22:57:59 [1] Eval Extra: inv=0.3901144
17-03-27 22:59:40 [1] Step: 82600 Acc: 0.67750 0.85349 Cost: 1.00543 0.78504 0.13388 0.08651 Time: 0.00086
17-03-27 22:59:40 [1] Train Extra: lr=0.0000279 inv=0.4175000 sub=0.0000000
17-03-27 23:00:59 [1] Step: 82700 Acc: 0.66250 0.85050 Cost: 1.17146 0.88574 0.19918 0.08654 Time: 0.00079
17-03-27 23:00:59 [1] Train Extra: lr=0.0000278 inv=0.4187500 sub=0.0000000
17-03-27 23:02:25 [1] Step: 82800 Acc: 0.66969 0.85568 Cost: 0.87825 0.70088 0.09083 0.08654 Time: 0.00082
17-03-27 23:02:25 [1] Train Extra: lr=0.0000277 inv=0.4309375 sub=0.0000000
17-03-27 23:04:02 [1] Step: 82900 Acc: 0.67812 0.85325 Cost: 1.05763 0.75773 0.21335 0.08655 Time: 0.00086
17-03-27 23:04:02 [1] Train Extra: lr=0.0000276 inv=0.4325000 sub=0.0000000
17-03-27 23:05:28 [1] Step: 83000 Acc: 0.67125 0.85792 Cost: 1.07646 0.66598 0.32389 0.08659 Time: 0.00083
17-03-27 23:05:28 [1] Train Extra: lr=0.0000276 inv=0.3848437 sub=0.0000000
17-03-27 23:06:37 [1] Step: 83000 Eval acc: 0.67596 0.86097 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 23:06:37 [1] Eval Extra: inv=0.3919016
17-03-27 23:08:10 [1] Step: 83100 Acc: 0.69094 0.85124 Cost: 1.29785 0.95538 0.25587 0.08661 Time: 0.00083
17-03-27 23:08:10 [1] Train Extra: lr=0.0000275 inv=0.4281250 sub=0.0000000
17-03-27 23:09:36 [1] Step: 83200 Acc: 0.67531 0.84634 Cost: 1.07329 0.78290 0.20371 0.08668 Time: 0.00080
17-03-27 23:09:36 [1] Train Extra: lr=0.0000274 inv=0.4228125 sub=0.0000000
17-03-27 23:10:59 [1] Step: 83300 Acc: 0.67437 0.84693 Cost: 1.08978 0.82201 0.18110 0.08666 Time: 0.00083
17-03-27 23:10:59 [1] Train Extra: lr=0.0000273 inv=0.4056250 sub=0.0000000
17-03-27 23:12:19 [1] Step: 83400 Acc: 0.67437 0.85485 Cost: 1.18648 0.79173 0.30800 0.08675 Time: 0.00080
17-03-27 23:12:19 [1] Train Extra: lr=0.0000272 inv=0.4070313 sub=0.0000000
17-03-27 23:13:56 [1] Step: 83500 Acc: 0.66969 0.85482 Cost: 0.79801 0.54949 0.16167 0.08685 Time: 0.00090
17-03-27 23:13:56 [1] Train Extra: lr=0.0000272 inv=0.4093750 sub=0.0000000
17-03-27 23:15:03 [1] Step: 83500 Eval acc: 0.66830 0.85554 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 23:15:03 [1] Eval Extra: inv=0.4104371
17-03-27 23:16:35 [1] Step: 83600 Acc: 0.68656 0.85035 Cost: 1.25095 0.90588 0.25821 0.08686 Time: 0.00084
17-03-27 23:16:35 [1] Train Extra: lr=0.0000271 inv=0.4335938 sub=0.0000000
17-03-27 23:18:06 [1] Step: 83700 Acc: 0.67437 0.84798 Cost: 0.78032 0.59478 0.09861 0.08693 Time: 0.00085
17-03-27 23:18:06 [1] Train Extra: lr=0.0000270 inv=0.4117188 sub=0.0000000
17-03-27 23:19:33 [1] Step: 83800 Acc: 0.66906 0.85238 Cost: 0.94080 0.62762 0.22619 0.08699 Time: 0.00081
17-03-27 23:19:33 [1] Train Extra: lr=0.0000269 inv=0.3971875 sub=0.0000000
17-03-27 23:21:04 [1] Step: 83900 Acc: 0.67156 0.84781 Cost: 0.89674 0.56767 0.24212 0.08695 Time: 0.00083
17-03-27 23:21:04 [1] Train Extra: lr=0.0000268 inv=0.4256250 sub=0.0000000
17-03-27 23:22:28 [1] Step: 84000 Acc: 0.67719 0.85090 Cost: 0.85818 0.61645 0.15473 0.08699 Time: 0.00082
17-03-27 23:22:28 [1] Train Extra: lr=0.0000268 inv=0.4054687 sub=0.0000000
17-03-27 23:23:37 [1] Step: 84000 Eval acc: 0.66350 0.85669 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 23:23:37 [1] Eval Extra: inv=0.4175858
17-03-27 23:25:02 [1] Step: 84100 Acc: 0.65969 0.85175 Cost: 0.84824 0.59958 0.16162 0.08704 Time: 0.00082
17-03-27 23:25:02 [1] Train Extra: lr=0.0000267 inv=0.4143750 sub=0.0000000
17-03-27 23:26:25 [1] Step: 84200 Acc: 0.66500 0.85639 Cost: 1.06600 0.84072 0.13825 0.08703 Time: 0.00083
17-03-27 23:26:25 [1] Train Extra: lr=0.0000266 inv=0.3843750 sub=0.0000000
17-03-27 23:27:54 [1] Step: 84300 Acc: 0.68031 0.85414 Cost: 1.12996 0.85802 0.18484 0.08711 Time: 0.00082
17-03-27 23:27:54 [1] Train Extra: lr=0.0000265 inv=0.4031250 sub=0.0000000
17-03-27 23:29:31 [1] Step: 84400 Acc: 0.67625 0.85129 Cost: 0.95283 0.61531 0.25036 0.08716 Time: 0.00086
17-03-27 23:29:31 [1] Train Extra: lr=0.0000265 inv=0.4428125 sub=0.0000000
17-03-27 23:31:03 [1] Step: 84500 Acc: 0.65500 0.85398 Cost: 1.38906 0.96486 0.33696 0.08725 Time: 0.00084
17-03-27 23:31:03 [1] Train Extra: lr=0.0000264 inv=0.4289062 sub=0.0000000
17-03-27 23:32:12 [1] Step: 84500 Eval acc: 0.67453 0.85968 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 23:32:12 [1] Eval Extra: inv=0.4426062
17-03-27 23:33:33 [1] Step: 84600 Acc: 0.67656 0.84945 Cost: 1.17498 0.84516 0.24263 0.08719 Time: 0.00079
17-03-27 23:33:33 [1] Train Extra: lr=0.0000263 inv=0.3920312 sub=0.0000000
17-03-27 23:35:20 [1] Step: 84700 Acc: 0.67656 0.85162 Cost: 0.73102 0.54663 0.09722 0.08718 Time: 0.00089
17-03-27 23:35:20 [1] Train Extra: lr=0.0000262 inv=0.4118750 sub=0.0000000
17-03-27 23:36:47 [1] Step: 84800 Acc: 0.67156 0.84890 Cost: 1.01841 0.71339 0.21779 0.08723 Time: 0.00083
17-03-27 23:36:47 [1] Train Extra: lr=0.0000262 inv=0.3950000 sub=0.0000000
17-03-27 23:38:16 [1] Step: 84900 Acc: 0.66469 0.84518 Cost: 0.96011 0.77214 0.10068 0.08729 Time: 0.00084
17-03-27 23:38:16 [1] Train Extra: lr=0.0000261 inv=0.4232812 sub=0.0000000
17-03-27 23:39:50 [1] Step: 85000 Acc: 0.67500 0.85123 Cost: 1.24569 0.90973 0.24858 0.08738 Time: 0.00083
17-03-27 23:39:50 [1] Train Extra: lr=0.0000260 inv=0.4215625 sub=0.0000000
17-03-27 23:40:59 [1] Step: 85000 Eval acc: 0.66432 0.85689 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 23:40:59 [1] Eval Extra: inv=0.4006332
17-03-27 23:40:59 [1] Checkpointing.
17-03-27 23:42:25 [1] Step: 85100 Acc: 0.66938 0.84552 Cost: 1.12327 0.77455 0.26132 0.08740 Time: 0.00081
17-03-27 23:42:25 [1] Train Extra: lr=0.0000259 inv=0.4018750 sub=0.0000000
17-03-27 23:43:50 [1] Step: 85200 Acc: 0.68219 0.85746 Cost: 1.13763 0.73378 0.31646 0.08739 Time: 0.00084
17-03-27 23:43:50 [1] Train Extra: lr=0.0000259 inv=0.3990625 sub=0.0000000
17-03-27 23:45:16 [1] Step: 85300 Acc: 0.66563 0.84632 Cost: 1.24707 0.87500 0.28465 0.08742 Time: 0.00081
17-03-27 23:45:16 [1] Train Extra: lr=0.0000258 inv=0.4140625 sub=0.0000000
17-03-27 23:46:48 [1] Step: 85400 Acc: 0.68031 0.85675 Cost: 1.21495 0.74099 0.38656 0.08740 Time: 0.00086
17-03-27 23:46:48 [1] Train Extra: lr=0.0000257 inv=0.4076562 sub=0.0000000
17-03-27 23:48:18 [1] Step: 85500 Acc: 0.66438 0.85168 Cost: 0.92762 0.67569 0.16454 0.08740 Time: 0.00086
17-03-27 23:48:18 [1] Train Extra: lr=0.0000256 inv=0.3962500 sub=0.0000000
17-03-27 23:49:24 [1] Step: 85500 Eval acc: 0.67361 0.85726 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-27 23:49:24 [1] Eval Extra: inv=0.4119179
17-03-27 23:50:45 [1] Step: 85600 Acc: 0.67000 0.84725 Cost: 1.01315 0.75025 0.17547 0.08743 Time: 0.00081
17-03-27 23:50:45 [1] Train Extra: lr=0.0000256 inv=0.4106250 sub=0.0000000
17-03-27 23:52:18 [1] Step: 85700 Acc: 0.67563 0.85707 Cost: 0.84230 0.57463 0.18022 0.08746 Time: 0.00087
17-03-27 23:52:18 [1] Train Extra: lr=0.0000255 inv=0.4143750 sub=0.0000000
17-03-27 23:53:53 [1] Step: 85800 Acc: 0.67219 0.85009 Cost: 0.94616 0.71650 0.14219 0.08747 Time: 0.00083
17-03-27 23:53:53 [1] Train Extra: lr=0.0000254 inv=0.4306250 sub=0.0000000
17-03-27 23:55:19 [1] Step: 85900 Acc: 0.68219 0.85603 Cost: 1.20573 0.83451 0.28371 0.08751 Time: 0.00084
17-03-27 23:55:19 [1] Train Extra: lr=0.0000253 inv=0.4028125 sub=0.0000000
17-03-27 23:56:46 [1] Step: 86000 Acc: 0.67000 0.85143 Cost: 1.14719 0.79584 0.26374 0.08761 Time: 0.00083
17-03-27 23:56:46 [1] Train Extra: lr=0.0000253 inv=0.4059375 sub=0.0000000
17-03-27 23:57:57 [1] Step: 86000 Eval acc: 0.66799 0.85742 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-27 23:57:57 [1] Eval Extra: inv=0.3794935
17-03-27 23:59:24 [1] Step: 86100 Acc: 0.67688 0.84465 Cost: 1.26934 0.94403 0.23768 0.08764 Time: 0.00082
17-03-27 23:59:24 [1] Train Extra: lr=0.0000252 inv=0.4310937 sub=0.0000000
17-03-28 00:00:56 [1] Step: 86200 Acc: 0.67781 0.85986 Cost: 1.06643 0.74358 0.23524 0.08761 Time: 0.00087
17-03-28 00:00:56 [1] Train Extra: lr=0.0000251 inv=0.4135937 sub=0.0000000
17-03-28 00:02:21 [1] Step: 86300 Acc: 0.67469 0.85195 Cost: 1.15121 0.81053 0.25305 0.08763 Time: 0.00081
17-03-28 00:02:21 [1] Train Extra: lr=0.0000251 inv=0.3906250 sub=0.0000000
17-03-28 00:04:04 [1] Step: 86400 Acc: 0.66281 0.85131 Cost: 0.86542 0.69242 0.08533 0.08767 Time: 0.00087
17-03-28 00:04:04 [1] Train Extra: lr=0.0000250 inv=0.4259375 sub=0.0000000
17-03-28 00:05:22 [1] Step: 86500 Acc: 0.67563 0.86068 Cost: 1.17864 0.91898 0.17198 0.08768 Time: 0.00082
17-03-28 00:05:22 [1] Train Extra: lr=0.0000249 inv=0.3787500 sub=0.0000000
17-03-28 00:06:32 [1] Step: 86500 Eval acc: 0.67535 0.85742 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 00:06:32 [1] Eval Extra: inv=0.3922590
17-03-28 00:07:58 [1] Step: 86600 Acc: 0.67937 0.85210 Cost: 1.03064 0.78069 0.16229 0.08767 Time: 0.00083
17-03-28 00:07:58 [1] Train Extra: lr=0.0000248 inv=0.3946875 sub=0.0000000
17-03-28 00:09:32 [1] Step: 86700 Acc: 0.66781 0.84457 Cost: 0.91869 0.65416 0.17681 0.08772 Time: 0.00081
17-03-28 00:09:32 [1] Train Extra: lr=0.0000248 inv=0.4385938 sub=0.0000000
17-03-28 00:11:00 [1] Step: 86800 Acc: 0.67531 0.85746 Cost: 0.74295 0.51218 0.14306 0.08772 Time: 0.00086
17-03-28 00:11:00 [1] Train Extra: lr=0.0000247 inv=0.3909375 sub=0.0000000
17-03-28 00:12:28 [1] Step: 86900 Acc: 0.66500 0.85066 Cost: 0.76619 0.52373 0.15478 0.08768 Time: 0.00081
17-03-28 00:12:28 [1] Train Extra: lr=0.0000246 inv=0.4253125 sub=0.0000000
17-03-28 00:14:00 [1] Step: 87000 Acc: 0.67437 0.85509 Cost: 0.93120 0.74992 0.09358 0.08769 Time: 0.00087
17-03-28 00:14:00 [1] Train Extra: lr=0.0000246 inv=0.3967188 sub=0.0000000
17-03-28 00:15:06 [1] Step: 87000 Eval acc: 0.67422 0.85742 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 00:15:06 [1] Eval Extra: inv=0.3813828
17-03-28 00:16:43 [1] Step: 87100 Acc: 0.65781 0.85983 Cost: 0.79773 0.56915 0.14087 0.08771 Time: 0.00089
17-03-28 00:16:43 [1] Train Extra: lr=0.0000245 inv=0.3823437 sub=0.0000000
17-03-28 00:18:07 [1] Step: 87200 Acc: 0.67156 0.84368 Cost: 1.25285 0.89936 0.26574 0.08775 Time: 0.00080
17-03-28 00:18:07 [1] Train Extra: lr=0.0000244 inv=0.4307813 sub=0.0000000
17-03-28 00:19:33 [1] Step: 87300 Acc: 0.66812 0.84865 Cost: 1.23411 0.91157 0.23478 0.08776 Time: 0.00082
17-03-28 00:19:33 [1] Train Extra: lr=0.0000243 inv=0.4162500 sub=0.0000000
17-03-28 00:21:13 [1] Step: 87400 Acc: 0.67594 0.86115 Cost: 0.78778 0.61899 0.08098 0.08781 Time: 0.00091
17-03-28 00:21:13 [1] Train Extra: lr=0.0000243 inv=0.4139062 sub=0.0000000
17-03-28 00:22:37 [1] Step: 87500 Acc: 0.67719 0.85233 Cost: 0.94804 0.62734 0.23288 0.08782 Time: 0.00083
17-03-28 00:22:37 [1] Train Extra: lr=0.0000242 inv=0.3996875 sub=0.0000000
17-03-28 00:23:44 [1] Step: 87500 Eval acc: 0.67208 0.85682 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 00:23:44 [1] Eval Extra: inv=0.4219261
17-03-28 00:25:07 [1] Step: 87600 Acc: 0.66000 0.84873 Cost: 1.07352 0.71148 0.27415 0.08789 Time: 0.00081
17-03-28 00:25:07 [1] Train Extra: lr=0.0000241 inv=0.3953125 sub=0.0000000
17-03-28 00:26:37 [1] Step: 87700 Acc: 0.67500 0.85908 Cost: 1.24767 0.92895 0.23080 0.08792 Time: 0.00086
17-03-28 00:26:37 [1] Train Extra: lr=0.0000241 inv=0.3821875 sub=0.0000000
17-03-28 00:28:10 [1] Step: 87800 Acc: 0.65812 0.85776 Cost: 1.05519 0.81218 0.15506 0.08795 Time: 0.00085
17-03-28 00:28:10 [1] Train Extra: lr=0.0000240 inv=0.3964063 sub=0.0000000
17-03-28 00:29:34 [1] Step: 87900 Acc: 0.67719 0.85192 Cost: 0.90798 0.70862 0.11145 0.08791 Time: 0.00083
17-03-28 00:29:34 [1] Train Extra: lr=0.0000239 inv=0.3967188 sub=0.0000000
17-03-28 00:31:02 [1] Step: 88000 Acc: 0.69312 0.85221 Cost: 0.97007 0.75538 0.12664 0.08805 Time: 0.00085
17-03-28 00:31:02 [1] Train Extra: lr=0.0000239 inv=0.3860938 sub=0.0000000
17-03-28 00:32:13 [1] Step: 88000 Eval acc: 0.67065 0.85694 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 00:32:13 [1] Eval Extra: inv=0.3866932
17-03-28 00:33:39 [1] Step: 88100 Acc: 0.68656 0.85540 Cost: 1.01626 0.69865 0.22935 0.08827 Time: 0.00084
17-03-28 00:33:39 [1] Train Extra: lr=0.0000238 inv=0.3853125 sub=0.0000000
17-03-28 00:35:05 [1] Step: 88200 Acc: 0.69000 0.85317 Cost: 1.10552 0.72817 0.28901 0.08833 Time: 0.00083
17-03-28 00:35:05 [1] Train Extra: lr=0.0000237 inv=0.4023438 sub=0.0000000
17-03-28 00:36:35 [1] Step: 88300 Acc: 0.69531 0.85234 Cost: 1.13985 0.88657 0.16480 0.08849 Time: 0.00084
17-03-28 00:36:35 [1] Train Extra: lr=0.0000237 inv=0.4314062 sub=0.0000000
17-03-28 00:37:58 [1] Step: 88400 Acc: 0.70000 0.85729 Cost: 0.82480 0.62605 0.11015 0.08860 Time: 0.00085
17-03-28 00:37:58 [1] Train Extra: lr=0.0000236 inv=0.3809375 sub=0.0000000
17-03-28 00:39:38 [1] Step: 88500 Acc: 0.69656 0.85754 Cost: 1.25448 0.81323 0.35256 0.08869 Time: 0.00089
17-03-28 00:39:38 [1] Train Extra: lr=0.0000235 inv=0.4042188 sub=0.0000000
17-03-28 00:40:43 [1] Step: 88500 Eval acc: 0.67770 0.85922 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-28 00:40:43 [1] Eval Extra: inv=0.4589971
17-03-28 00:42:19 [1] Step: 88600 Acc: 0.68344 0.85847 Cost: 1.04206 0.73669 0.21658 0.08879 Time: 0.00088
17-03-28 00:42:19 [1] Train Extra: lr=0.0000235 inv=0.4134375 sub=0.0000000
17-03-28 00:43:44 [1] Step: 88700 Acc: 0.69250 0.84821 Cost: 1.36255 0.92021 0.35348 0.08886 Time: 0.00080
17-03-28 00:43:44 [1] Train Extra: lr=0.0000234 inv=0.4246875 sub=0.0000000
17-03-28 00:45:16 [1] Step: 88800 Acc: 0.69156 0.84526 Cost: 1.26391 0.90985 0.26507 0.08899 Time: 0.00083
17-03-28 00:45:16 [1] Train Extra: lr=0.0000233 inv=0.4298438 sub=0.0000000
17-03-28 00:46:51 [1] Step: 88900 Acc: 0.68250 0.84482 Cost: 0.80634 0.60713 0.11015 0.08906 Time: 0.00081
17-03-28 00:46:51 [1] Train Extra: lr=0.0000232 inv=0.4343750 sub=0.0000000
17-03-28 00:48:24 [1] Step: 89000 Acc: 0.68625 0.84818 Cost: 1.11975 0.71429 0.31630 0.08917 Time: 0.00088
17-03-28 00:48:24 [1] Train Extra: lr=0.0000232 inv=0.4050000 sub=0.0000000
17-03-28 00:49:32 [1] Step: 89000 Eval acc: 0.67606 0.86140 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 00:49:32 [1] Eval Extra: inv=0.4084967
17-03-28 00:51:08 [1] Step: 89100 Acc: 0.66063 0.85703 Cost: 0.85940 0.61612 0.15407 0.08920 Time: 0.00088
17-03-28 00:51:08 [1] Train Extra: lr=0.0000231 inv=0.4207812 sub=0.0000000
17-03-28 00:52:53 [1] Step: 89200 Acc: 0.68312 0.85507 Cost: 0.93039 0.68943 0.15176 0.08921 Time: 0.00088
17-03-28 00:52:53 [1] Train Extra: lr=0.0000230 inv=0.4635938 sub=0.0000000
17-03-28 00:54:17 [1] Step: 89300 Acc: 0.69437 0.85392 Cost: 0.81553 0.56641 0.15979 0.08934 Time: 0.00083
17-03-28 00:54:17 [1] Train Extra: lr=0.0000230 inv=0.3917188 sub=0.0000000
17-03-28 00:55:47 [1] Step: 89400 Acc: 0.67063 0.84852 Cost: 0.98946 0.69239 0.20768 0.08939 Time: 0.00083
17-03-28 00:55:47 [1] Train Extra: lr=0.0000229 inv=0.4225000 sub=0.0000000
17-03-28 00:57:22 [1] Step: 89500 Acc: 0.68719 0.85863 Cost: 0.81298 0.53797 0.18557 0.08943 Time: 0.00086
17-03-28 00:57:22 [1] Train Extra: lr=0.0000229 inv=0.3965625 sub=0.0000000
17-03-28 00:58:32 [1] Step: 89500 Eval acc: 0.67913 0.85707 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 00:58:32 [1] Eval Extra: inv=0.3793403
17-03-28 00:58:32 [1] Checkpointing with new best dev accuracy of 0.679126
17-03-28 00:59:57 [1] Step: 89600 Acc: 0.69594 0.85230 Cost: 0.68320 0.48984 0.10381 0.08955 Time: 0.00083
17-03-28 00:59:57 [1] Train Extra: lr=0.0000228 inv=0.3917188 sub=0.0000000
17-03-28 01:01:23 [1] Step: 89700 Acc: 0.68563 0.84938 Cost: 1.06932 0.79885 0.18083 0.08964 Time: 0.00081
17-03-28 01:01:23 [1] Train Extra: lr=0.0000227 inv=0.4400000 sub=0.0000000
17-03-28 01:03:03 [1] Step: 89800 Acc: 0.67594 0.85525 Cost: 0.97876 0.54462 0.34442 0.08971 Time: 0.00087
17-03-28 01:03:03 [1] Train Extra: lr=0.0000227 inv=0.4220313 sub=0.0000000
17-03-28 01:04:32 [1] Step: 89900 Acc: 0.70156 0.85425 Cost: 1.01873 0.76281 0.16614 0.08978 Time: 0.00084
17-03-28 01:04:32 [1] Train Extra: lr=0.0000226 inv=0.4209375 sub=0.0000000
17-03-28 01:06:00 [1] Step: 90000 Acc: 0.68969 0.85305 Cost: 1.14808 0.80137 0.25682 0.08988 Time: 0.00083
17-03-28 01:06:00 [1] Train Extra: lr=0.0000225 inv=0.4075000 sub=0.0000000
17-03-28 01:07:09 [1] Step: 90000 Eval acc: 0.67596 0.85939 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 01:07:09 [1] Eval Extra: inv=0.4246324
17-03-28 01:07:09 [1] Checkpointing.
17-03-28 01:08:35 [1] Step: 90100 Acc: 0.69406 0.85738 Cost: 1.08290 0.87820 0.11470 0.08999 Time: 0.00084
17-03-28 01:08:35 [1] Train Extra: lr=0.0000225 inv=0.3950000 sub=0.0000000
17-03-28 01:10:06 [1] Step: 90200 Acc: 0.67875 0.84494 Cost: 0.85473 0.62565 0.13901 0.09007 Time: 0.00082
17-03-28 01:10:06 [1] Train Extra: lr=0.0000224 inv=0.4367187 sub=0.0000000
17-03-28 01:11:37 [1] Step: 90300 Acc: 0.67625 0.84139 Cost: 0.77504 0.60668 0.07820 0.09015 Time: 0.00081
17-03-28 01:11:37 [1] Train Extra: lr=0.0000223 inv=0.4228125 sub=0.0000000
17-03-28 01:13:02 [1] Step: 90400 Acc: 0.68219 0.84809 Cost: 0.88780 0.62208 0.17548 0.09025 Time: 0.00080
17-03-28 01:13:02 [1] Train Extra: lr=0.0000223 inv=0.4110937 sub=0.0000000
17-03-28 01:14:26 [1] Step: 90500 Acc: 0.68281 0.85277 Cost: 0.98172 0.70439 0.18699 0.09033 Time: 0.00083
17-03-28 01:14:26 [1] Train Extra: lr=0.0000222 inv=0.3903125 sub=0.0000000
17-03-28 01:15:35 [1] Step: 90500 Eval acc: 0.67177 0.85868 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 01:15:35 [1] Eval Extra: inv=0.4231516
17-03-28 01:17:01 [1] Step: 90600 Acc: 0.68750 0.85214 Cost: 0.98508 0.64253 0.25219 0.09036 Time: 0.00081
17-03-28 01:17:01 [1] Train Extra: lr=0.0000221 inv=0.4026562 sub=0.0000000
17-03-28 01:18:25 [1] Step: 90700 Acc: 0.68750 0.84758 Cost: 0.78844 0.53512 0.16290 0.09042 Time: 0.00082
17-03-28 01:18:25 [1] Train Extra: lr=0.0000221 inv=0.4270312 sub=0.0000000
17-03-28 01:19:52 [1] Step: 90800 Acc: 0.66656 0.84512 Cost: 1.19784 0.89708 0.21027 0.09049 Time: 0.00079
17-03-28 01:19:52 [1] Train Extra: lr=0.0000220 inv=0.4493750 sub=0.0000000
17-03-28 01:21:23 [1] Step: 90900 Acc: 0.68844 0.85051 Cost: 1.17111 0.83133 0.24923 0.09054 Time: 0.00083
17-03-28 01:21:23 [1] Train Extra: lr=0.0000219 inv=0.4321875 sub=0.0000000
17-03-28 01:22:59 [1] Step: 91000 Acc: 0.67281 0.84978 Cost: 0.86751 0.68747 0.08939 0.09065 Time: 0.00085
17-03-28 01:22:59 [1] Train Extra: lr=0.0000219 inv=0.4506250 sub=0.0000000
17-03-28 01:24:04 [1] Step: 91000 Eval acc: 0.67208 0.86112 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 01:24:04 [1] Eval Extra: inv=0.3849060
17-03-28 01:25:33 [1] Step: 91100 Acc: 0.69219 0.84960 Cost: 0.81890 0.59836 0.12986 0.09068 Time: 0.00084
17-03-28 01:25:33 [1] Train Extra: lr=0.0000218 inv=0.4125000 sub=0.0000000
17-03-28 01:27:05 [1] Step: 91200 Acc: 0.68625 0.85526 Cost: 0.91838 0.56641 0.26126 0.09071 Time: 0.00084
17-03-28 01:27:05 [1] Train Extra: lr=0.0000218 inv=0.4237500 sub=0.0000000
17-03-28 01:28:23 [1] Step: 91300 Acc: 0.68906 0.84732 Cost: 0.99750 0.61835 0.28838 0.09076 Time: 0.00078
17-03-28 01:28:23 [1] Train Extra: lr=0.0000217 inv=0.3998437 sub=0.0000000
17-03-28 01:30:02 [1] Step: 91400 Acc: 0.66719 0.84709 Cost: 0.91888 0.63313 0.19487 0.09087 Time: 0.00084
17-03-28 01:30:02 [1] Train Extra: lr=0.0000216 inv=0.4231250 sub=0.0000000
17-03-28 01:31:31 [1] Step: 91500 Acc: 0.68500 0.85128 Cost: 0.90203 0.56053 0.25058 0.09092 Time: 0.00084
17-03-28 01:31:31 [1] Train Extra: lr=0.0000216 inv=0.3990625 sub=0.0000000
17-03-28 01:32:37 [1] Step: 91500 Eval acc: 0.66718 0.85991 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 01:32:37 [1] Eval Extra: inv=0.4037480
17-03-28 01:34:02 [1] Step: 91600 Acc: 0.67781 0.85123 Cost: 0.94947 0.74350 0.11507 0.09090 Time: 0.00083
17-03-28 01:34:02 [1] Train Extra: lr=0.0000215 inv=0.4037500 sub=0.0000000
17-03-28 01:35:28 [1] Step: 91700 Acc: 0.67563 0.84511 Cost: 0.79778 0.53292 0.17387 0.09098 Time: 0.00080
17-03-28 01:35:28 [1] Train Extra: lr=0.0000215 inv=0.4182812 sub=0.0000000
17-03-28 01:36:58 [1] Step: 91800 Acc: 0.68469 0.84843 Cost: 0.91187 0.71878 0.10199 0.09109 Time: 0.00085
17-03-28 01:36:58 [1] Train Extra: lr=0.0000214 inv=0.4006250 sub=0.0000000
17-03-28 01:38:26 [1] Step: 91900 Acc: 0.67563 0.85367 Cost: 0.74097 0.48549 0.16436 0.09111 Time: 0.00083
17-03-28 01:38:26 [1] Train Extra: lr=0.0000213 inv=0.4071875 sub=0.0000000
17-03-28 01:39:58 [1] Step: 92000 Acc: 0.68469 0.85871 Cost: 1.13842 0.82600 0.22129 0.09113 Time: 0.00087
17-03-28 01:39:58 [1] Train Extra: lr=0.0000213 inv=0.3979687 sub=0.0000000
17-03-28 01:41:07 [1] Step: 92000 Eval acc: 0.67290 0.85646 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 01:41:07 [1] Eval Extra: inv=0.4442913
17-03-28 01:42:30 [1] Step: 92100 Acc: 0.69281 0.85673 Cost: 0.76895 0.58149 0.09627 0.09119 Time: 0.00084
17-03-28 01:42:30 [1] Train Extra: lr=0.0000212 inv=0.3864063 sub=0.0000000
17-03-28 01:43:51 [1] Step: 92200 Acc: 0.68531 0.84820 Cost: 1.05924 0.73936 0.22866 0.09123 Time: 0.00077
17-03-28 01:43:51 [1] Train Extra: lr=0.0000211 inv=0.4126563 sub=0.0000000
17-03-28 01:45:18 [1] Step: 92300 Acc: 0.67750 0.85638 Cost: 1.19592 0.81046 0.29412 0.09133 Time: 0.00086
17-03-28 01:45:18 [1] Train Extra: lr=0.0000211 inv=0.3948437 sub=0.0000000
17-03-28 01:46:38 [1] Step: 92400 Acc: 0.68875 0.85259 Cost: 1.30961 1.03534 0.18295 0.09132 Time: 0.00078
17-03-28 01:46:38 [1] Train Extra: lr=0.0000210 inv=0.4062500 sub=0.0000000
17-03-28 01:48:02 [1] Step: 92500 Acc: 0.69031 0.84547 Cost: 1.13387 0.84884 0.19364 0.09139 Time: 0.00080
17-03-28 01:48:02 [1] Train Extra: lr=0.0000210 inv=0.4167188 sub=0.0000000
17-03-28 01:49:11 [1] Step: 92500 Eval acc: 0.67198 0.86043 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 01:49:11 [1] Eval Extra: inv=0.3909314
17-03-28 01:50:35 [1] Step: 92600 Acc: 0.67625 0.84866 Cost: 1.19929 0.78387 0.32398 0.09144 Time: 0.00081
17-03-28 01:50:35 [1] Train Extra: lr=0.0000209 inv=0.4228125 sub=0.0000000
17-03-28 01:52:08 [1] Step: 92700 Acc: 0.67812 0.85657 Cost: 0.90301 0.65084 0.16069 0.09148 Time: 0.00084
17-03-28 01:52:08 [1] Train Extra: lr=0.0000208 inv=0.4323438 sub=0.0000000
17-03-28 01:53:34 [1] Step: 92800 Acc: 0.67969 0.84937 Cost: 1.12068 0.78893 0.24029 0.09147 Time: 0.00082
17-03-28 01:53:34 [1] Train Extra: lr=0.0000208 inv=0.4535938 sub=0.0000000
17-03-28 01:55:07 [1] Step: 92900 Acc: 0.68188 0.84955 Cost: 0.84309 0.61978 0.13180 0.09150 Time: 0.00085
17-03-28 01:55:07 [1] Train Extra: lr=0.0000207 inv=0.4054687 sub=0.0000000
17-03-28 01:56:33 [1] Step: 93000 Acc: 0.67750 0.85918 Cost: 1.16548 0.86298 0.21095 0.09155 Time: 0.00084
17-03-28 01:56:33 [1] Train Extra: lr=0.0000207 inv=0.4082812 sub=0.0000000
17-03-28 01:57:44 [1] Step: 93000 Eval acc: 0.67586 0.85675 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 01:57:44 [1] Eval Extra: inv=0.3710172
17-03-28 01:59:24 [1] Step: 93100 Acc: 0.68094 0.85019 Cost: 1.25081 0.84379 0.31543 0.09159 Time: 0.00085
17-03-28 01:59:24 [1] Train Extra: lr=0.0000206 inv=0.4492188 sub=0.0000000
17-03-28 02:00:55 [1] Step: 93200 Acc: 0.66594 0.85532 Cost: 1.26029 0.91482 0.25386 0.09161 Time: 0.00086
17-03-28 02:00:55 [1] Train Extra: lr=0.0000205 inv=0.4020313 sub=0.0000000
17-03-28 02:02:21 [1] Step: 93300 Acc: 0.67781 0.85749 Cost: 1.07562 0.76270 0.22130 0.09162 Time: 0.00084
17-03-28 02:02:21 [1] Train Extra: lr=0.0000205 inv=0.4023438 sub=0.0000000
17-03-28 02:03:53 [1] Step: 93400 Acc: 0.66750 0.85376 Cost: 1.17851 0.86849 0.21835 0.09167 Time: 0.00084
17-03-28 02:03:53 [1] Train Extra: lr=0.0000204 inv=0.4329688 sub=0.0000000
17-03-28 02:05:11 [1] Step: 93500 Acc: 0.69125 0.84993 Cost: 1.07112 0.82389 0.15549 0.09174 Time: 0.00080
17-03-28 02:05:11 [1] Train Extra: lr=0.0000204 inv=0.3775000 sub=0.0000000
17-03-28 02:06:20 [1] Step: 93500 Eval acc: 0.67463 0.85661 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 02:06:20 [1] Eval Extra: inv=0.4243770
17-03-28 02:07:51 [1] Step: 93600 Acc: 0.68156 0.85428 Cost: 0.77788 0.55259 0.13352 0.09177 Time: 0.00084
17-03-28 02:07:51 [1] Train Extra: lr=0.0000203 inv=0.4143750 sub=0.0000000
17-03-28 02:09:21 [1] Step: 93700 Acc: 0.68406 0.85341 Cost: 1.07791 0.77718 0.20896 0.09177 Time: 0.00084
17-03-28 02:09:21 [1] Train Extra: lr=0.0000203 inv=0.4054687 sub=0.0000000
17-03-28 02:10:48 [1] Step: 93800 Acc: 0.66469 0.84196 Cost: 0.60070 0.39532 0.11360 0.09178 Time: 0.00078
17-03-28 02:10:48 [1] Train Extra: lr=0.0000202 inv=0.4382813 sub=0.0000000
17-03-28 02:12:23 [1] Step: 93900 Acc: 0.67844 0.84127 Cost: 1.03899 0.67675 0.27041 0.09183 Time: 0.00082
17-03-28 02:12:23 [1] Train Extra: lr=0.0000201 inv=0.4635938 sub=0.0000000
17-03-28 02:13:50 [1] Step: 94000 Acc: 0.68781 0.85207 Cost: 0.87493 0.64339 0.13968 0.09186 Time: 0.00083
17-03-28 02:13:50 [1] Train Extra: lr=0.0000201 inv=0.4129687 sub=0.0000000
17-03-28 02:15:00 [1] Step: 94000 Eval acc: 0.66993 0.85593 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 02:15:00 [1] Eval Extra: inv=0.4070670
17-03-28 02:16:25 [1] Step: 94100 Acc: 0.67750 0.84890 Cost: 0.88871 0.67044 0.12636 0.09191 Time: 0.00081
17-03-28 02:16:25 [1] Train Extra: lr=0.0000200 inv=0.4120313 sub=0.0000000
17-03-28 02:18:00 [1] Step: 94200 Acc: 0.69031 0.85303 Cost: 1.01603 0.69106 0.23309 0.09187 Time: 0.00085
17-03-28 02:18:00 [1] Train Extra: lr=0.0000200 inv=0.4240625 sub=0.0000000
17-03-28 02:19:33 [1] Step: 94300 Acc: 0.67688 0.85654 Cost: 1.12773 0.86174 0.17409 0.09189 Time: 0.00087
17-03-28 02:19:33 [1] Train Extra: lr=0.0000199 inv=0.4015625 sub=0.0000000
17-03-28 02:20:50 [1] Step: 94400 Acc: 0.67625 0.85677 Cost: 1.09351 0.78324 0.21835 0.09192 Time: 0.00082
17-03-28 02:20:50 [1] Train Extra: lr=0.0000198 inv=0.3792187 sub=0.0000000
17-03-28 02:22:15 [1] Step: 94500 Acc: 0.67594 0.85249 Cost: 1.10437 0.85743 0.15498 0.09196 Time: 0.00082
17-03-28 02:22:15 [1] Train Extra: lr=0.0000198 inv=0.4098438 sub=0.0000000
17-03-28 02:23:25 [1] Step: 94500 Eval acc: 0.67085 0.85471 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 02:23:25 [1] Eval Extra: inv=0.4322406
17-03-28 02:24:56 [1] Step: 94600 Acc: 0.66938 0.85352 Cost: 1.26770 1.00124 0.17447 0.09199 Time: 0.00082
17-03-28 02:24:56 [1] Train Extra: lr=0.0000197 inv=0.4139062 sub=0.0000000
17-03-28 02:26:21 [1] Step: 94700 Acc: 0.67781 0.85145 Cost: 1.01046 0.77700 0.14147 0.09199 Time: 0.00083
17-03-28 02:26:21 [1] Train Extra: lr=0.0000197 inv=0.4292187 sub=0.0000000
17-03-28 02:27:39 [1] Step: 94800 Acc: 0.68281 0.85673 Cost: 1.15129 0.85729 0.20192 0.09207 Time: 0.00080
17-03-28 02:27:39 [1] Train Extra: lr=0.0000196 inv=0.4217187 sub=0.0000000
17-03-28 02:29:09 [1] Step: 94900 Acc: 0.66969 0.84982 Cost: 1.03377 0.83515 0.10658 0.09204 Time: 0.00085
17-03-28 02:29:09 [1] Train Extra: lr=0.0000196 inv=0.4078125 sub=0.0000000
17-03-28 02:30:42 [1] Step: 95000 Acc: 0.68531 0.85715 Cost: 1.06322 0.73899 0.23215 0.09208 Time: 0.00084
17-03-28 02:30:42 [1] Train Extra: lr=0.0000195 inv=0.4232812 sub=0.0000000
17-03-28 02:31:47 [1] Step: 95000 Eval acc: 0.67055 0.85974 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-28 02:31:47 [1] Eval Extra: inv=0.4201389
17-03-28 02:31:47 [1] Checkpointing.
17-03-28 02:33:16 [1] Step: 95100 Acc: 0.67500 0.85598 Cost: 0.98215 0.67975 0.21034 0.09205 Time: 0.00085
17-03-28 02:33:16 [1] Train Extra: lr=0.0000195 inv=0.4004687 sub=0.0000000
17-03-28 02:34:51 [1] Step: 95200 Acc: 0.67563 0.85307 Cost: 1.12422 0.74986 0.28226 0.09210 Time: 0.00083
17-03-28 02:34:51 [1] Train Extra: lr=0.0000194 inv=0.4184375 sub=0.0000000
17-03-28 02:36:12 [1] Step: 95300 Acc: 0.69000 0.86042 Cost: 1.17720 0.75511 0.32992 0.09217 Time: 0.00085
17-03-28 02:36:12 [1] Train Extra: lr=0.0000193 inv=0.3676563 sub=0.0000000
17-03-28 02:37:36 [1] Step: 95400 Acc: 0.67781 0.85053 Cost: 0.91511 0.56056 0.26241 0.09213 Time: 0.00080
17-03-28 02:37:36 [1] Train Extra: lr=0.0000193 inv=0.4154687 sub=0.0000000
17-03-28 02:38:59 [1] Step: 95500 Acc: 0.68906 0.85918 Cost: 1.10477 0.78841 0.22415 0.09221 Time: 0.00084
17-03-28 02:38:59 [1] Train Extra: lr=0.0000192 inv=0.4171875 sub=0.0000000
17-03-28 02:40:07 [1] Step: 95500 Eval acc: 0.67330 0.85778 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 02:40:07 [1] Eval Extra: inv=0.4025225
17-03-28 02:41:25 [1] Step: 95600 Acc: 0.68688 0.85303 Cost: 0.93864 0.70259 0.14384 0.09221 Time: 0.00081
17-03-28 02:41:25 [1] Train Extra: lr=0.0000192 inv=0.3789062 sub=0.0000000
17-03-28 02:42:56 [1] Step: 95700 Acc: 0.66594 0.85872 Cost: 1.07627 0.69056 0.29347 0.09225 Time: 0.00088
17-03-28 02:42:56 [1] Train Extra: lr=0.0000191 inv=0.3728125 sub=0.0000000
17-03-28 02:44:36 [1] Step: 95800 Acc: 0.66969 0.85864 Cost: 1.18717 0.82603 0.26889 0.09225 Time: 0.00088
17-03-28 02:44:36 [1] Train Extra: lr=0.0000191 inv=0.4253125 sub=0.0000000
17-03-28 02:46:01 [1] Step: 95900 Acc: 0.67688 0.84800 Cost: 0.98578 0.81019 0.08334 0.09225 Time: 0.00080
17-03-28 02:46:01 [1] Train Extra: lr=0.0000190 inv=0.3989063 sub=0.0000000
17-03-28 02:47:44 [1] Step: 96000 Acc: 0.67719 0.83848 Cost: 0.75966 0.58332 0.08405 0.09229 Time: 0.00085
17-03-28 02:47:44 [1] Train Extra: lr=0.0000190 inv=0.4734375 sub=0.0000000
17-03-28 02:48:52 [1] Step: 96000 Eval acc: 0.67964 0.85622 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 02:48:52 [1] Eval Extra: inv=0.4030331
17-03-28 02:50:18 [1] Step: 96100 Acc: 0.67625 0.85649 Cost: 0.85461 0.69527 0.06701 0.09233 Time: 0.00085
17-03-28 02:50:18 [1] Train Extra: lr=0.0000189 inv=0.3956250 sub=0.0000000
17-03-28 02:51:48 [1] Step: 96200 Acc: 0.68594 0.86008 Cost: 1.06310 0.76595 0.20478 0.09237 Time: 0.00089
17-03-28 02:51:48 [1] Train Extra: lr=0.0000188 inv=0.3915625 sub=0.0000000
17-03-28 02:53:23 [1] Step: 96300 Acc: 0.68188 0.84691 Cost: 1.06454 0.65612 0.31605 0.09237 Time: 0.00082
17-03-28 02:53:23 [1] Train Extra: lr=0.0000188 inv=0.4531250 sub=0.0000000
17-03-28 02:55:01 [1] Step: 96400 Acc: 0.66312 0.85801 Cost: 0.85203 0.64654 0.11309 0.09240 Time: 0.00088
17-03-28 02:55:01 [1] Train Extra: lr=0.0000187 inv=0.4076562 sub=0.0000000
17-03-28 02:56:23 [1] Step: 96500 Acc: 0.68437 0.85660 Cost: 0.92634 0.60873 0.22526 0.09235 Time: 0.00084
17-03-28 02:56:23 [1] Train Extra: lr=0.0000187 inv=0.3803125 sub=0.0000000
17-03-28 02:57:30 [1] Step: 96500 Eval acc: 0.67422 0.86021 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 02:57:30 [1] Eval Extra: inv=0.3741830
17-03-28 02:59:00 [1] Step: 96600 Acc: 0.68125 0.84813 Cost: 0.99461 0.62290 0.27933 0.09238 Time: 0.00084
17-03-28 02:59:00 [1] Train Extra: lr=0.0000186 inv=0.4207812 sub=0.0000000
17-03-28 03:00:35 [1] Step: 96700 Acc: 0.68625 0.84792 Cost: 1.10636 0.85889 0.15495 0.09251 Time: 0.00084
17-03-28 03:00:35 [1] Train Extra: lr=0.0000186 inv=0.4248438 sub=0.0000000
17-03-28 03:02:15 [1] Step: 96800 Acc: 0.67500 0.85527 Cost: 1.20910 0.88029 0.23631 0.09249 Time: 0.00086
17-03-28 03:02:15 [1] Train Extra: lr=0.0000185 inv=0.4301563 sub=0.0000000
17-03-28 03:03:41 [1] Step: 96900 Acc: 0.67531 0.85426 Cost: 1.22525 0.95558 0.17713 0.09254 Time: 0.00084
17-03-28 03:03:41 [1] Train Extra: lr=0.0000185 inv=0.3984375 sub=0.0000000
17-03-28 03:05:12 [1] Step: 97000 Acc: 0.67812 0.84087 Cost: 1.10025 0.76428 0.24340 0.09258 Time: 0.00082
17-03-28 03:05:12 [1] Train Extra: lr=0.0000184 inv=0.4201563 sub=0.0000000
17-03-28 03:06:21 [1] Step: 97000 Eval acc: 0.66871 0.85674 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 03:06:21 [1] Eval Extra: inv=0.4003779
17-03-28 03:07:55 [1] Step: 97100 Acc: 0.68344 0.84800 Cost: 0.97564 0.73015 0.15287 0.09262 Time: 0.00084
17-03-28 03:07:55 [1] Train Extra: lr=0.0000184 inv=0.4470312 sub=0.0000000
17-03-28 03:09:36 [1] Step: 97200 Acc: 0.65531 0.85549 Cost: 1.24306 0.88044 0.27005 0.09257 Time: 0.00086
17-03-28 03:09:36 [1] Train Extra: lr=0.0000183 inv=0.4325000 sub=0.0000000
17-03-28 03:11:07 [1] Step: 97300 Acc: 0.68031 0.85378 Cost: 1.36491 1.04643 0.22591 0.09257 Time: 0.00085
17-03-28 03:11:07 [1] Train Extra: lr=0.0000183 inv=0.4165625 sub=0.0000000
17-03-28 03:12:27 [1] Step: 97400 Acc: 0.67969 0.85761 Cost: 0.73618 0.52804 0.11565 0.09249 Time: 0.00082
17-03-28 03:12:27 [1] Train Extra: lr=0.0000182 inv=0.3750000 sub=0.0000000
17-03-28 03:13:59 [1] Step: 97500 Acc: 0.68906 0.85047 Cost: 1.40335 1.06130 0.24954 0.09251 Time: 0.00086
17-03-28 03:13:59 [1] Train Extra: lr=0.0000182 inv=0.4142188 sub=0.0000000
17-03-28 03:15:09 [1] Step: 97500 Eval acc: 0.67320 0.85807 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 03:15:09 [1] Eval Extra: inv=0.3814849
17-03-28 03:16:28 [1] Step: 97600 Acc: 0.67063 0.86056 Cost: 0.82052 0.62434 0.10369 0.09248 Time: 0.00079
17-03-28 03:16:28 [1] Train Extra: lr=0.0000181 inv=0.3750000 sub=0.0000000
17-03-28 03:18:02 [1] Step: 97700 Acc: 0.66469 0.85202 Cost: 0.91845 0.67717 0.14878 0.09249 Time: 0.00085
17-03-28 03:18:02 [1] Train Extra: lr=0.0000180 inv=0.4162500 sub=0.0000000
17-03-28 03:19:26 [1] Step: 97800 Acc: 0.68969 0.85498 Cost: 0.97283 0.66977 0.21045 0.09261 Time: 0.00082
17-03-28 03:19:26 [1] Train Extra: lr=0.0000180 inv=0.4178125 sub=0.0000000
17-03-28 03:20:47 [1] Step: 97900 Acc: 0.71313 0.85262 Cost: 1.05903 0.73128 0.23502 0.09273 Time: 0.00082
17-03-28 03:20:47 [1] Train Extra: lr=0.0000179 inv=0.3937500 sub=0.0000000
17-03-28 03:22:14 [1] Step: 98000 Acc: 0.70156 0.85299 Cost: 0.87604 0.66350 0.11971 0.09282 Time: 0.00080
17-03-28 03:22:14 [1] Train Extra: lr=0.0000179 inv=0.4245313 sub=0.0000000
17-03-28 03:23:23 [1] Step: 98000 Eval acc: 0.67249 0.86046 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 03:23:23 [1] Eval Extra: inv=0.3899101
17-03-28 03:24:45 [1] Step: 98100 Acc: 0.69312 0.85064 Cost: 1.07735 0.82320 0.16121 0.09295 Time: 0.00082
17-03-28 03:24:45 [1] Train Extra: lr=0.0000178 inv=0.4143750 sub=0.0000000
17-03-28 03:26:13 [1] Step: 98200 Acc: 0.70000 0.84908 Cost: 1.00852 0.59507 0.32039 0.09307 Time: 0.00080
17-03-28 03:26:13 [1] Train Extra: lr=0.0000178 inv=0.4362500 sub=0.0000000
17-03-28 03:27:42 [1] Step: 98300 Acc: 0.70594 0.85748 Cost: 1.23887 0.82392 0.32181 0.09315 Time: 0.00084
17-03-28 03:27:42 [1] Train Extra: lr=0.0000177 inv=0.4117188 sub=0.0000000
17-03-28 03:29:01 [1] Step: 98400 Acc: 0.70469 0.85903 Cost: 0.94282 0.60474 0.24484 0.09323 Time: 0.00083
17-03-28 03:29:01 [1] Train Extra: lr=0.0000177 inv=0.3782813 sub=0.0000000
17-03-28 03:30:40 [1] Step: 98500 Acc: 0.70094 0.85792 Cost: 1.04576 0.74673 0.20566 0.09336 Time: 0.00089
17-03-28 03:30:40 [1] Train Extra: lr=0.0000176 inv=0.4242187 sub=0.0000000
17-03-28 03:31:46 [1] Step: 98500 Eval acc: 0.67014 0.85975 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 03:31:46 [1] Eval Extra: inv=0.3893484
17-03-28 03:33:09 [1] Step: 98600 Acc: 0.69688 0.85056 Cost: 1.30210 0.92720 0.28148 0.09341 Time: 0.00081
17-03-28 03:33:09 [1] Train Extra: lr=0.0000176 inv=0.4007812 sub=0.0000000
17-03-28 03:34:49 [1] Step: 98700 Acc: 0.68500 0.85498 Cost: 0.86527 0.57341 0.19827 0.09359 Time: 0.00085
17-03-28 03:34:49 [1] Train Extra: lr=0.0000175 inv=0.4298438 sub=0.0000000
17-03-28 03:36:19 [1] Step: 98800 Acc: 0.69781 0.85442 Cost: 1.12851 0.84734 0.18750 0.09367 Time: 0.00084
17-03-28 03:36:19 [1] Train Extra: lr=0.0000175 inv=0.4310937 sub=0.0000000
17-03-28 03:37:46 [1] Step: 98900 Acc: 0.69688 0.86223 Cost: 1.08457 0.75659 0.23425 0.09373 Time: 0.00084
17-03-28 03:37:46 [1] Train Extra: lr=0.0000174 inv=0.3904687 sub=0.0000000
17-03-28 03:39:15 [1] Step: 99000 Acc: 0.70344 0.84454 Cost: 1.03690 0.66027 0.28283 0.09380 Time: 0.00082
17-03-28 03:39:15 [1] Train Extra: lr=0.0000174 inv=0.4387500 sub=0.0000000
17-03-28 03:40:20 [1] Step: 99000 Eval acc: 0.67228 0.85968 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-28 03:40:20 [1] Eval Extra: inv=0.4070159
17-03-28 03:41:50 [1] Step: 99100 Acc: 0.67781 0.85465 Cost: 1.03378 0.72885 0.21106 0.09386 Time: 0.00083
17-03-28 03:41:50 [1] Train Extra: lr=0.0000173 inv=0.3978125 sub=0.0000000
17-03-28 03:43:22 [1] Step: 99200 Acc: 0.68719 0.84843 Cost: 1.15303 0.86031 0.19874 0.09398 Time: 0.00082
17-03-28 03:43:22 [1] Train Extra: lr=0.0000173 inv=0.4346875 sub=0.0000000
17-03-28 03:44:45 [1] Step: 99300 Acc: 0.69156 0.85251 Cost: 1.12337 0.82476 0.20456 0.09406 Time: 0.00082
17-03-28 03:44:45 [1] Train Extra: lr=0.0000172 inv=0.4103125 sub=0.0000000
17-03-28 03:46:15 [1] Step: 99400 Acc: 0.68594 0.85290 Cost: 0.84192 0.64795 0.09984 0.09413 Time: 0.00083
17-03-28 03:46:15 [1] Train Extra: lr=0.0000172 inv=0.4125000 sub=0.0000000
17-03-28 03:47:48 [1] Step: 99500 Acc: 0.68937 0.85128 Cost: 1.21103 0.86032 0.25650 0.09421 Time: 0.00086
17-03-28 03:47:48 [1] Train Extra: lr=0.0000171 inv=0.4148438 sub=0.0000000
17-03-28 03:48:59 [1] Step: 99500 Eval acc: 0.67157 0.85851 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 03:48:59 [1] Eval Extra: inv=0.3944036
17-03-28 03:50:27 [1] Step: 99600 Acc: 0.69875 0.85460 Cost: 1.17819 0.78885 0.29504 0.09431 Time: 0.00083
17-03-28 03:50:27 [1] Train Extra: lr=0.0000171 inv=0.4279688 sub=0.0000000
17-03-28 03:52:05 [1] Step: 99700 Acc: 0.68750 0.85684 Cost: 1.12367 0.77413 0.25511 0.09443 Time: 0.00087
17-03-28 03:52:05 [1] Train Extra: lr=0.0000170 inv=0.4381250 sub=0.0000000
17-03-28 03:53:37 [1] Step: 99800 Acc: 0.69156 0.85369 Cost: 1.13396 0.83658 0.20283 0.09455 Time: 0.00086
17-03-28 03:53:37 [1] Train Extra: lr=0.0000170 inv=0.3990625 sub=0.0000000
17-03-28 03:55:18 [1] Step: 99900 Acc: 0.69469 0.85224 Cost: 1.07419 0.80207 0.17749 0.09463 Time: 0.00088
17-03-28 03:55:18 [1] Train Extra: lr=0.0000169 inv=0.4276563 sub=0.0000000
17-03-28 03:56:59 [1] Step: 100000 Acc: 0.68531 0.85444 Cost: 1.03657 0.73019 0.21168 0.09470 Time: 0.00088
17-03-28 03:56:59 [1] Train Extra: lr=0.0000169 inv=0.4217187 sub=0.0000000
17-03-28 03:58:07 [1] Step: 100000 Eval acc: 0.67361 0.86232 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 03:58:07 [1] Eval Extra: inv=0.3862847
17-03-28 03:58:07 [1] Checkpointing.
17-03-28 03:59:37 [1] Step: 100100 Acc: 0.69281 0.84830 Cost: 1.03184 0.76033 0.17677 0.09475 Time: 0.00087
17-03-28 03:59:37 [1] Train Extra: lr=0.0000168 inv=0.3712500 sub=0.0000000
17-03-28 04:01:04 [1] Step: 100200 Acc: 0.69375 0.85212 Cost: 1.21734 0.80544 0.31709 0.09481 Time: 0.00083
17-03-28 04:01:04 [1] Train Extra: lr=0.0000168 inv=0.3995313 sub=0.0000000
17-03-28 04:02:31 [1] Step: 100300 Acc: 0.68000 0.85350 Cost: 0.91591 0.70171 0.11941 0.09479 Time: 0.00082
17-03-28 04:02:31 [1] Train Extra: lr=0.0000167 inv=0.3812500 sub=0.0000000
17-03-28 04:03:59 [1] Step: 100400 Acc: 0.68844 0.85457 Cost: 0.84615 0.53797 0.21326 0.09492 Time: 0.00085
17-03-28 04:03:59 [1] Train Extra: lr=0.0000167 inv=0.4114062 sub=0.0000000
17-03-28 04:05:24 [1] Step: 100500 Acc: 0.69094 0.85266 Cost: 1.17616 0.97212 0.10903 0.09501 Time: 0.00082
17-03-28 04:05:24 [1] Train Extra: lr=0.0000167 inv=0.3878125 sub=0.0000000
17-03-28 04:06:32 [1] Step: 100500 Eval acc: 0.67412 0.85871 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 04:06:32 [1] Eval Extra: inv=0.3947100
17-03-28 04:07:58 [1] Step: 100600 Acc: 0.69000 0.85666 Cost: 0.91694 0.63115 0.19070 0.09510 Time: 0.00081
17-03-28 04:07:58 [1] Train Extra: lr=0.0000166 inv=0.3806250 sub=0.0000000
17-03-28 04:09:35 [1] Step: 100700 Acc: 0.69094 0.85203 Cost: 0.96798 0.65984 0.21300 0.09514 Time: 0.00085
17-03-28 04:09:35 [1] Train Extra: lr=0.0000166 inv=0.4201563 sub=0.0000000
17-03-28 04:11:00 [1] Step: 100800 Acc: 0.69406 0.85329 Cost: 0.86691 0.56333 0.20832 0.09527 Time: 0.00081
17-03-28 04:11:00 [1] Train Extra: lr=0.0000165 inv=0.4176563 sub=0.0000000
17-03-28 04:12:36 [1] Step: 100900 Acc: 0.69500 0.84760 Cost: 0.86272 0.62945 0.13793 0.09534 Time: 0.00084
17-03-28 04:12:36 [1] Train Extra: lr=0.0000165 inv=0.4390625 sub=0.0000000
17-03-28 04:14:07 [1] Step: 101000 Acc: 0.69312 0.85994 Cost: 1.23726 0.89241 0.24952 0.09533 Time: 0.00085
17-03-28 04:14:07 [1] Train Extra: lr=0.0000164 inv=0.3890625 sub=0.0000000
17-03-28 04:15:13 [1] Step: 101000 Eval acc: 0.67800 0.86020 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 04:15:13 [1] Eval Extra: inv=0.4043096
17-03-28 04:16:37 [1] Step: 101100 Acc: 0.70375 0.85864 Cost: 1.18014 0.90864 0.17607 0.09543 Time: 0.00082
17-03-28 04:16:37 [1] Train Extra: lr=0.0000164 inv=0.3910938 sub=0.0000000
17-03-28 04:18:13 [1] Step: 101200 Acc: 0.66312 0.84940 Cost: 0.99613 0.60725 0.29332 0.09556 Time: 0.00086
17-03-28 04:18:13 [1] Train Extra: lr=0.0000163 inv=0.4300000 sub=0.0000000
17-03-28 04:19:46 [1] Step: 101300 Acc: 0.68031 0.84807 Cost: 0.81686 0.45101 0.27025 0.09560 Time: 0.00082
17-03-28 04:19:46 [1] Train Extra: lr=0.0000163 inv=0.4301563 sub=0.0000000
17-03-28 04:21:16 [1] Step: 101400 Acc: 0.69094 0.85846 Cost: 0.96491 0.66074 0.20855 0.09562 Time: 0.00086
17-03-28 04:21:16 [1] Train Extra: lr=0.0000162 inv=0.3970313 sub=0.0000000
17-03-28 04:22:44 [1] Step: 101500 Acc: 0.68937 0.85588 Cost: 1.04914 0.69564 0.25784 0.09567 Time: 0.00083
17-03-28 04:22:44 [1] Train Extra: lr=0.0000162 inv=0.4128125 sub=0.0000000
17-03-28 04:23:53 [1] Step: 101500 Eval acc: 0.67555 0.85806 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 04:23:53 [1] Eval Extra: inv=0.4146752
17-03-28 04:25:16 [1] Step: 101600 Acc: 0.69937 0.85062 Cost: 1.13914 0.80396 0.23945 0.09574 Time: 0.00083
17-03-28 04:25:16 [1] Train Extra: lr=0.0000161 inv=0.3879687 sub=0.0000000
17-03-28 04:26:40 [1] Step: 101700 Acc: 0.70906 0.85727 Cost: 0.99958 0.69462 0.20919 0.09577 Time: 0.00083
17-03-28 04:26:40 [1] Train Extra: lr=0.0000161 inv=0.4096875 sub=0.0000000
17-03-28 04:28:05 [1] Step: 101800 Acc: 0.68281 0.84934 Cost: 1.17808 0.81904 0.26325 0.09579 Time: 0.00080
17-03-28 04:28:05 [1] Train Extra: lr=0.0000160 inv=0.4237500 sub=0.0000000
17-03-28 04:29:32 [1] Step: 101900 Acc: 0.69219 0.85124 Cost: 0.99362 0.65893 0.23877 0.09593 Time: 0.00081
17-03-28 04:29:32 [1] Train Extra: lr=0.0000160 inv=0.4009375 sub=0.0000000
17-03-28 04:31:02 [1] Step: 102000 Acc: 0.66969 0.85651 Cost: 1.25938 0.95279 0.21060 0.09598 Time: 0.00086
17-03-28 04:31:02 [1] Train Extra: lr=0.0000159 inv=0.3903125 sub=0.0000000
17-03-28 04:32:07 [1] Step: 102000 Eval acc: 0.67586 0.85507 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-28 04:32:07 [1] Eval Extra: inv=0.4043607
17-03-28 04:33:36 [1] Step: 102100 Acc: 0.67719 0.85092 Cost: 1.20756 0.87006 0.24147 0.09604 Time: 0.00083
17-03-28 04:33:36 [1] Train Extra: lr=0.0000159 inv=0.4112500 sub=0.0000000
17-03-28 04:35:07 [1] Step: 102200 Acc: 0.69812 0.85194 Cost: 1.20788 0.90520 0.20661 0.09607 Time: 0.00084
17-03-28 04:35:07 [1] Train Extra: lr=0.0000159 inv=0.4034375 sub=0.0000000
17-03-28 04:36:30 [1] Step: 102300 Acc: 0.70188 0.85336 Cost: 0.76289 0.51086 0.15591 0.09612 Time: 0.00082
17-03-28 04:36:30 [1] Train Extra: lr=0.0000158 inv=0.4157812 sub=0.0000000
17-03-28 04:38:03 [1] Step: 102400 Acc: 0.67781 0.85295 Cost: 0.81835 0.53626 0.18586 0.09622 Time: 0.00083
17-03-28 04:38:03 [1] Train Extra: lr=0.0000158 inv=0.4276563 sub=0.0000000
17-03-28 04:39:26 [1] Step: 102500 Acc: 0.69344 0.84900 Cost: 1.24030 0.85402 0.29008 0.09620 Time: 0.00081
17-03-28 04:39:26 [1] Train Extra: lr=0.0000157 inv=0.4267187 sub=0.0000000
17-03-28 04:40:35 [1] Step: 102500 Eval acc: 0.67116 0.85981 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 04:40:35 [1] Eval Extra: inv=0.4287684
17-03-28 04:42:00 [1] Step: 102600 Acc: 0.68375 0.85135 Cost: 0.66874 0.43966 0.13280 0.09629 Time: 0.00080
17-03-28 04:42:00 [1] Train Extra: lr=0.0000157 inv=0.4231250 sub=0.0000000
17-03-28 04:43:26 [1] Step: 102700 Acc: 0.69063 0.85752 Cost: 0.92525 0.58013 0.24879 0.09633 Time: 0.00083
17-03-28 04:43:26 [1] Train Extra: lr=0.0000156 inv=0.3762500 sub=0.0000000
17-03-28 04:45:04 [1] Step: 102800 Acc: 0.66938 0.85025 Cost: 0.91487 0.63728 0.18125 0.09635 Time: 0.00085
17-03-28 04:45:04 [1] Train Extra: lr=0.0000156 inv=0.4096875 sub=0.0000000
17-03-28 04:46:33 [1] Step: 102900 Acc: 0.68094 0.85224 Cost: 1.16625 0.86431 0.20555 0.09639 Time: 0.00083
17-03-28 04:46:33 [1] Train Extra: lr=0.0000155 inv=0.4023438 sub=0.0000000
17-03-28 04:48:03 [1] Step: 103000 Acc: 0.69250 0.85023 Cost: 1.02399 0.72650 0.20095 0.09654 Time: 0.00083
17-03-28 04:48:03 [1] Train Extra: lr=0.0000155 inv=0.4196875 sub=0.0000000
17-03-28 04:49:12 [1] Step: 103000 Eval acc: 0.67371 0.85783 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 04:49:12 [1] Eval Extra: inv=0.3715788
17-03-28 04:50:31 [1] Step: 103100 Acc: 0.68531 0.85507 Cost: 0.60563 0.42107 0.08799 0.09657 Time: 0.00079
17-03-28 04:50:31 [1] Train Extra: lr=0.0000155 inv=0.3900000 sub=0.0000000
17-03-28 04:51:54 [1] Step: 103200 Acc: 0.68844 0.85373 Cost: 1.15320 0.83944 0.21718 0.09658 Time: 0.00083
17-03-28 04:51:54 [1] Train Extra: lr=0.0000154 inv=0.4106250 sub=0.0000000
17-03-28 04:53:26 [1] Step: 103300 Acc: 0.68563 0.85564 Cost: 1.30931 1.06462 0.14809 0.09660 Time: 0.00084
17-03-28 04:53:26 [1] Train Extra: lr=0.0000154 inv=0.4303125 sub=0.0000000
17-03-28 04:55:04 [1] Step: 103400 Acc: 0.69156 0.85455 Cost: 1.00149 0.61060 0.29431 0.09659 Time: 0.00087
17-03-28 04:55:04 [1] Train Extra: lr=0.0000153 inv=0.4201563 sub=0.0000000
17-03-28 04:56:36 [1] Step: 103500 Acc: 0.69219 0.85865 Cost: 0.87533 0.60044 0.17838 0.09652 Time: 0.00086
17-03-28 04:56:36 [1] Train Extra: lr=0.0000153 inv=0.4190625 sub=0.0000000
17-03-28 04:57:41 [1] Step: 103500 Eval acc: 0.67249 0.86135 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-28 04:57:41 [1] Eval Extra: inv=0.3861826
17-03-28 04:59:09 [1] Step: 103600 Acc: 0.67781 0.85193 Cost: 1.03459 0.80654 0.13143 0.09662 Time: 0.00084
17-03-28 04:59:09 [1] Train Extra: lr=0.0000152 inv=0.3956250 sub=0.0000000
17-03-28 05:00:33 [1] Step: 103700 Acc: 0.69625 0.85661 Cost: 0.82644 0.59661 0.13317 0.09666 Time: 0.00083
17-03-28 05:00:33 [1] Train Extra: lr=0.0000152 inv=0.3834375 sub=0.0000000
17-03-28 05:01:58 [1] Step: 103800 Acc: 0.66875 0.85109 Cost: 0.98261 0.72377 0.16220 0.09664 Time: 0.00081
17-03-28 05:01:58 [1] Train Extra: lr=0.0000151 inv=0.4387500 sub=0.0000000
17-03-28 05:03:29 [1] Step: 103900 Acc: 0.67906 0.84894 Cost: 0.97263 0.72151 0.15444 0.09668 Time: 0.00083
17-03-28 05:03:29 [1] Train Extra: lr=0.0000151 inv=0.4287500 sub=0.0000000
17-03-28 05:04:48 [1] Step: 104000 Acc: 0.68000 0.85223 Cost: 1.10212 0.73576 0.26972 0.09664 Time: 0.00080
17-03-28 05:04:48 [1] Train Extra: lr=0.0000151 inv=0.3839063 sub=0.0000000
17-03-28 05:05:57 [1] Step: 104000 Eval acc: 0.67147 0.86063 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 05:05:57 [1] Eval Extra: inv=0.3932292
17-03-28 05:07:30 [1] Step: 104100 Acc: 0.68188 0.85209 Cost: 1.12290 0.69675 0.32938 0.09677 Time: 0.00085
17-03-28 05:07:30 [1] Train Extra: lr=0.0000150 inv=0.3848437 sub=0.0000000
17-03-28 05:08:54 [1] Step: 104200 Acc: 0.66812 0.85055 Cost: 1.16496 0.81821 0.24986 0.09688 Time: 0.00081
17-03-28 05:08:54 [1] Train Extra: lr=0.0000150 inv=0.4048438 sub=0.0000000
17-03-28 05:10:17 [1] Step: 104300 Acc: 0.68625 0.84701 Cost: 1.29754 0.94804 0.25266 0.09685 Time: 0.00080
17-03-28 05:10:17 [1] Train Extra: lr=0.0000149 inv=0.3995313 sub=0.0000000
17-03-28 05:11:48 [1] Step: 104400 Acc: 0.68750 0.84330 Cost: 1.05784 0.64723 0.31375 0.09687 Time: 0.00084
17-03-28 05:11:48 [1] Train Extra: lr=0.0000149 inv=0.4189062 sub=0.0000000
17-03-28 05:13:43 [1] Step: 104500 Acc: 0.67281 0.84574 Cost: 1.26144 0.97062 0.19386 0.09696 Time: 0.00097
17-03-28 05:13:43 [1] Train Extra: lr=0.0000148 inv=0.4432813 sub=0.0000000
17-03-28 05:14:50 [1] Step: 104500 Eval acc: 0.67749 0.85607 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 05:14:50 [1] Eval Extra: inv=0.4017565
17-03-28 05:16:19 [1] Step: 104600 Acc: 0.68000 0.85520 Cost: 1.37442 1.00048 0.27706 0.09688 Time: 0.00085
17-03-28 05:16:19 [1] Train Extra: lr=0.0000148 inv=0.3901562 sub=0.0000000
17-03-28 05:17:44 [1] Step: 104700 Acc: 0.67656 0.85531 Cost: 1.17175 0.84271 0.23220 0.09685 Time: 0.00081
17-03-28 05:17:44 [1] Train Extra: lr=0.0000148 inv=0.4057812 sub=0.0000000
17-03-28 05:19:16 [1] Step: 104800 Acc: 0.66719 0.84779 Cost: 1.13780 0.87758 0.16339 0.09684 Time: 0.00085
17-03-28 05:19:16 [1] Train Extra: lr=0.0000147 inv=0.4104687 sub=0.0000000
17-03-28 05:20:50 [1] Step: 104900 Acc: 0.68188 0.84656 Cost: 1.39750 0.97480 0.32582 0.09688 Time: 0.00083
17-03-28 05:20:50 [1] Train Extra: lr=0.0000147 inv=0.4289062 sub=0.0000000
17-03-28 05:22:26 [1] Step: 105000 Acc: 0.69281 0.85003 Cost: 1.11367 0.77366 0.24315 0.09686 Time: 0.00086
17-03-28 05:22:26 [1] Train Extra: lr=0.0000146 inv=0.4420312 sub=0.0000000
17-03-28 05:23:36 [1] Step: 105000 Eval acc: 0.67606 0.85976 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 05:23:36 [1] Eval Extra: inv=0.3853656
17-03-28 05:23:36 [1] Checkpointing.
17-03-28 05:25:06 [1] Step: 105100 Acc: 0.67469 0.85255 Cost: 1.12100 0.72657 0.29749 0.09695 Time: 0.00083
17-03-28 05:25:06 [1] Train Extra: lr=0.0000146 inv=0.4371875 sub=0.0000000
17-03-28 05:26:37 [1] Step: 105200 Acc: 0.68312 0.85203 Cost: 0.95996 0.70869 0.15439 0.09687 Time: 0.00084
17-03-28 05:26:37 [1] Train Extra: lr=0.0000145 inv=0.4395312 sub=0.0000000
17-03-28 05:27:56 [1] Step: 105300 Acc: 0.68312 0.85681 Cost: 0.99879 0.71046 0.19148 0.09685 Time: 0.00079
17-03-28 05:27:56 [1] Train Extra: lr=0.0000145 inv=0.4223438 sub=0.0000000
17-03-28 05:29:20 [1] Step: 105400 Acc: 0.67188 0.85268 Cost: 1.17485 0.86851 0.20948 0.09686 Time: 0.00083
17-03-28 05:29:20 [1] Train Extra: lr=0.0000145 inv=0.4170313 sub=0.0000000
17-03-28 05:30:49 [1] Step: 105500 Acc: 0.69437 0.84554 Cost: 1.00984 0.70619 0.20679 0.09686 Time: 0.00082
17-03-28 05:30:49 [1] Train Extra: lr=0.0000144 inv=0.4509375 sub=0.0000000
17-03-28 05:32:02 [1] Step: 105500 Eval acc: 0.67443 0.85852 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00021
17-03-28 05:32:02 [1] Eval Extra: inv=0.4161560
17-03-28 05:33:37 [1] Step: 105600 Acc: 0.66500 0.85564 Cost: 0.78260 0.61087 0.07485 0.09687 Time: 0.00087
17-03-28 05:33:37 [1] Train Extra: lr=0.0000144 inv=0.3878125 sub=0.0000000
17-03-28 05:35:10 [1] Step: 105700 Acc: 0.68219 0.85114 Cost: 1.18912 0.75583 0.33643 0.09686 Time: 0.00084
17-03-28 05:35:10 [1] Train Extra: lr=0.0000143 inv=0.4190625 sub=0.0000000
17-03-28 05:36:41 [1] Step: 105800 Acc: 0.68312 0.85409 Cost: 0.91840 0.74904 0.07245 0.09690 Time: 0.00087
17-03-28 05:36:41 [1] Train Extra: lr=0.0000143 inv=0.4089063 sub=0.0000000
17-03-28 05:37:57 [1] Step: 105900 Acc: 0.67063 0.85400 Cost: 0.69286 0.50320 0.09270 0.09696 Time: 0.00078
17-03-28 05:37:57 [1] Train Extra: lr=0.0000143 inv=0.3990625 sub=0.0000000
17-03-28 05:39:27 [1] Step: 106000 Acc: 0.68937 0.84512 Cost: 0.67522 0.42974 0.14846 0.09702 Time: 0.00080
17-03-28 05:39:27 [1] Train Extra: lr=0.0000142 inv=0.4620313 sub=0.0000000
17-03-28 05:40:37 [1] Step: 106000 Eval acc: 0.67249 0.85647 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 05:40:37 [1] Eval Extra: inv=0.4086499
17-03-28 05:42:04 [1] Step: 106100 Acc: 0.69031 0.85682 Cost: 1.22691 0.99396 0.13599 0.09695 Time: 0.00082
17-03-28 05:42:04 [1] Train Extra: lr=0.0000142 inv=0.4025000 sub=0.0000000
17-03-28 05:43:34 [1] Step: 106200 Acc: 0.67625 0.85059 Cost: 0.75020 0.54514 0.10805 0.09700 Time: 0.00086
17-03-28 05:43:34 [1] Train Extra: lr=0.0000141 inv=0.3917188 sub=0.0000000
17-03-28 05:45:08 [1] Step: 106300 Acc: 0.67812 0.84910 Cost: 0.91193 0.65736 0.15759 0.09698 Time: 0.00083
17-03-28 05:45:08 [1] Train Extra: lr=0.0000141 inv=0.4462500 sub=0.0000000
17-03-28 05:46:38 [1] Step: 106400 Acc: 0.66625 0.84678 Cost: 1.28908 0.93262 0.25947 0.09698 Time: 0.00082
17-03-28 05:46:38 [1] Train Extra: lr=0.0000141 inv=0.4178125 sub=0.0000000
17-03-28 05:48:02 [1] Step: 106500 Acc: 0.68344 0.85368 Cost: 1.06931 0.82839 0.14393 0.09699 Time: 0.00082
17-03-28 05:48:02 [1] Train Extra: lr=0.0000140 inv=0.3785938 sub=0.0000000
17-03-28 05:49:11 [1] Step: 106500 Eval acc: 0.67463 0.86002 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 05:49:11 [1] Eval Extra: inv=0.3840380
17-03-28 05:50:33 [1] Step: 106600 Acc: 0.68844 0.85193 Cost: 0.96456 0.64374 0.22384 0.09698 Time: 0.00079
17-03-28 05:50:33 [1] Train Extra: lr=0.0000140 inv=0.4042188 sub=0.0000000
17-03-28 05:51:56 [1] Step: 106700 Acc: 0.68969 0.84859 Cost: 1.06492 0.74918 0.21871 0.09702 Time: 0.00081
17-03-28 05:51:56 [1] Train Extra: lr=0.0000139 inv=0.3984375 sub=0.0000000
17-03-28 05:53:32 [1] Step: 106800 Acc: 0.68375 0.85579 Cost: 1.05259 0.71317 0.24232 0.09710 Time: 0.00088
17-03-28 05:53:32 [1] Train Extra: lr=0.0000139 inv=0.4010938 sub=0.0000000
17-03-28 05:55:05 [1] Step: 106900 Acc: 0.70063 0.86032 Cost: 1.12041 0.82522 0.19808 0.09711 Time: 0.00085
17-03-28 05:55:05 [1] Train Extra: lr=0.0000139 inv=0.3976562 sub=0.0000000
17-03-28 05:56:37 [1] Step: 107000 Acc: 0.69250 0.84705 Cost: 0.78768 0.58444 0.10619 0.09705 Time: 0.00082
17-03-28 05:56:37 [1] Train Extra: lr=0.0000138 inv=0.4446875 sub=0.0000000
17-03-28 05:57:42 [1] Step: 107000 Eval acc: 0.67545 0.85605 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 05:57:42 [1] Eval Extra: inv=0.4235600
17-03-28 05:59:13 [1] Step: 107100 Acc: 0.67281 0.85630 Cost: 1.13108 0.78051 0.25351 0.09706 Time: 0.00084
17-03-28 05:59:13 [1] Train Extra: lr=0.0000138 inv=0.3998437 sub=0.0000000
17-03-28 06:00:35 [1] Step: 107200 Acc: 0.68406 0.85071 Cost: 0.90522 0.67083 0.13733 0.09706 Time: 0.00083
17-03-28 06:00:35 [1] Train Extra: lr=0.0000137 inv=0.4132812 sub=0.0000000
17-03-28 06:02:07 [1] Step: 107300 Acc: 0.69156 0.85415 Cost: 1.22204 0.83989 0.28510 0.09706 Time: 0.00086
17-03-28 06:02:07 [1] Train Extra: lr=0.0000137 inv=0.4173438 sub=0.0000000
17-03-28 06:03:31 [1] Step: 107400 Acc: 0.68937 0.84737 Cost: 0.74620 0.51267 0.13648 0.09705 Time: 0.00080
17-03-28 06:03:31 [1] Train Extra: lr=0.0000137 inv=0.4215625 sub=0.0000000
17-03-28 06:05:02 [1] Step: 107500 Acc: 0.69656 0.85226 Cost: 0.74987 0.56458 0.08828 0.09701 Time: 0.00084
17-03-28 06:05:02 [1] Train Extra: lr=0.0000136 inv=0.4118750 sub=0.0000000
17-03-28 06:06:12 [1] Step: 107500 Eval acc: 0.67320 0.85807 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 06:06:12 [1] Eval Extra: inv=0.4009906
17-03-28 06:07:42 [1] Step: 107600 Acc: 0.70625 0.86835 Cost: 1.12453 0.76679 0.26061 0.09713 Time: 0.00089
17-03-28 06:07:42 [1] Train Extra: lr=0.0000136 inv=0.3771875 sub=0.0000000
17-03-28 06:09:13 [1] Step: 107700 Acc: 0.71156 0.85840 Cost: 0.82696 0.50263 0.22700 0.09733 Time: 0.00085
17-03-28 06:09:13 [1] Train Extra: lr=0.0000135 inv=0.4059375 sub=0.0000000
17-03-28 06:10:44 [1] Step: 107800 Acc: 0.70063 0.85225 Cost: 0.94730 0.74908 0.10079 0.09743 Time: 0.00084
17-03-28 06:10:44 [1] Train Extra: lr=0.0000135 inv=0.4126563 sub=0.0000000
17-03-28 06:12:03 [1] Step: 107900 Acc: 0.70594 0.85518 Cost: 0.98916 0.65326 0.23835 0.09755 Time: 0.00079
17-03-28 06:12:03 [1] Train Extra: lr=0.0000135 inv=0.3940625 sub=0.0000000
17-03-28 06:13:36 [1] Step: 108000 Acc: 0.69469 0.84540 Cost: 1.08157 0.77023 0.21368 0.09766 Time: 0.00081
17-03-28 06:13:36 [1] Train Extra: lr=0.0000134 inv=0.4490625 sub=0.0000000
17-03-28 06:14:44 [1] Step: 108000 Eval acc: 0.67228 0.85892 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 06:14:44 [1] Eval Extra: inv=0.4007864
17-03-28 06:16:16 [1] Step: 108100 Acc: 0.69906 0.85213 Cost: 0.84229 0.59258 0.15199 0.09772 Time: 0.00083
17-03-28 06:16:16 [1] Train Extra: lr=0.0000134 inv=0.4350000 sub=0.0000000
17-03-28 06:17:48 [1] Step: 108200 Acc: 0.70719 0.85236 Cost: 1.42468 0.96057 0.36633 0.09778 Time: 0.00085
17-03-28 06:17:48 [1] Train Extra: lr=0.0000133 inv=0.4157812 sub=0.0000000
17-03-28 06:19:22 [1] Step: 108300 Acc: 0.68656 0.84933 Cost: 0.75932 0.51373 0.14766 0.09793 Time: 0.00084
17-03-28 06:19:22 [1] Train Extra: lr=0.0000133 inv=0.4457813 sub=0.0000000
17-03-28 06:20:53 [1] Step: 108400 Acc: 0.70125 0.85179 Cost: 0.82200 0.51286 0.21109 0.09805 Time: 0.00083
17-03-28 06:20:53 [1] Train Extra: lr=0.0000133 inv=0.4098438 sub=0.0000000
17-03-28 06:22:12 [1] Step: 108500 Acc: 0.70594 0.85299 Cost: 0.98604 0.65489 0.23303 0.09812 Time: 0.00079
17-03-28 06:22:12 [1] Train Extra: lr=0.0000132 inv=0.3868750 sub=0.0000000
17-03-28 06:23:21 [1] Step: 108500 Eval acc: 0.67596 0.85794 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 06:23:21 [1] Eval Extra: inv=0.4295343
17-03-28 06:24:43 [1] Step: 108600 Acc: 0.70063 0.85185 Cost: 1.07915 0.83474 0.14619 0.09822 Time: 0.00082
17-03-28 06:24:43 [1] Train Extra: lr=0.0000132 inv=0.3931250 sub=0.0000000
17-03-28 06:26:07 [1] Step: 108700 Acc: 0.68781 0.85183 Cost: 1.12340 0.72895 0.29615 0.09831 Time: 0.00081
17-03-28 06:26:07 [1] Train Extra: lr=0.0000132 inv=0.4242187 sub=0.0000000
17-03-28 06:27:43 [1] Step: 108800 Acc: 0.70188 0.85814 Cost: 1.09070 0.75240 0.23986 0.09843 Time: 0.00084
17-03-28 06:27:43 [1] Train Extra: lr=0.0000131 inv=0.4310937 sub=0.0000000
17-03-28 06:29:23 [1] Step: 108900 Acc: 0.69406 0.85194 Cost: 1.16478 0.82569 0.24057 0.09852 Time: 0.00088
17-03-28 06:29:23 [1] Train Extra: lr=0.0000131 inv=0.4210937 sub=0.0000000
17-03-28 06:31:01 [1] Step: 109000 Acc: 0.69094 0.85196 Cost: 1.31663 0.87686 0.34123 0.09854 Time: 0.00088
17-03-28 06:31:01 [1] Train Extra: lr=0.0000130 inv=0.4367187 sub=0.0000000
17-03-28 06:32:07 [1] Step: 109000 Eval acc: 0.67412 0.85820 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 06:32:07 [1] Eval Extra: inv=0.4664522
17-03-28 06:33:32 [1] Step: 109100 Acc: 0.69563 0.84663 Cost: 0.89032 0.65320 0.13845 0.09867 Time: 0.00081
17-03-28 06:33:32 [1] Train Extra: lr=0.0000130 inv=0.4037500 sub=0.0000000
17-03-28 06:35:12 [1] Step: 109200 Acc: 0.68563 0.85219 Cost: 1.25935 0.77551 0.38506 0.09878 Time: 0.00086
17-03-28 06:35:12 [1] Train Extra: lr=0.0000130 inv=0.4376563 sub=0.0000000
17-03-28 06:36:39 [1] Step: 109300 Acc: 0.70750 0.84845 Cost: 0.82286 0.56604 0.15800 0.09882 Time: 0.00080
17-03-28 06:36:39 [1] Train Extra: lr=0.0000129 inv=0.4153125 sub=0.0000000
17-03-28 06:38:10 [1] Step: 109400 Acc: 0.69125 0.86317 Cost: 1.15167 0.81023 0.24255 0.09889 Time: 0.00087
17-03-28 06:38:10 [1] Train Extra: lr=0.0000129 inv=0.3754688 sub=0.0000000
17-03-28 06:39:34 [1] Step: 109500 Acc: 0.71375 0.86495 Cost: 0.88899 0.55789 0.23210 0.09901 Time: 0.00085
17-03-28 06:39:34 [1] Train Extra: lr=0.0000129 inv=0.3909375 sub=0.0000000
17-03-28 06:40:43 [1] Step: 109500 Eval acc: 0.66840 0.85237 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 06:40:43 [1] Eval Extra: inv=0.4256536
17-03-28 06:42:12 [1] Step: 109600 Acc: 0.69969 0.85913 Cost: 0.95930 0.68540 0.17485 0.09906 Time: 0.00086
17-03-28 06:42:12 [1] Train Extra: lr=0.0000128 inv=0.4150000 sub=0.0000000
17-03-28 06:43:46 [1] Step: 109700 Acc: 0.69281 0.84903 Cost: 0.86883 0.62839 0.14136 0.09908 Time: 0.00084
17-03-28 06:43:46 [1] Train Extra: lr=0.0000128 inv=0.4351563 sub=0.0000000
17-03-28 06:45:14 [1] Step: 109800 Acc: 0.69625 0.84851 Cost: 1.16684 0.82938 0.23833 0.09913 Time: 0.00081
17-03-28 06:45:14 [1] Train Extra: lr=0.0000127 inv=0.4084375 sub=0.0000000
17-03-28 06:46:39 [1] Step: 109900 Acc: 0.68406 0.84688 Cost: 1.19647 0.89700 0.20027 0.09920 Time: 0.00080
17-03-28 06:46:39 [1] Train Extra: lr=0.0000127 inv=0.4089063 sub=0.0000000
17-03-28 06:48:18 [1] Step: 110000 Acc: 0.67219 0.84894 Cost: 1.22912 0.90613 0.22366 0.09933 Time: 0.00084
17-03-28 06:48:18 [1] Train Extra: lr=0.0000127 inv=0.4332813 sub=0.0000000
17-03-28 06:49:27 [1] Step: 110000 Eval acc: 0.67096 0.85972 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 06:49:27 [1] Eval Extra: inv=0.3625919
17-03-28 06:49:27 [1] Checkpointing.
17-03-28 06:50:51 [1] Step: 110100 Acc: 0.69188 0.85631 Cost: 1.53668 1.15220 0.28505 0.09943 Time: 0.00084
17-03-28 06:50:51 [1] Train Extra: lr=0.0000126 inv=0.3978125 sub=0.0000000
17-03-28 06:52:25 [1] Step: 110200 Acc: 0.70406 0.85167 Cost: 0.84348 0.61319 0.13080 0.09949 Time: 0.00082
17-03-28 06:52:25 [1] Train Extra: lr=0.0000126 inv=0.4459375 sub=0.0000000
17-03-28 06:53:57 [1] Step: 110300 Acc: 0.69375 0.85592 Cost: 1.00491 0.67378 0.23159 0.09953 Time: 0.00085
17-03-28 06:53:57 [1] Train Extra: lr=0.0000126 inv=0.4093750 sub=0.0000000
17-03-28 06:55:29 [1] Step: 110400 Acc: 0.68406 0.85034 Cost: 1.05012 0.70638 0.24409 0.09964 Time: 0.00084
17-03-28 06:55:29 [1] Train Extra: lr=0.0000125 inv=0.4167188 sub=0.0000000
17-03-28 06:57:03 [1] Step: 110500 Acc: 0.70063 0.85554 Cost: 0.91509 0.68896 0.12648 0.09964 Time: 0.00087
17-03-28 06:57:03 [1] Train Extra: lr=0.0000125 inv=0.4003125 sub=0.0000000
17-03-28 06:58:12 [1] Step: 110500 Eval acc: 0.67055 0.85689 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 06:58:12 [1] Eval Extra: inv=0.3794935
17-03-28 06:59:38 [1] Step: 110600 Acc: 0.68812 0.85030 Cost: 1.00719 0.78050 0.12701 0.09968 Time: 0.00080
17-03-28 06:59:38 [1] Train Extra: lr=0.0000125 inv=0.4171875 sub=0.0000000
17-03-28 07:01:09 [1] Step: 110700 Acc: 0.69844 0.84970 Cost: 1.15997 0.86705 0.19318 0.09974 Time: 0.00082
17-03-28 07:01:09 [1] Train Extra: lr=0.0000124 inv=0.4446875 sub=0.0000000
17-03-28 07:02:39 [1] Step: 110800 Acc: 0.69937 0.85500 Cost: 1.15689 0.87890 0.17820 0.09979 Time: 0.00084
17-03-28 07:02:39 [1] Train Extra: lr=0.0000124 inv=0.4235937 sub=0.0000000
17-03-28 07:04:12 [1] Step: 110900 Acc: 0.69469 0.85357 Cost: 0.90607 0.63932 0.16690 0.09985 Time: 0.00085
17-03-28 07:04:12 [1] Train Extra: lr=0.0000123 inv=0.3793750 sub=0.0000000
17-03-28 07:05:36 [1] Step: 111000 Acc: 0.69750 0.84898 Cost: 0.80615 0.59257 0.11364 0.09994 Time: 0.00082
17-03-28 07:05:36 [1] Train Extra: lr=0.0000123 inv=0.3764062 sub=0.0000000
17-03-28 07:06:41 [1] Step: 111000 Eval acc: 0.67555 0.85827 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 07:06:41 [1] Eval Extra: inv=0.3794424
17-03-28 07:08:05 [1] Step: 111100 Acc: 0.69250 0.85327 Cost: 1.22436 0.91887 0.20551 0.09998 Time: 0.00081
17-03-28 07:08:05 [1] Train Extra: lr=0.0000123 inv=0.3982812 sub=0.0000000
17-03-28 07:09:36 [1] Step: 111200 Acc: 0.68781 0.85151 Cost: 1.19110 0.82400 0.26715 0.09995 Time: 0.00084
17-03-28 07:09:36 [1] Train Extra: lr=0.0000122 inv=0.4221875 sub=0.0000000
17-03-28 07:10:59 [1] Step: 111300 Acc: 0.68156 0.84891 Cost: 0.95553 0.67584 0.17970 0.09999 Time: 0.00082
17-03-28 07:10:59 [1] Train Extra: lr=0.0000122 inv=0.3970313 sub=0.0000000
17-03-28 07:12:25 [1] Step: 111400 Acc: 0.69281 0.85147 Cost: 1.14696 0.76372 0.28314 0.10010 Time: 0.00081
17-03-28 07:12:25 [1] Train Extra: lr=0.0000122 inv=0.4135937 sub=0.0000000
17-03-28 07:13:55 [1] Step: 111500 Acc: 0.69594 0.85969 Cost: 1.33620 0.86563 0.37039 0.10017 Time: 0.00086
17-03-28 07:13:55 [1] Train Extra: lr=0.0000121 inv=0.4068750 sub=0.0000000
17-03-28 07:15:00 [1] Step: 111500 Eval acc: 0.67188 0.85789 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 07:15:00 [1] Eval Extra: inv=0.4471507
17-03-28 07:16:31 [1] Step: 111600 Acc: 0.68437 0.85692 Cost: 1.02446 0.73208 0.19212 0.10025 Time: 0.00086
17-03-28 07:16:31 [1] Train Extra: lr=0.0000121 inv=0.4112500 sub=0.0000000
17-03-28 07:18:05 [1] Step: 111700 Acc: 0.69469 0.85085 Cost: 1.32099 0.94510 0.27564 0.10025 Time: 0.00083
17-03-28 07:18:05 [1] Train Extra: lr=0.0000121 inv=0.4592188 sub=0.0000000
17-03-28 07:19:32 [1] Step: 111800 Acc: 0.69844 0.84941 Cost: 1.07473 0.72338 0.25108 0.10027 Time: 0.00084
17-03-28 07:19:32 [1] Train Extra: lr=0.0000120 inv=0.4029687 sub=0.0000000
17-03-28 07:21:03 [1] Step: 111900 Acc: 0.68500 0.83976 Cost: 0.92336 0.69945 0.12365 0.10025 Time: 0.00081
17-03-28 07:21:03 [1] Train Extra: lr=0.0000120 inv=0.4334375 sub=0.0000000
17-03-28 07:22:31 [1] Step: 112000 Acc: 0.68906 0.85603 Cost: 0.95468 0.67538 0.17902 0.10028 Time: 0.00081
17-03-28 07:22:31 [1] Train Extra: lr=0.0000120 inv=0.4040625 sub=0.0000000
17-03-28 07:23:42 [1] Step: 112000 Eval acc: 0.67412 0.85779 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 07:23:42 [1] Eval Extra: inv=0.4193730
17-03-28 07:25:12 [1] Step: 112100 Acc: 0.69437 0.84734 Cost: 0.83870 0.64341 0.09490 0.10039 Time: 0.00082
17-03-28 07:25:12 [1] Train Extra: lr=0.0000119 inv=0.4479688 sub=0.0000000
17-03-28 07:26:37 [1] Step: 112200 Acc: 0.69281 0.85178 Cost: 1.07935 0.85401 0.12489 0.10045 Time: 0.00082
17-03-28 07:26:37 [1] Train Extra: lr=0.0000119 inv=0.4100000 sub=0.0000000
17-03-28 07:28:02 [1] Step: 112300 Acc: 0.69312 0.85120 Cost: 0.92297 0.57912 0.24332 0.10053 Time: 0.00081
17-03-28 07:28:02 [1] Train Extra: lr=0.0000119 inv=0.3993750 sub=0.0000000
17-03-28 07:29:34 [1] Step: 112400 Acc: 0.68563 0.84880 Cost: 1.18480 0.84080 0.24341 0.10059 Time: 0.00084
17-03-28 07:29:34 [1] Train Extra: lr=0.0000118 inv=0.4187500 sub=0.0000000
17-03-28 07:31:07 [1] Step: 112500 Acc: 0.67781 0.85131 Cost: 1.02026 0.71641 0.20325 0.10061 Time: 0.00083
17-03-28 07:31:07 [1] Train Extra: lr=0.0000118 inv=0.4165625 sub=0.0000000
17-03-28 07:32:16 [1] Step: 112500 Eval acc: 0.67555 0.86094 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 07:32:16 [1] Eval Extra: inv=0.3925654
17-03-28 07:33:39 [1] Step: 112600 Acc: 0.68125 0.85260 Cost: 0.95344 0.65148 0.20131 0.10065 Time: 0.00081
17-03-28 07:33:39 [1] Train Extra: lr=0.0000118 inv=0.4095313 sub=0.0000000
17-03-28 07:35:03 [1] Step: 112700 Acc: 0.67250 0.85120 Cost: 1.15537 0.76907 0.28565 0.10066 Time: 0.00081
17-03-28 07:35:03 [1] Train Extra: lr=0.0000117 inv=0.4060937 sub=0.0000000
17-03-28 07:36:27 [1] Step: 112800 Acc: 0.68531 0.85338 Cost: 1.01397 0.77367 0.13965 0.10065 Time: 0.00082
17-03-28 07:36:27 [1] Train Extra: lr=0.0000117 inv=0.4139062 sub=0.0000000
17-03-28 07:38:00 [1] Step: 112900 Acc: 0.69125 0.85098 Cost: 1.06753 0.71491 0.25190 0.10072 Time: 0.00082
17-03-28 07:38:00 [1] Train Extra: lr=0.0000117 inv=0.4485938 sub=0.0000000
17-03-28 07:39:26 [1] Step: 113000 Acc: 0.69656 0.84929 Cost: 1.17209 0.86359 0.20764 0.10086 Time: 0.00080
17-03-28 07:39:26 [1] Train Extra: lr=0.0000116 inv=0.4253125 sub=0.0000000
17-03-28 07:40:35 [1] Step: 113000 Eval acc: 0.66973 0.85801 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 07:40:35 [1] Eval Extra: inv=0.4154412
17-03-28 07:41:53 [1] Step: 113100 Acc: 0.69125 0.85680 Cost: 0.90030 0.62498 0.17444 0.10088 Time: 0.00081
17-03-28 07:41:53 [1] Train Extra: lr=0.0000116 inv=0.3507812 sub=0.0000000
17-03-28 07:43:25 [1] Step: 113200 Acc: 0.69125 0.84862 Cost: 0.89663 0.65705 0.13866 0.10093 Time: 0.00083
17-03-28 07:43:25 [1] Train Extra: lr=0.0000116 inv=0.4090625 sub=0.0000000
17-03-28 07:44:54 [1] Step: 113300 Acc: 0.67406 0.85084 Cost: 0.84525 0.62851 0.11582 0.10092 Time: 0.00083
17-03-28 07:44:54 [1] Train Extra: lr=0.0000115 inv=0.4118750 sub=0.0000000
17-03-28 07:46:16 [1] Step: 113400 Acc: 0.70500 0.85402 Cost: 0.95749 0.69997 0.15661 0.10090 Time: 0.00082
17-03-28 07:46:16 [1] Train Extra: lr=0.0000115 inv=0.4131250 sub=0.0000000
17-03-28 07:47:40 [1] Step: 113500 Acc: 0.68563 0.85580 Cost: 1.07278 0.69904 0.27279 0.10094 Time: 0.00083
17-03-28 07:47:40 [1] Train Extra: lr=0.0000115 inv=0.4001562 sub=0.0000000
17-03-28 07:48:49 [1] Step: 113500 Eval acc: 0.67320 0.85874 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 07:48:49 [1] Eval Extra: inv=0.4171773
17-03-28 07:50:29 [1] Step: 113600 Acc: 0.68812 0.85881 Cost: 0.93548 0.68748 0.14708 0.10092 Time: 0.00086
17-03-28 07:50:29 [1] Train Extra: lr=0.0000114 inv=0.4250000 sub=0.0000000
17-03-28 07:52:00 [1] Step: 113700 Acc: 0.70063 0.85232 Cost: 1.00966 0.63369 0.27500 0.10097 Time: 0.00084
17-03-28 07:52:00 [1] Train Extra: lr=0.0000114 inv=0.4262500 sub=0.0000000
17-03-28 07:53:33 [1] Step: 113800 Acc: 0.68625 0.85728 Cost: 0.79654 0.56686 0.12870 0.10098 Time: 0.00085
17-03-28 07:53:33 [1] Train Extra: lr=0.0000114 inv=0.4273438 sub=0.0000000
17-03-28 07:54:57 [1] Step: 113900 Acc: 0.69281 0.84521 Cost: 1.12964 0.84071 0.18788 0.10105 Time: 0.00081
17-03-28 07:54:57 [1] Train Extra: lr=0.0000113 inv=0.4082812 sub=0.0000000
17-03-28 07:56:20 [1] Step: 114000 Acc: 0.70188 0.85647 Cost: 0.97290 0.71685 0.15494 0.10111 Time: 0.00083
17-03-28 07:56:20 [1] Train Extra: lr=0.0000113 inv=0.3740625 sub=0.0000000
17-03-28 07:57:25 [1] Step: 114000 Eval acc: 0.67024 0.85571 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 07:57:25 [1] Eval Extra: inv=0.3949142
17-03-28 07:58:57 [1] Step: 114100 Acc: 0.68656 0.84936 Cost: 1.27966 0.95174 0.22676 0.10116 Time: 0.00082
17-03-28 07:58:57 [1] Train Extra: lr=0.0000113 inv=0.4328125 sub=0.0000000
17-03-28 08:00:35 [1] Step: 114200 Acc: 0.69219 0.85603 Cost: 0.97370 0.66072 0.21185 0.10113 Time: 0.00088
17-03-28 08:00:35 [1] Train Extra: lr=0.0000112 inv=0.4212500 sub=0.0000000
17-03-28 08:02:14 [1] Step: 114300 Acc: 0.68437 0.86122 Cost: 1.16381 0.78845 0.27420 0.10116 Time: 0.00090
17-03-28 08:02:14 [1] Train Extra: lr=0.0000112 inv=0.3968750 sub=0.0000000
17-03-28 08:03:38 [1] Step: 114400 Acc: 0.70531 0.85083 Cost: 1.04759 0.77581 0.17062 0.10115 Time: 0.00083
17-03-28 08:03:38 [1] Train Extra: lr=0.0000112 inv=0.3887500 sub=0.0000000
17-03-28 08:05:04 [1] Step: 114500 Acc: 0.69500 0.85615 Cost: 1.04842 0.65127 0.29600 0.10116 Time: 0.00083
17-03-28 08:05:04 [1] Train Extra: lr=0.0000111 inv=0.4018750 sub=0.0000000
17-03-28 08:06:11 [1] Step: 114500 Eval acc: 0.67157 0.86117 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 08:06:11 [1] Eval Extra: inv=0.4120711
17-03-28 08:07:35 [1] Step: 114600 Acc: 0.69063 0.85081 Cost: 1.17353 0.79996 0.27242 0.10115 Time: 0.00082
17-03-28 08:07:35 [1] Train Extra: lr=0.0000111 inv=0.4153125 sub=0.0000000
17-03-28 08:09:07 [1] Step: 114700 Acc: 0.68312 0.85385 Cost: 0.76248 0.56936 0.09187 0.10125 Time: 0.00085
17-03-28 08:09:07 [1] Train Extra: lr=0.0000111 inv=0.4287500 sub=0.0000000
17-03-28 08:10:36 [1] Step: 114800 Acc: 0.70063 0.84862 Cost: 1.12854 0.79396 0.23329 0.10129 Time: 0.00082
17-03-28 08:10:36 [1] Train Extra: lr=0.0000110 inv=0.4351563 sub=0.0000000
17-03-28 08:12:07 [1] Step: 114900 Acc: 0.68812 0.85553 Cost: 0.88767 0.64748 0.13888 0.10131 Time: 0.00085
17-03-28 08:12:07 [1] Train Extra: lr=0.0000110 inv=0.4026562 sub=0.0000000
17-03-28 08:13:35 [1] Step: 115000 Acc: 0.69937 0.85177 Cost: 1.22348 0.89606 0.22610 0.10132 Time: 0.00081
17-03-28 08:13:35 [1] Train Extra: lr=0.0000110 inv=0.4289062 sub=0.0000000
17-03-28 08:14:46 [1] Step: 115000 Eval acc: 0.67188 0.85605 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 08:14:46 [1] Eval Extra: inv=0.4406658
17-03-28 08:14:46 [1] Checkpointing.
17-03-28 08:16:14 [1] Step: 115100 Acc: 0.68063 0.85587 Cost: 1.10332 0.77093 0.23106 0.10132 Time: 0.00082
17-03-28 08:16:14 [1] Train Extra: lr=0.0000109 inv=0.4337500 sub=0.0000000
17-03-28 08:17:36 [1] Step: 115200 Acc: 0.69000 0.85225 Cost: 1.09522 0.74673 0.24719 0.10130 Time: 0.00083
17-03-28 08:17:36 [1] Train Extra: lr=0.0000109 inv=0.4050000 sub=0.0000000
17-03-28 08:19:09 [1] Step: 115300 Acc: 0.67844 0.84916 Cost: 1.03935 0.71933 0.21872 0.10130 Time: 0.00083
17-03-28 08:19:09 [1] Train Extra: lr=0.0000109 inv=0.4317187 sub=0.0000000
17-03-28 08:20:35 [1] Step: 115400 Acc: 0.68594 0.85404 Cost: 0.79145 0.57146 0.11864 0.10135 Time: 0.00082
17-03-28 08:20:35 [1] Train Extra: lr=0.0000108 inv=0.4104687 sub=0.0000000
17-03-28 08:21:57 [1] Step: 115500 Acc: 0.69000 0.85590 Cost: 1.04632 0.66326 0.28167 0.10139 Time: 0.00084
17-03-28 08:21:57 [1] Train Extra: lr=0.0000108 inv=0.3757813 sub=0.0000000
17-03-28 08:23:03 [1] Step: 115500 Eval acc: 0.67300 0.85935 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 08:23:03 [1] Eval Extra: inv=0.4218750
17-03-28 08:24:41 [1] Step: 115600 Acc: 0.68312 0.85260 Cost: 1.03519 0.71771 0.21605 0.10143 Time: 0.00086
17-03-28 08:24:41 [1] Train Extra: lr=0.0000108 inv=0.4310937 sub=0.0000000
17-03-28 08:26:05 [1] Step: 115700 Acc: 0.68656 0.85703 Cost: 1.20439 0.84018 0.26278 0.10143 Time: 0.00083
17-03-28 08:26:05 [1] Train Extra: lr=0.0000108 inv=0.4078125 sub=0.0000000
17-03-28 08:27:35 [1] Step: 115800 Acc: 0.67344 0.86056 Cost: 1.15703 0.83812 0.21752 0.10138 Time: 0.00087
17-03-28 08:27:35 [1] Train Extra: lr=0.0000107 inv=0.3971875 sub=0.0000000
17-03-28 08:29:03 [1] Step: 115900 Acc: 0.69031 0.85037 Cost: 1.15358 0.81676 0.23544 0.10137 Time: 0.00082
17-03-28 08:29:03 [1] Train Extra: lr=0.0000107 inv=0.4173438 sub=0.0000000
17-03-28 08:30:42 [1] Step: 116000 Acc: 0.68625 0.86529 Cost: 1.17019 0.75052 0.31832 0.10135 Time: 0.00091
17-03-28 08:30:42 [1] Train Extra: lr=0.0000107 inv=0.4001562 sub=0.0000000
17-03-28 08:31:48 [1] Step: 116000 Eval acc: 0.67004 0.85965 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 08:31:48 [1] Eval Extra: inv=0.4411254
17-03-28 08:33:12 [1] Step: 116100 Acc: 0.69094 0.84990 Cost: 0.93590 0.62804 0.20640 0.10146 Time: 0.00082
17-03-28 08:33:12 [1] Train Extra: lr=0.0000106 inv=0.4273438 sub=0.0000000
17-03-28 08:34:36 [1] Step: 116200 Acc: 0.69969 0.85216 Cost: 0.99199 0.60056 0.28993 0.10150 Time: 0.00083
17-03-28 08:34:36 [1] Train Extra: lr=0.0000106 inv=0.4037500 sub=0.0000000
17-03-28 08:36:06 [1] Step: 116300 Acc: 0.67656 0.85029 Cost: 0.84001 0.59159 0.14688 0.10153 Time: 0.00081
17-03-28 08:36:06 [1] Train Extra: lr=0.0000106 inv=0.4442187 sub=0.0000000
17-03-28 08:37:36 [1] Step: 116400 Acc: 0.68563 0.85938 Cost: 1.05663 0.76757 0.18751 0.10155 Time: 0.00086
17-03-28 08:37:36 [1] Train Extra: lr=0.0000105 inv=0.4131250 sub=0.0000000
17-03-28 08:38:55 [1] Step: 116500 Acc: 0.69219 0.85593 Cost: 0.97207 0.69363 0.17685 0.10158 Time: 0.00079
17-03-28 08:38:55 [1] Train Extra: lr=0.0000105 inv=0.3896875 sub=0.0000000
17-03-28 08:40:05 [1] Step: 116500 Eval acc: 0.67596 0.85948 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 08:40:05 [1] Eval Extra: inv=0.3721405
17-03-28 08:41:34 [1] Step: 116600 Acc: 0.69625 0.84885 Cost: 0.86898 0.60366 0.16378 0.10155 Time: 0.00082
17-03-28 08:41:34 [1] Train Extra: lr=0.0000105 inv=0.4098438 sub=0.0000000
17-03-28 08:42:59 [1] Step: 116700 Acc: 0.68156 0.85685 Cost: 0.90179 0.57824 0.22194 0.10161 Time: 0.00084
17-03-28 08:42:59 [1] Train Extra: lr=0.0000104 inv=0.3821875 sub=0.0000000
17-03-28 08:44:34 [1] Step: 116800 Acc: 0.66938 0.85911 Cost: 1.29105 0.84882 0.34053 0.10169 Time: 0.00090
17-03-28 08:44:34 [1] Train Extra: lr=0.0000104 inv=0.3921875 sub=0.0000000
17-03-28 08:46:01 [1] Step: 116900 Acc: 0.67688 0.84368 Cost: 0.78966 0.56572 0.12225 0.10168 Time: 0.00079
17-03-28 08:46:01 [1] Train Extra: lr=0.0000104 inv=0.4265625 sub=0.0000000
17-03-28 08:47:26 [1] Step: 117000 Acc: 0.69219 0.85069 Cost: 1.31666 0.95968 0.25530 0.10168 Time: 0.00081
17-03-28 08:47:26 [1] Train Extra: lr=0.0000104 inv=0.4115625 sub=0.0000000
17-03-28 08:48:35 [1] Step: 117000 Eval acc: 0.67913 0.85790 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 08:48:35 [1] Eval Extra: inv=0.4400020
17-03-28 08:50:00 [1] Step: 117100 Acc: 0.68875 0.85640 Cost: 1.22127 0.85837 0.26115 0.10175 Time: 0.00084
17-03-28 08:50:00 [1] Train Extra: lr=0.0000103 inv=0.3767187 sub=0.0000000
17-03-28 08:51:35 [1] Step: 117200 Acc: 0.68156 0.84845 Cost: 1.14649 0.73149 0.31327 0.10174 Time: 0.00083
17-03-28 08:51:35 [1] Train Extra: lr=0.0000103 inv=0.4420312 sub=0.0000000
17-03-28 08:53:00 [1] Step: 117300 Acc: 0.70625 0.85191 Cost: 1.03915 0.77237 0.16498 0.10180 Time: 0.00085
17-03-28 08:53:00 [1] Train Extra: lr=0.0000103 inv=0.3937500 sub=0.0000000
17-03-28 08:54:32 [1] Step: 117400 Acc: 0.71406 0.84908 Cost: 0.77951 0.50389 0.17374 0.10188 Time: 0.00081
17-03-28 08:54:32 [1] Train Extra: lr=0.0000102 inv=0.4654687 sub=0.0000000
17-03-28 08:56:05 [1] Step: 117500 Acc: 0.72500 0.85009 Cost: 1.05687 0.66539 0.28951 0.10197 Time: 0.00083
17-03-28 08:56:05 [1] Train Extra: lr=0.0000102 inv=0.4151563 sub=0.0000000
17-03-28 08:57:14 [1] Step: 117500 Eval acc: 0.66912 0.85751 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 08:57:14 [1] Eval Extra: inv=0.4334150
17-03-28 08:58:45 [1] Step: 117600 Acc: 0.71250 0.85303 Cost: 1.09292 0.64655 0.34432 0.10205 Time: 0.00084
17-03-28 08:58:45 [1] Train Extra: lr=0.0000102 inv=0.4035937 sub=0.0000000
17-03-28 09:00:10 [1] Step: 117700 Acc: 0.72062 0.85696 Cost: 1.10415 0.77680 0.22516 0.10219 Time: 0.00081
17-03-28 09:00:10 [1] Train Extra: lr=0.0000102 inv=0.3981250 sub=0.0000000
17-03-28 09:01:33 [1] Step: 117800 Acc: 0.70937 0.85735 Cost: 0.90500 0.59553 0.20717 0.10229 Time: 0.00082
17-03-28 09:01:33 [1] Train Extra: lr=0.0000101 inv=0.4006250 sub=0.0000000
17-03-28 09:03:04 [1] Step: 117900 Acc: 0.70063 0.84636 Cost: 0.70352 0.49573 0.10539 0.10240 Time: 0.00081
17-03-28 09:03:04 [1] Train Extra: lr=0.0000101 inv=0.4368750 sub=0.0000000
17-03-28 09:04:36 [1] Step: 118000 Acc: 0.69719 0.86141 Cost: 1.76300 1.29781 0.36267 0.10252 Time: 0.00085
17-03-28 09:04:36 [1] Train Extra: lr=0.0000101 inv=0.4175000 sub=0.0000000
17-03-28 09:05:41 [1] Step: 118000 Eval acc: 0.67065 0.86212 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-28 09:05:41 [1] Eval Extra: inv=0.3875102
17-03-28 09:07:13 [1] Step: 118100 Acc: 0.69500 0.85217 Cost: 0.85417 0.60145 0.15003 0.10269 Time: 0.00084
17-03-28 09:07:13 [1] Train Extra: lr=0.0000100 inv=0.4279688 sub=0.0000000
17-03-28 09:08:43 [1] Step: 118200 Acc: 0.69000 0.85372 Cost: 0.81416 0.58135 0.13003 0.10278 Time: 0.00086
17-03-28 09:08:43 [1] Train Extra: lr=0.0000100 inv=0.4039063 sub=0.0000000
17-03-28 09:10:10 [1] Step: 118300 Acc: 0.71375 0.85680 Cost: 1.16029 0.91750 0.13991 0.10287 Time: 0.00084
17-03-28 09:10:10 [1] Train Extra: lr=0.0000100 inv=0.4020313 sub=0.0000000
17-03-28 09:11:42 [1] Step: 118400 Acc: 0.69844 0.85055 Cost: 0.97613 0.63983 0.23329 0.10301 Time: 0.00082
17-03-28 09:11:42 [1] Train Extra: lr=0.0000100 inv=0.4331250 sub=0.0000000
17-03-28 09:13:08 [1] Step: 118500 Acc: 0.71750 0.85830 Cost: 1.13863 0.78875 0.24678 0.10310 Time: 0.00083
17-03-28 09:13:08 [1] Train Extra: lr=0.0000099 inv=0.3987500 sub=0.0000000
17-03-28 09:14:16 [1] Step: 118500 Eval acc: 0.67719 0.85592 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 09:14:16 [1] Eval Extra: inv=0.4381127
17-03-28 09:15:40 [1] Step: 118600 Acc: 0.69656 0.84691 Cost: 0.95162 0.67717 0.17126 0.10319 Time: 0.00081
17-03-28 09:15:40 [1] Train Extra: lr=0.0000099 inv=0.4043750 sub=0.0000000
17-03-28 09:17:07 [1] Step: 118700 Acc: 0.71594 0.84875 Cost: 0.86425 0.55374 0.20725 0.10327 Time: 0.00081
17-03-28 09:17:07 [1] Train Extra: lr=0.0000099 inv=0.4250000 sub=0.0000000
17-03-28 09:18:38 [1] Step: 118800 Acc: 0.70688 0.84848 Cost: 0.94774 0.61441 0.23001 0.10333 Time: 0.00083
17-03-28 09:18:38 [1] Train Extra: lr=0.0000098 inv=0.4142188 sub=0.0000000
17-03-28 09:20:03 [1] Step: 118900 Acc: 0.71437 0.85783 Cost: 0.79728 0.56159 0.13228 0.10341 Time: 0.00083
17-03-28 09:20:03 [1] Train Extra: lr=0.0000098 inv=0.3920312 sub=0.0000000
17-03-28 09:21:34 [1] Step: 119000 Acc: 0.69656 0.85436 Cost: 1.14227 0.80952 0.22930 0.10345 Time: 0.00085
17-03-28 09:21:34 [1] Train Extra: lr=0.0000098 inv=0.3962500 sub=0.0000000
17-03-28 09:22:44 [1] Step: 119000 Eval acc: 0.66983 0.85608 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 09:22:44 [1] Eval Extra: inv=0.4020629
17-03-28 09:24:24 [1] Step: 119100 Acc: 0.71437 0.85951 Cost: 1.26926 0.88502 0.28073 0.10350 Time: 0.00089
17-03-28 09:24:24 [1] Train Extra: lr=0.0000098 inv=0.4096875 sub=0.0000000
17-03-28 09:25:49 [1] Step: 119200 Acc: 0.70437 0.84701 Cost: 0.98886 0.65697 0.22831 0.10357 Time: 0.00081
17-03-28 09:25:49 [1] Train Extra: lr=0.0000097 inv=0.4226563 sub=0.0000000
17-03-28 09:27:24 [1] Step: 119300 Acc: 0.69875 0.84638 Cost: 0.99246 0.67166 0.21720 0.10359 Time: 0.00083
17-03-28 09:27:24 [1] Train Extra: lr=0.0000097 inv=0.4487500 sub=0.0000000
17-03-28 09:29:02 [1] Step: 119400 Acc: 0.72562 0.85612 Cost: 0.89635 0.62896 0.16372 0.10367 Time: 0.00090
17-03-28 09:29:02 [1] Train Extra: lr=0.0000097 inv=0.4020313 sub=0.0000000
17-03-28 09:30:27 [1] Step: 119500 Acc: 0.69406 0.85560 Cost: 0.89224 0.67284 0.11568 0.10373 Time: 0.00082
17-03-28 09:30:27 [1] Train Extra: lr=0.0000096 inv=0.4275000 sub=0.0000000
17-03-28 09:31:37 [1] Step: 119500 Eval acc: 0.67422 0.86060 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 09:31:37 [1] Eval Extra: inv=0.4054330
17-03-28 09:33:15 [1] Step: 119600 Acc: 0.67844 0.85215 Cost: 1.23603 0.83413 0.29808 0.10383 Time: 0.00085
17-03-28 09:33:15 [1] Train Extra: lr=0.0000096 inv=0.4334375 sub=0.0000000
17-03-28 09:34:40 [1] Step: 119700 Acc: 0.70531 0.85227 Cost: 0.87481 0.55267 0.21825 0.10389 Time: 0.00084
17-03-28 09:34:40 [1] Train Extra: lr=0.0000096 inv=0.3956250 sub=0.0000000
17-03-28 09:36:12 [1] Step: 119800 Acc: 0.70344 0.85606 Cost: 0.92628 0.64901 0.17331 0.10396 Time: 0.00086
17-03-28 09:36:12 [1] Train Extra: lr=0.0000096 inv=0.4121875 sub=0.0000000
17-03-28 09:37:52 [1] Step: 119900 Acc: 0.70312 0.84864 Cost: 0.88741 0.61931 0.16410 0.10401 Time: 0.00087
17-03-28 09:37:52 [1] Train Extra: lr=0.0000095 inv=0.3943750 sub=0.0000000
17-03-28 09:39:18 [1] Step: 120000 Acc: 0.68531 0.85141 Cost: 1.31158 0.96353 0.24389 0.10416 Time: 0.00081
17-03-28 09:39:18 [1] Train Extra: lr=0.0000095 inv=0.4440625 sub=0.0000000
17-03-28 09:40:27 [1] Step: 120000 Eval acc: 0.67821 0.85661 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 09:40:27 [1] Eval Extra: inv=0.4177900
17-03-28 09:40:27 [1] Checkpointing.
17-03-28 09:41:49 [1] Step: 120100 Acc: 0.69531 0.85001 Cost: 0.83886 0.53234 0.20231 0.10421 Time: 0.00082
17-03-28 09:41:49 [1] Train Extra: lr=0.0000095 inv=0.4128125 sub=0.0000000
17-03-28 09:43:17 [1] Step: 120200 Acc: 0.69594 0.85257 Cost: 1.24360 0.88614 0.25320 0.10425 Time: 0.00083
17-03-28 09:43:17 [1] Train Extra: lr=0.0000094 inv=0.3823437 sub=0.0000000
17-03-28 09:44:40 [1] Step: 120300 Acc: 0.69875 0.85100 Cost: 1.21821 0.81787 0.29613 0.10420 Time: 0.00082
17-03-28 09:44:40 [1] Train Extra: lr=0.0000094 inv=0.4160937 sub=0.0000000
17-03-28 09:46:11 [1] Step: 120400 Acc: 0.69469 0.85097 Cost: 1.23386 0.85704 0.27257 0.10425 Time: 0.00084
17-03-28 09:46:11 [1] Train Extra: lr=0.0000094 inv=0.4120313 sub=0.0000000
17-03-28 09:47:38 [1] Step: 120500 Acc: 0.69594 0.85518 Cost: 0.99558 0.71008 0.18115 0.10435 Time: 0.00082
17-03-28 09:47:38 [1] Train Extra: lr=0.0000094 inv=0.3926562 sub=0.0000000
17-03-28 09:48:47 [1] Step: 120500 Eval acc: 0.67167 0.85754 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 09:48:47 [1] Eval Extra: inv=0.4106413
17-03-28 09:50:11 [1] Step: 120600 Acc: 0.68594 0.85110 Cost: 1.15606 0.78149 0.27013 0.10444 Time: 0.00082
17-03-28 09:50:11 [1] Train Extra: lr=0.0000093 inv=0.4014063 sub=0.0000000
17-03-28 09:51:35 [1] Step: 120700 Acc: 0.69094 0.84800 Cost: 1.30429 0.88206 0.31771 0.10452 Time: 0.00082
17-03-28 09:51:35 [1] Train Extra: lr=0.0000093 inv=0.3975000 sub=0.0000000
17-03-28 09:53:06 [1] Step: 120800 Acc: 0.69844 0.85335 Cost: 0.88254 0.62165 0.15637 0.10451 Time: 0.00082
17-03-28 09:53:06 [1] Train Extra: lr=0.0000093 inv=0.4350000 sub=0.0000000
17-03-28 09:54:25 [1] Step: 120900 Acc: 0.69437 0.85138 Cost: 0.68253 0.48295 0.09501 0.10457 Time: 0.00078
17-03-28 09:54:25 [1] Train Extra: lr=0.0000093 inv=0.3814063 sub=0.0000000
17-03-28 09:56:18 [1] Step: 121000 Acc: 0.70250 0.85602 Cost: 0.83037 0.60818 0.11757 0.10462 Time: 0.00092
17-03-28 09:56:18 [1] Train Extra: lr=0.0000092 inv=0.4214062 sub=0.0000000
17-03-28 09:57:28 [1] Step: 121000 Eval acc: 0.67688 0.85552 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 09:57:28 [1] Eval Extra: inv=0.4083435
17-03-28 09:58:48 [1] Step: 121100 Acc: 0.70031 0.85330 Cost: 0.85894 0.63218 0.12206 0.10470 Time: 0.00078
17-03-28 09:58:48 [1] Train Extra: lr=0.0000092 inv=0.4070313 sub=0.0000000
17-03-28 10:00:09 [1] Step: 121200 Acc: 0.70594 0.85588 Cost: 1.10934 0.83184 0.17278 0.10472 Time: 0.00084
17-03-28 10:00:09 [1] Train Extra: lr=0.0000092 inv=0.3803125 sub=0.0000000
17-03-28 10:01:50 [1] Step: 121300 Acc: 0.70437 0.84624 Cost: 0.90271 0.67910 0.11887 0.10474 Time: 0.00083
17-03-28 10:01:50 [1] Train Extra: lr=0.0000092 inv=0.4681250 sub=0.0000000
17-03-28 10:03:23 [1] Step: 121400 Acc: 0.69156 0.85364 Cost: 1.18424 0.96877 0.11068 0.10480 Time: 0.00084
17-03-28 10:03:23 [1] Train Extra: lr=0.0000091 inv=0.4068750 sub=0.0000000
17-03-28 10:04:51 [1] Step: 121500 Acc: 0.70312 0.85348 Cost: 1.25890 0.84899 0.30504 0.10487 Time: 0.00084
17-03-28 10:04:51 [1] Train Extra: lr=0.0000091 inv=0.3945312 sub=0.0000000
17-03-28 10:05:56 [1] Step: 121500 Eval acc: 0.67208 0.85774 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-28 10:05:56 [1] Eval Extra: inv=0.3959355
17-03-28 10:07:19 [1] Step: 121600 Acc: 0.70562 0.85152 Cost: 0.90245 0.61984 0.17768 0.10492 Time: 0.00082
17-03-28 10:07:19 [1] Train Extra: lr=0.0000091 inv=0.4043750 sub=0.0000000
17-03-28 10:08:43 [1] Step: 121700 Acc: 0.71094 0.84903 Cost: 1.03401 0.77469 0.15435 0.10496 Time: 0.00082
17-03-28 10:08:43 [1] Train Extra: lr=0.0000090 inv=0.4076562 sub=0.0000000
17-03-28 10:10:08 [1] Step: 121800 Acc: 0.70156 0.86259 Cost: 1.21536 0.90098 0.20936 0.10501 Time: 0.00084
17-03-28 10:10:08 [1] Train Extra: lr=0.0000090 inv=0.3932813 sub=0.0000000
17-03-28 10:11:46 [1] Step: 121900 Acc: 0.70594 0.85596 Cost: 1.08485 0.67118 0.30863 0.10504 Time: 0.00087
17-03-28 10:11:46 [1] Train Extra: lr=0.0000090 inv=0.4407813 sub=0.0000000
17-03-28 10:13:14 [1] Step: 122000 Acc: 0.70156 0.85715 Cost: 1.02669 0.69089 0.23065 0.10515 Time: 0.00086
17-03-28 10:13:14 [1] Train Extra: lr=0.0000090 inv=0.3665625 sub=0.0000000
17-03-28 10:14:24 [1] Step: 122000 Eval acc: 0.67177 0.85687 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 10:14:24 [1] Eval Extra: inv=0.3815359
17-03-28 10:15:57 [1] Step: 122100 Acc: 0.69250 0.85142 Cost: 0.88587 0.62740 0.15330 0.10517 Time: 0.00083
17-03-28 10:15:57 [1] Train Extra: lr=0.0000089 inv=0.4223438 sub=0.0000000
17-03-28 10:17:23 [1] Step: 122200 Acc: 0.69906 0.84524 Cost: 0.89824 0.70974 0.08327 0.10522 Time: 0.00079
17-03-28 10:17:23 [1] Train Extra: lr=0.0000089 inv=0.4456250 sub=0.0000000
17-03-28 10:18:47 [1] Step: 122300 Acc: 0.68469 0.85754 Cost: 0.85245 0.64429 0.10291 0.10525 Time: 0.00082
17-03-28 10:18:47 [1] Train Extra: lr=0.0000089 inv=0.4187500 sub=0.0000000
17-03-28 10:20:25 [1] Step: 122400 Acc: 0.68031 0.85401 Cost: 0.98929 0.73646 0.14751 0.10532 Time: 0.00086
17-03-28 10:20:25 [1] Train Extra: lr=0.0000089 inv=0.4509375 sub=0.0000000
17-03-28 10:21:51 [1] Step: 122500 Acc: 0.69969 0.85402 Cost: 0.96476 0.72687 0.13248 0.10541 Time: 0.00082
17-03-28 10:21:51 [1] Train Extra: lr=0.0000088 inv=0.3976562 sub=0.0000000
17-03-28 10:23:01 [1] Step: 122500 Eval acc: 0.66973 0.86126 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 10:23:01 [1] Eval Extra: inv=0.4046160
17-03-28 10:24:24 [1] Step: 122600 Acc: 0.67750 0.85835 Cost: 1.11171 0.85196 0.15428 0.10547 Time: 0.00083
17-03-28 10:24:24 [1] Train Extra: lr=0.0000088 inv=0.3862500 sub=0.0000000
17-03-28 10:25:42 [1] Step: 122700 Acc: 0.69937 0.85272 Cost: 0.90847 0.68642 0.11656 0.10548 Time: 0.00079
17-03-28 10:25:42 [1] Train Extra: lr=0.0000088 inv=0.4076562 sub=0.0000000
17-03-28 10:27:15 [1] Step: 122800 Acc: 0.69688 0.84686 Cost: 0.85038 0.63410 0.11073 0.10554 Time: 0.00081
17-03-28 10:27:15 [1] Train Extra: lr=0.0000088 inv=0.4437500 sub=0.0000000
17-03-28 10:28:43 [1] Step: 122900 Acc: 0.70500 0.85196 Cost: 0.81442 0.60525 0.10359 0.10558 Time: 0.00085
17-03-28 10:28:43 [1] Train Extra: lr=0.0000087 inv=0.4112500 sub=0.0000000
17-03-28 10:30:02 [1] Step: 123000 Acc: 0.69750 0.85263 Cost: 1.21508 0.95150 0.15793 0.10565 Time: 0.00079
17-03-28 10:30:02 [1] Train Extra: lr=0.0000087 inv=0.3975000 sub=0.0000000
17-03-28 10:31:11 [1] Step: 123000 Eval acc: 0.67688 0.85212 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 10:31:11 [1] Eval Extra: inv=0.4126838
17-03-28 10:32:28 [1] Step: 123100 Acc: 0.70719 0.85985 Cost: 1.00103 0.74384 0.15151 0.10568 Time: 0.00082
17-03-28 10:32:28 [1] Train Extra: lr=0.0000087 inv=0.3675000 sub=0.0000000
17-03-28 10:34:01 [1] Step: 123200 Acc: 0.70562 0.85397 Cost: 1.21020 0.88053 0.22396 0.10570 Time: 0.00084
17-03-28 10:34:01 [1] Train Extra: lr=0.0000087 inv=0.4343750 sub=0.0000000
17-03-28 10:35:34 [1] Step: 123300 Acc: 0.69844 0.84904 Cost: 0.87595 0.61318 0.15707 0.10569 Time: 0.00082
17-03-28 10:35:34 [1] Train Extra: lr=0.0000086 inv=0.4471875 sub=0.0000000
17-03-28 10:37:11 [1] Step: 123400 Acc: 0.69531 0.85664 Cost: 0.74946 0.54670 0.09706 0.10571 Time: 0.00087
17-03-28 10:37:11 [1] Train Extra: lr=0.0000086 inv=0.4478125 sub=0.0000000
17-03-28 10:38:45 [1] Step: 123500 Acc: 0.69781 0.84764 Cost: 1.13772 0.87376 0.15818 0.10577 Time: 0.00083
17-03-28 10:38:45 [1] Train Extra: lr=0.0000086 inv=0.4153125 sub=0.0000000
17-03-28 10:39:55 [1] Step: 123500 Eval acc: 0.67218 0.85289 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 10:39:55 [1] Eval Extra: inv=0.3985907
17-03-28 10:41:22 [1] Step: 123600 Acc: 0.69750 0.85820 Cost: 0.98793 0.69438 0.18767 0.10588 Time: 0.00084
17-03-28 10:41:22 [1] Train Extra: lr=0.0000086 inv=0.3815625 sub=0.0000000
17-03-28 10:42:59 [1] Step: 123700 Acc: 0.69344 0.85257 Cost: 0.96816 0.66174 0.20058 0.10584 Time: 0.00086
17-03-28 10:42:59 [1] Train Extra: lr=0.0000085 inv=0.4268750 sub=0.0000000
17-03-28 10:44:19 [1] Step: 123800 Acc: 0.68937 0.85553 Cost: 0.72848 0.51848 0.10411 0.10588 Time: 0.00080
17-03-28 10:44:19 [1] Train Extra: lr=0.0000085 inv=0.3884375 sub=0.0000000
17-03-28 10:45:50 [1] Step: 123900 Acc: 0.69594 0.84326 Cost: 1.17402 0.77504 0.29315 0.10584 Time: 0.00084
17-03-28 10:45:50 [1] Train Extra: lr=0.0000085 inv=0.4548437 sub=0.0000000
17-03-28 10:47:18 [1] Step: 124000 Acc: 0.68031 0.85207 Cost: 1.12255 0.81702 0.19976 0.10577 Time: 0.00083
17-03-28 10:47:18 [1] Train Extra: lr=0.0000085 inv=0.4120313 sub=0.0000000
17-03-28 10:48:28 [1] Step: 124000 Eval acc: 0.67167 0.86125 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 10:48:28 [1] Eval Extra: inv=0.3898591
17-03-28 10:49:48 [1] Step: 124100 Acc: 0.70375 0.84864 Cost: 1.16264 0.75930 0.29762 0.10573 Time: 0.00078
17-03-28 10:49:48 [1] Train Extra: lr=0.0000084 inv=0.4204688 sub=0.0000000
17-03-28 10:51:14 [1] Step: 124200 Acc: 0.69531 0.84685 Cost: 1.04965 0.68561 0.25823 0.10582 Time: 0.00082
17-03-28 10:51:14 [1] Train Extra: lr=0.0000084 inv=0.4129687 sub=0.0000000
17-03-28 10:52:38 [1] Step: 124300 Acc: 0.69750 0.85273 Cost: 1.12893 0.87498 0.14809 0.10587 Time: 0.00082
17-03-28 10:52:38 [1] Train Extra: lr=0.0000084 inv=0.4114062 sub=0.0000000
17-03-28 10:54:08 [1] Step: 124400 Acc: 0.70000 0.85591 Cost: 1.17143 0.70844 0.35715 0.10584 Time: 0.00087
17-03-28 10:54:08 [1] Train Extra: lr=0.0000084 inv=0.4225000 sub=0.0000000
17-03-28 10:55:41 [1] Step: 124500 Acc: 0.69469 0.85101 Cost: 0.84279 0.62019 0.11665 0.10595 Time: 0.00085
17-03-28 10:55:41 [1] Train Extra: lr=0.0000083 inv=0.3995313 sub=0.0000000
17-03-28 10:56:50 [1] Step: 124500 Eval acc: 0.67627 0.86091 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 10:56:50 [1] Eval Extra: inv=0.4075266
17-03-28 10:58:16 [1] Step: 124600 Acc: 0.70094 0.85532 Cost: 0.87991 0.60936 0.16463 0.10592 Time: 0.00081
17-03-28 10:58:16 [1] Train Extra: lr=0.0000083 inv=0.4260937 sub=0.0000000
17-03-28 10:59:38 [1] Step: 124700 Acc: 0.70281 0.86031 Cost: 0.74833 0.49313 0.14926 0.10593 Time: 0.00085
17-03-28 10:59:38 [1] Train Extra: lr=0.0000083 inv=0.3804688 sub=0.0000000
17-03-28 11:00:58 [1] Step: 124800 Acc: 0.70000 0.85596 Cost: 0.98974 0.65392 0.22984 0.10598 Time: 0.00082
17-03-28 11:00:58 [1] Train Extra: lr=0.0000083 inv=0.3971875 sub=0.0000000
17-03-28 11:02:31 [1] Step: 124900 Acc: 0.67937 0.85012 Cost: 1.12823 0.75522 0.26697 0.10605 Time: 0.00085
17-03-28 11:02:31 [1] Train Extra: lr=0.0000083 inv=0.4201563 sub=0.0000000
17-03-28 11:03:57 [1] Step: 125000 Acc: 0.69219 0.84898 Cost: 1.08289 0.72643 0.25046 0.10600 Time: 0.00080
17-03-28 11:03:57 [1] Train Extra: lr=0.0000082 inv=0.4335938 sub=0.0000000
17-03-28 11:05:07 [1] Step: 125000 Eval acc: 0.67351 0.85999 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 11:05:07 [1] Eval Extra: inv=0.4056373
17-03-28 11:05:07 [1] Checkpointing.
17-03-28 11:06:39 [1] Step: 125100 Acc: 0.71500 0.85078 Cost: 1.20631 0.79102 0.30921 0.10608 Time: 0.00084
17-03-28 11:06:39 [1] Train Extra: lr=0.0000082 inv=0.3964063 sub=0.0000000
17-03-28 11:08:04 [1] Step: 125200 Acc: 0.69406 0.84557 Cost: 1.11162 0.72443 0.28112 0.10606 Time: 0.00081
17-03-28 11:08:04 [1] Train Extra: lr=0.0000082 inv=0.4325000 sub=0.0000000
17-03-28 11:09:26 [1] Step: 125300 Acc: 0.70469 0.85133 Cost: 1.00463 0.70968 0.18889 0.10605 Time: 0.00081
17-03-28 11:09:26 [1] Train Extra: lr=0.0000082 inv=0.3973437 sub=0.0000000
17-03-28 11:11:00 [1] Step: 125400 Acc: 0.70312 0.85057 Cost: 1.06503 0.76588 0.19309 0.10606 Time: 0.00081
17-03-28 11:11:00 [1] Train Extra: lr=0.0000081 inv=0.4671875 sub=0.0000000
17-03-28 11:12:33 [1] Step: 125500 Acc: 0.68031 0.85368 Cost: 0.98136 0.69558 0.17969 0.10609 Time: 0.00085
17-03-28 11:12:33 [1] Train Extra: lr=0.0000081 inv=0.4285937 sub=0.0000000
17-03-28 11:13:45 [1] Step: 125500 Eval acc: 0.67514 0.85835 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 11:13:45 [1] Eval Extra: inv=0.3906250
17-03-28 11:15:02 [1] Step: 125600 Acc: 0.68719 0.85452 Cost: 0.74283 0.56888 0.06787 0.10609 Time: 0.00081
17-03-28 11:15:02 [1] Train Extra: lr=0.0000081 inv=0.3721875 sub=0.0000000
17-03-28 11:16:36 [1] Step: 125700 Acc: 0.68125 0.85352 Cost: 1.30865 0.94176 0.26078 0.10611 Time: 0.00084
17-03-28 11:16:36 [1] Train Extra: lr=0.0000081 inv=0.4243750 sub=0.0000000
17-03-28 11:18:04 [1] Step: 125800 Acc: 0.69875 0.85094 Cost: 0.81406 0.58438 0.12367 0.10602 Time: 0.00084
17-03-28 11:18:04 [1] Train Extra: lr=0.0000080 inv=0.4032812 sub=0.0000000
17-03-28 11:19:38 [1] Step: 125900 Acc: 0.69563 0.85078 Cost: 0.94776 0.70217 0.13957 0.10602 Time: 0.00082
17-03-28 11:19:38 [1] Train Extra: lr=0.0000080 inv=0.4446875 sub=0.0000000
17-03-28 11:21:11 [1] Step: 126000 Acc: 0.67563 0.85009 Cost: 0.90975 0.60151 0.20222 0.10602 Time: 0.00084
17-03-28 11:21:11 [1] Train Extra: lr=0.0000080 inv=0.4350000 sub=0.0000000
17-03-28 11:22:17 [1] Step: 126000 Eval acc: 0.67290 0.85995 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00019
17-03-28 11:22:17 [1] Eval Extra: inv=0.3706087
17-03-28 11:23:50 [1] Step: 126100 Acc: 0.69437 0.85642 Cost: 0.71842 0.53147 0.08094 0.10602 Time: 0.00086
17-03-28 11:23:50 [1] Train Extra: lr=0.0000080 inv=0.4223438 sub=0.0000000
17-03-28 11:25:31 [1] Step: 126200 Acc: 0.68125 0.85876 Cost: 1.24755 0.92070 0.22077 0.10609 Time: 0.00089
17-03-28 11:25:31 [1] Train Extra: lr=0.0000080 inv=0.4075000 sub=0.0000000
17-03-28 11:27:09 [1] Step: 126300 Acc: 0.69937 0.85606 Cost: 1.16627 0.93841 0.12177 0.10608 Time: 0.00089
17-03-28 11:27:09 [1] Train Extra: lr=0.0000079 inv=0.4095313 sub=0.0000000
17-03-28 11:28:42 [1] Step: 126400 Acc: 0.70656 0.85562 Cost: 0.75266 0.53071 0.11585 0.10610 Time: 0.00085
17-03-28 11:28:42 [1] Train Extra: lr=0.0000079 inv=0.4115625 sub=0.0000000
17-03-28 11:30:02 [1] Step: 126500 Acc: 0.69781 0.86059 Cost: 1.00357 0.64221 0.25521 0.10616 Time: 0.00084
17-03-28 11:30:02 [1] Train Extra: lr=0.0000079 inv=0.3678125 sub=0.0000000
17-03-28 11:31:13 [1] Step: 126500 Eval acc: 0.67953 0.86273 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 11:31:13 [1] Eval Extra: inv=0.3812806
17-03-28 11:32:40 [1] Step: 126600 Acc: 0.68625 0.85783 Cost: 1.09750 0.76571 0.22569 0.10610 Time: 0.00083
17-03-28 11:32:40 [1] Train Extra: lr=0.0000079 inv=0.4162500 sub=0.0000000
17-03-28 11:34:10 [1] Step: 126700 Acc: 0.68281 0.84542 Cost: 1.04553 0.84107 0.09834 0.10613 Time: 0.00082
17-03-28 11:34:10 [1] Train Extra: lr=0.0000078 inv=0.4170313 sub=0.0000000
17-03-28 11:35:42 [1] Step: 126800 Acc: 0.68000 0.84232 Cost: 1.27053 0.89908 0.26524 0.10622 Time: 0.00080
17-03-28 11:35:42 [1] Train Extra: lr=0.0000078 inv=0.4521875 sub=0.0000000
17-03-28 11:37:21 [1] Step: 126900 Acc: 0.67344 0.85582 Cost: 1.11194 0.81701 0.18872 0.10622 Time: 0.00087
17-03-28 11:37:21 [1] Train Extra: lr=0.0000078 inv=0.4315625 sub=0.0000000
17-03-28 11:38:50 [1] Step: 127000 Acc: 0.67469 0.85586 Cost: 0.93096 0.66114 0.16355 0.10627 Time: 0.00086
17-03-28 11:38:50 [1] Train Extra: lr=0.0000078 inv=0.4148438 sub=0.0000000
17-03-28 11:40:01 [1] Step: 127000 Eval acc: 0.67269 0.85461 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 11:40:01 [1] Eval Extra: inv=0.4149816
17-03-28 11:41:44 [1] Step: 127100 Acc: 0.71719 0.85381 Cost: 1.10131 0.75745 0.23751 0.10634 Time: 0.00091
17-03-28 11:41:44 [1] Train Extra: lr=0.0000077 inv=0.4157812 sub=0.0000000
17-03-28 11:43:16 [1] Step: 127200 Acc: 0.71688 0.85618 Cost: 0.97661 0.65280 0.21739 0.10641 Time: 0.00087
17-03-28 11:43:16 [1] Train Extra: lr=0.0000077 inv=0.3971875 sub=0.0000000
17-03-28 11:44:49 [1] Step: 127300 Acc: 0.71469 0.85132 Cost: 1.04957 0.69439 0.24864 0.10654 Time: 0.00088
17-03-28 11:44:49 [1] Train Extra: lr=0.0000077 inv=0.4009375 sub=0.0000000
17-03-28 11:46:15 [1] Step: 127400 Acc: 0.71594 0.85504 Cost: 0.64774 0.53424 0.00682 0.10668 Time: 0.00084
17-03-28 11:46:15 [1] Train Extra: lr=0.0000077 inv=0.3926562 sub=0.0000000
17-03-28 11:47:49 [1] Step: 127500 Acc: 0.70969 0.85446 Cost: 1.01238 0.72587 0.17973 0.10678 Time: 0.00084
17-03-28 11:47:49 [1] Train Extra: lr=0.0000077 inv=0.4457813 sub=0.0000000
17-03-28 11:48:57 [1] Step: 127500 Eval acc: 0.67320 0.85682 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 11:48:57 [1] Eval Extra: inv=0.4098243
17-03-28 11:50:22 [1] Step: 127600 Acc: 0.70250 0.85041 Cost: 0.71639 0.49616 0.11336 0.10687 Time: 0.00081
17-03-28 11:50:22 [1] Train Extra: lr=0.0000076 inv=0.4376563 sub=0.0000000
17-03-28 11:51:49 [1] Step: 127700 Acc: 0.70937 0.85310 Cost: 0.71641 0.48766 0.12177 0.10698 Time: 0.00081
17-03-28 11:51:49 [1] Train Extra: lr=0.0000076 inv=0.4073438 sub=0.0000000
17-03-28 11:53:18 [1] Step: 127800 Acc: 0.69594 0.85002 Cost: 1.15612 0.83196 0.21709 0.10706 Time: 0.00084
17-03-28 11:53:18 [1] Train Extra: lr=0.0000076 inv=0.4170313 sub=0.0000000
17-03-28 11:54:43 [1] Step: 127900 Acc: 0.71156 0.84489 Cost: 1.00393 0.65222 0.24456 0.10714 Time: 0.00080
17-03-28 11:54:43 [1] Train Extra: lr=0.0000076 inv=0.4264062 sub=0.0000000
17-03-28 11:56:08 [1] Step: 128000 Acc: 0.70781 0.85509 Cost: 0.79449 0.47485 0.21232 0.10732 Time: 0.00082
17-03-28 11:56:08 [1] Train Extra: lr=0.0000075 inv=0.4203125 sub=0.0000000
17-03-28 11:57:17 [1] Step: 128000 Eval acc: 0.67310 0.85592 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 11:57:17 [1] Eval Extra: inv=0.4398999
17-03-28 11:58:55 [1] Step: 128100 Acc: 0.70531 0.85176 Cost: 1.24956 0.81416 0.32797 0.10744 Time: 0.00085
17-03-28 11:58:55 [1] Train Extra: lr=0.0000075 inv=0.4529687 sub=0.0000000
17-03-28 12:00:29 [1] Step: 128200 Acc: 0.71125 0.85546 Cost: 1.01732 0.74685 0.16295 0.10752 Time: 0.00084
17-03-28 12:00:29 [1] Train Extra: lr=0.0000075 inv=0.4215625 sub=0.0000000
17-03-28 12:01:52 [1] Step: 128300 Acc: 0.72000 0.85848 Cost: 0.86698 0.57338 0.18599 0.10761 Time: 0.00084
17-03-28 12:01:52 [1] Train Extra: lr=0.0000075 inv=0.3959375 sub=0.0000000
17-03-28 12:03:19 [1] Step: 128400 Acc: 0.71062 0.85365 Cost: 1.04919 0.71912 0.22234 0.10774 Time: 0.00082
17-03-28 12:03:19 [1] Train Extra: lr=0.0000075 inv=0.4225000 sub=0.0000000
17-03-28 12:04:58 [1] Step: 128500 Acc: 0.70437 0.84397 Cost: 0.97826 0.70324 0.16720 0.10781 Time: 0.00083
17-03-28 12:04:58 [1] Train Extra: lr=0.0000074 inv=0.4743750 sub=0.0000000
17-03-28 12:06:09 [1] Step: 128500 Eval acc: 0.67371 0.86070 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00020
17-03-28 12:06:09 [1] Eval Extra: inv=0.3968546
17-03-28 12:07:34 [1] Step: 128600 Acc: 0.70531 0.85543 Cost: 0.99641 0.61907 0.26948 0.10786 Time: 0.00083
17-03-28 12:07:34 [1] Train Extra: lr=0.0000074 inv=0.3985938 sub=0.0000000
17-03-28 12:08:52 [1] Step: 128700 Acc: 0.70344 0.85486 Cost: 1.16536 0.77979 0.27765 0.10792 Time: 0.00079
17-03-28 12:08:52 [1] Train Extra: lr=0.0000074 inv=0.4053125 sub=0.0000000
17-03-28 12:10:24 [1] Step: 128800 Acc: 0.69875 0.84828 Cost: 0.70536 0.49891 0.09849 0.10796 Time: 0.00084
17-03-28 12:10:24 [1] Train Extra: lr=0.0000074 inv=0.4067188 sub=0.0000000
17-03-28 12:11:54 [1] Step: 128900 Acc: 0.72000 0.85222 Cost: 0.98103 0.67994 0.19309 0.10800 Time: 0.00083
17-03-28 12:11:54 [1] Train Extra: lr=0.0000074 inv=0.4304688 sub=0.0000000
17-03-28 12:13:31 [1] Step: 129000 Acc: 0.70969 0.85462 Cost: 1.10740 0.79717 0.20211 0.10811 Time: 0.00087
17-03-28 12:13:31 [1] Train Extra: lr=0.0000073 inv=0.4321875 sub=0.0000000
17-03-28 12:14:36 [1] Step: 129000 Eval acc: 0.67004 0.85939 /home/dexter/data/multinli_0.2/multinli_0.2_dev_matched.jsonl Time: 0.00018
17-03-28 12:14:36 [1] Eval Extra: inv=0.4078840