multinli.log

multinli.log

17-03-25 12:22:02 [1] Flag values:
{   '?': None,
    'actively_decay_learning_rate': True,
    'batch_size': 32,
    'branch_name': 'master',
    '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': 'multisnli',
    'debug': False,
    'deque_length': None,
    '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': None,
    'eval_data_limit': -1,
    'eval_data_path': '/home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl',
    'eval_interval_steps': 500,
    'eval_report_use_preds': True,
    'eval_seq_length': None,
    'evalb': False,
    'expanded_eval_only_mode': False,
    'experiment_name': 'spinn-multisnli-eid_01',
    'gen_h': True,
    'gpu': 0,
    'help': None,
    'helpshort': None,
    'helpxml': None,
    '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': '2bf8089be8b4737c6097cba003f0931b0283242c',
    '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_0.1/multinli_0.1_train.jsonl',
    'training_steps': 250000,
    'transition_weight': 0.6,
    '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}
17-03-25 12:22:25 [1] In open vocabulary mode. Using loaded embeddings without fine-tuning.
17-03-25 12:22:25 [1] Constructing vocabulary...
17-03-25 12:22:26 [1] Found 82433 word types.
17-03-25 12:22:39 [1] Loading vocabulary with 73546 words from /home/dexter/data/glove/glove.840B.300d.txt
17-03-25 12:23:07 [1] Preprocessing training data.
17-03-25 12:23:49 [1] Preprocessing eval data: /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl
17-03-25 12:23:55 [1] Building model.
17-03-25 12:23:56 [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)
  )
)
17-03-25 12:23:56 [1] Total params: 3581817.0
17-03-25 12:24:03 [1]

# ----- BEGIN: Log Configuration ----- #
17-03-25 12:24:03 [1] Flag-JSON: {"eval_seq_length": null, "lowercase": false, "clipping_max_value": 5.0, "use_peano": true, "log_path": "/home/dexter/logs/spinn", "embedding_keep_rate": 0.9, "rl_mu": 0.1, "training_data_path": "/home/dexter/data/multinli_0.1/multinli_0.1_train.jsonl", "use_difference_feature": true, "init_range": 0.005, "evalb": false, "rl_entropy": false, "rl_whiten": false, "show_progress_bar": true, "use_l2_cost": true, "actively_decay_learning_rate": true, "use_encode": false, "encode_style": null, "encode_num_layers": 1, "help": null, "use_lengths": false, "rl_entropy_beta": 0.001, "embedding_data_path": "/home/dexter/data/glove/glove.840B.300d.txt", "write_eval_report": false, "model_dim": 600, "ckpt_on_best_dev_error": true, "deque_length": null, "seq_length": 500, "predict_use_cell": true, "eval_data_limit": -1, "word_embedding_dim": 300, "use_internal_parser": true, "ckpt_path": "/home/dexter/logs/spinn", "expanded_eval_only_mode": false, "eval_report_use_preds": true, "?": null, "helpxml": null, "bucket_eval": true, "semantic_classifier_keep_rate": 0.9, "lateral_tracking": true, "eval_interval_steps": 500, "data_type": "multisnli", "metrics_interval_steps": 10, "helpshort": null, "rl_weight": 1.0, "learning_rate": 0.0003, "metrics_path": "/home/dexter/logs/spinn-runs", "gpu": 0, "batch_size": 32, "use_product_feature": true, "smart_batching": true, "branch_name": "master", "encode_bidirectional": false, "validate_transitions": true, "optimizer_type": "RMSprop", "rl_baseline": "ema", "shuffle_eval": false, "shuffle_eval_seed": 123, "l2_lambda": 2.75e-05, "training_steps": 250000, "debug": false, "gen_h": true, "use_tracking_in_composition": true, "tracking_lstm_hidden_dim": 40, "rl_reward": "standard", "rl_epsilon_decay": 50000.0, "mlp_dim": 1024, "statistics_interval_steps": 100, "predict_leaf": true, "encode_reverse": false, "learning_rate_decay_per_10k_steps": 0.75, "num_mlp_layers": 2, "load_best": false, "sha": "2bf8089be8b4737c6097cba003f0931b0283242c", "experiment_name": "spinn-multisnli-eid_01", "num_samples": 0, "model_type": "SPINN", "ckpt_interval_steps": 5000, "mlp_bn": true, "rl_epsilon": 1.0, "transition_weight": 0.6, "ckpt_step": 1000, "eval_data_path": "/home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl"}
17-03-25 12:24:03 [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}
17-03-25 12:24:03 [1] Train-Extra-Format: Train Extra: lr={learning_rate:.7f} inv={invalid:.7f} sub={struct:.7f}
17-03-25 12:24:03 [1] Eval-Format: Step: {step} Eval acc: {class_acc:.5f} {transition_acc:.5f} {filename} Time: {time:.5f}
17-03-25 12:24:03 [1] Eval-Extra-Format: Eval Extra: inv={inv:.7f}
17-03-25 12:24:03 [1] # ----- END: Log Configuration ----- #


17-03-25 12:24:03 [1] Training.
17-03-25 12:24:03 [1] Step: 0 Acc: 0.50000 0.65000 Cost: 1.95439 1.32103 0.41352 0.21984 Time: 0.00032
17-03-25 12:24:03 [1] Train Extra: lr=0.0003000 inv=1.0000000 sub=0.0000000
17-03-25 12:25:08 [1] Step: 100 Acc: 0.34250 0.76268 Cost: 2.05374 1.50317 0.32934 0.22123 Time: 0.00064
17-03-25 12:25:08 [1] Train Extra: lr=0.0002991 inv=0.8121875 sub=0.0000000
17-03-25 12:26:23 [1] Step: 200 Acc: 0.37938 0.76642 Cost: 1.85071 1.24511 0.38526 0.22034 Time: 0.00071
17-03-25 12:26:23 [1] Train Extra: lr=0.0002983 inv=0.6070313 sub=0.0000000
17-03-25 12:27:48 [1] Step: 300 Acc: 0.39813 0.76079 Cost: 1.66557 1.09645 0.34982 0.21930 Time: 0.00070
17-03-25 12:27:48 [1] Train Extra: lr=0.0002974 inv=0.4343750 sub=0.0000000
17-03-25 12:29:03 [1] Step: 400 Acc: 0.39313 0.76887 Cost: 1.68176 1.10038 0.36312 0.21826 Time: 0.00070
17-03-25 12:29:03 [1] Train Extra: lr=0.0002966 inv=0.3964063 sub=0.0000000
17-03-25 12:30:19 [1] Step: 500 Acc: 0.39844 0.76578 Cost: 1.83541 1.27153 0.34681 0.21706 Time: 0.00069
17-03-25 12:30:19 [1] Train Extra: lr=0.0002957 inv=0.3464063 sub=0.0000000
17-03-25 12:31:11 [1] Step: 500 Eval acc: 0.39852 0.77655 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 12:31:11 [1] Eval Extra: inv=0.3164753
17-03-25 12:32:26 [1] Step: 600 Acc: 0.40969 0.76364 Cost: 1.61752 1.07206 0.32967 0.21579 Time: 0.00069
17-03-25 12:32:26 [1] Train Extra: lr=0.0002949 inv=0.2776562 sub=0.0000000
17-03-25 12:33:42 [1] Step: 700 Acc: 0.44156 0.76262 Cost: 1.76364 1.19679 0.35248 0.21437 Time: 0.00068
17-03-25 12:33:42 [1] Train Extra: lr=0.0002940 inv=0.3323437 sub=0.0000000
17-03-25 12:34:54 [1] Step: 800 Acc: 0.40813 0.75632 Cost: 1.63475 1.13143 0.29044 0.21288 Time: 0.00065
17-03-25 12:34:54 [1] Train Extra: lr=0.0002932 inv=0.4165625 sub=0.0000000
17-03-25 12:36:15 [1] Step: 900 Acc: 0.43062 0.77212 Cost: 1.68800 1.11302 0.36383 0.21115 Time: 0.00070
17-03-25 12:36:15 [1] Train Extra: lr=0.0002923 inv=0.5193750 sub=0.0000000
17-03-25 12:37:34 [1] Step: 1000 Acc: 0.43250 0.77250 Cost: 1.69298 1.19135 0.29237 0.20926 Time: 0.00071
17-03-25 12:37:34 [1] Train Extra: lr=0.0002915 inv=0.5990625 sub=0.0000000
17-03-25 12:38:26 [1] Step: 1000 Eval acc: 0.46190 0.78542 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 12:38:26 [1] Eval Extra: inv=0.5682420
17-03-25 12:39:48 [1] Step: 1100 Acc: 0.44250 0.77437 Cost: 1.74088 1.21294 0.32066 0.20728 Time: 0.00070
17-03-25 12:39:48 [1] Train Extra: lr=0.0002907 inv=0.6142188 sub=0.0000000
17-03-25 12:41:09 [1] Step: 1200 Acc: 0.45188 0.78035 Cost: 1.52425 1.02425 0.29476 0.20525 Time: 0.00071
17-03-25 12:41:09 [1] Train Extra: lr=0.0002898 inv=0.6270313 sub=0.0000000
17-03-25 12:42:23 [1] Step: 1300 Acc: 0.42625 0.79125 Cost: 1.34814 0.86456 0.28059 0.20298 Time: 0.00072
17-03-25 12:42:23 [1] Train Extra: lr=0.0002890 inv=0.6712500 sub=0.0000000
17-03-25 12:43:42 [1] Step: 1400 Acc: 0.43875 0.78914 Cost: 1.55171 0.99565 0.35552 0.20053 Time: 0.00071
17-03-25 12:43:42 [1] Train Extra: lr=0.0002882 inv=0.7209375 sub=0.0000000
17-03-25 12:44:55 [1] Step: 1500 Acc: 0.44656 0.77754 Cost: 1.78472 1.28337 0.30328 0.19808 Time: 0.00070
17-03-25 12:44:55 [1] Train Extra: lr=0.0002873 inv=0.7131250 sub=0.0000000
17-03-25 12:45:47 [1] Step: 1500 Eval acc: 0.47383 0.79337 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 12:45:47 [1] Eval Extra: inv=0.7365835
17-03-25 12:45:47 [1] Checkpointing with new best dev accuracy of 0.473830
17-03-25 12:46:59 [1] Step: 1600 Acc: 0.45594 0.78119 Cost: 1.44303 0.99603 0.25142 0.19558 Time: 0.00068
17-03-25 12:46:59 [1] Train Extra: lr=0.0002865 inv=0.7526563 sub=0.0000000
17-03-25 12:48:19 [1] Step: 1700 Acc: 0.44844 0.79489 Cost: 1.58261 1.08556 0.30403 0.19302 Time: 0.00072
17-03-25 12:48:19 [1] Train Extra: lr=0.0002857 inv=0.7492188 sub=0.0000000
17-03-25 12:49:30 [1] Step: 1800 Acc: 0.45250 0.79429 Cost: 1.59793 1.10512 0.30239 0.19042 Time: 0.00068
17-03-25 12:49:30 [1] Train Extra: lr=0.0002849 inv=0.7306250 sub=0.0000000
17-03-25 12:50:38 [1] Step: 1900 Acc: 0.48156 0.78956 Cost: 1.64687 1.19953 0.25936 0.18798 Time: 0.00069
17-03-25 12:50:38 [1] Train Extra: lr=0.0002840 inv=0.6548437 sub=0.0000000
17-03-25 12:51:54 [1] Step: 2000 Acc: 0.46750 0.79571 Cost: 1.50675 1.08365 0.23772 0.18538 Time: 0.00071
17-03-25 12:51:54 [1] Train Extra: lr=0.0002832 inv=0.6664062 sub=0.0000000
17-03-25 12:52:47 [1] Step: 2000 Eval acc: 0.50398 0.80907 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 12:52:47 [1] Eval Extra: inv=0.6366497
17-03-25 12:52:47 [1] Checkpointing with new best dev accuracy of 0.503975
17-03-25 12:54:04 [1] Step: 2100 Acc: 0.49281 0.80517 Cost: 1.43064 0.95514 0.29265 0.18284 Time: 0.00069
17-03-25 12:54:04 [1] Train Extra: lr=0.0002824 inv=0.6156250 sub=0.0000000
17-03-25 12:55:12 [1] Step: 2200 Acc: 0.48438 0.80812 Cost: 1.45349 0.99554 0.27766 0.18030 Time: 0.00070
17-03-25 12:55:12 [1] Train Extra: lr=0.0002816 inv=0.5689062 sub=0.0000000
17-03-25 12:56:31 [1] Step: 2300 Acc: 0.49375 0.79922 Cost: 1.56481 1.04250 0.34449 0.17782 Time: 0.00072
17-03-25 12:56:31 [1] Train Extra: lr=0.0002808 inv=0.5881250 sub=0.0000000
17-03-25 12:57:48 [1] Step: 2400 Acc: 0.49312 0.80316 Cost: 1.40881 0.97310 0.26049 0.17521 Time: 0.00068
17-03-25 12:57:48 [1] Train Extra: lr=0.0002800 inv=0.6132812 sub=0.0000000
17-03-25 12:59:08 [1] Step: 2500 Acc: 0.49906 0.80039 Cost: 1.46869 1.11782 0.17823 0.17264 Time: 0.00071
17-03-25 12:59:08 [1] Train Extra: lr=0.0002792 inv=0.5679688 sub=0.0000000
17-03-25 13:00:01 [1] Step: 2500 Eval acc: 0.52142 0.80933 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 13:00:01 [1] Eval Extra: inv=0.5692911
17-03-25 13:00:01 [1] Checkpointing with new best dev accuracy of 0.521422
17-03-25 13:01:26 [1] Step: 2600 Acc: 0.50281 0.81646 Cost: 1.63090 1.14184 0.31891 0.17015 Time: 0.00072
17-03-25 13:01:26 [1] Train Extra: lr=0.0002784 inv=0.5957813 sub=0.0000000
17-03-25 13:02:45 [1] Step: 2700 Acc: 0.48000 0.80180 Cost: 1.51136 1.03261 0.31118 0.16757 Time: 0.00070
17-03-25 13:02:45 [1] Train Extra: lr=0.0002776 inv=0.5890625 sub=0.0000000
17-03-25 13:03:59 [1] Step: 2800 Acc: 0.50906 0.80804 Cost: 1.33601 0.90909 0.26182 0.16510 Time: 0.00070
17-03-25 13:03:59 [1] Train Extra: lr=0.0002768 inv=0.5420312 sub=0.0000000
17-03-25 13:05:16 [1] Step: 2900 Acc: 0.50938 0.80636 Cost: 1.33899 0.87865 0.29759 0.16275 Time: 0.00071
17-03-25 13:05:16 [1] Train Extra: lr=0.0002760 inv=0.5462500 sub=0.0000000
17-03-25 13:06:33 [1] Step: 3000 Acc: 0.53094 0.81199 Cost: 1.49967 1.00565 0.33348 0.16054 Time: 0.00072
17-03-25 13:06:33 [1] Train Extra: lr=0.0002752 inv=0.5117188 sub=0.0000000
17-03-25 13:07:27 [1] Step: 3000 Eval acc: 0.54042 0.81654 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 13:07:27 [1] Eval Extra: inv=0.5187721
17-03-25 13:07:27 [1] Checkpointing with new best dev accuracy of 0.540415
17-03-25 13:08:41 [1] Step: 3100 Acc: 0.50125 0.81180 Cost: 1.28894 0.86797 0.26257 0.15840 Time: 0.00070
17-03-25 13:08:41 [1] Train Extra: lr=0.0002744 inv=0.5264062 sub=0.0000000
17-03-25 13:09:59 [1] Step: 3200 Acc: 0.51969 0.81173 Cost: 1.33330 0.91302 0.26406 0.15622 Time: 0.00069
17-03-25 13:09:59 [1] Train Extra: lr=0.0002736 inv=0.5468750 sub=0.0000000
17-03-25 13:11:26 [1] Step: 3300 Acc: 0.53187 0.82178 Cost: 1.42956 0.95605 0.31939 0.15412 Time: 0.00075
17-03-25 13:11:26 [1] Train Extra: lr=0.0002728 inv=0.5354688 sub=0.0000000
17-03-25 13:12:36 [1] Step: 3400 Acc: 0.52719 0.81252 Cost: 1.40823 0.98658 0.26959 0.15206 Time: 0.00068
17-03-25 13:12:36 [1] Train Extra: lr=0.0002720 inv=0.5142187 sub=0.0000000
17-03-25 13:13:58 [1] Step: 3500 Acc: 0.52156 0.81686 Cost: 1.32857 0.96513 0.21345 0.14999 Time: 0.00074
17-03-25 13:13:58 [1] Train Extra: lr=0.0002713 inv=0.4973437 sub=0.0000000
17-03-25 13:14:53 [1] Step: 3500 Eval acc: 0.54770 0.82634 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 13:14:53 [1] Eval Extra: inv=0.5067359
17-03-25 13:14:53 [1] Checkpointing with new best dev accuracy of 0.547703
17-03-25 13:16:06 [1] Step: 3600 Acc: 0.52719 0.81069 Cost: 1.35194 0.97433 0.22955 0.14806 Time: 0.00070
17-03-25 13:16:06 [1] Train Extra: lr=0.0002705 inv=0.5070313 sub=0.0000000
17-03-25 13:17:28 [1] Step: 3700 Acc: 0.52625 0.82616 Cost: 1.39095 1.03218 0.21255 0.14622 Time: 0.00074
17-03-25 13:17:28 [1] Train Extra: lr=0.0002697 inv=0.5184375 sub=0.0000000
17-03-25 13:18:44 [1] Step: 3800 Acc: 0.53031 0.81177 Cost: 1.20582 0.77464 0.28681 0.14436 Time: 0.00068
17-03-25 13:18:44 [1] Train Extra: lr=0.0002689 inv=0.5135937 sub=0.0000000
17-03-25 13:20:06 [1] Step: 3900 Acc: 0.52531 0.81840 Cost: 1.30600 0.86957 0.29397 0.14246 Time: 0.00073
17-03-25 13:20:06 [1] Train Extra: lr=0.0002682 inv=0.5006250 sub=0.0000000
17-03-25 13:21:17 [1] Step: 4000 Acc: 0.53469 0.82776 Cost: 1.31233 0.92686 0.24480 0.14067 Time: 0.00070
17-03-25 13:21:17 [1] Train Extra: lr=0.0002674 inv=0.4832812 sub=0.0000000
17-03-25 13:22:10 [1] Step: 4000 Eval acc: 0.54826 0.82175 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 13:22:10 [1] Eval Extra: inv=0.5276060
17-03-25 13:23:38 [1] Step: 4100 Acc: 0.53219 0.82407 Cost: 1.28495 0.91747 0.22866 0.13882 Time: 0.00074
17-03-25 13:23:38 [1] Train Extra: lr=0.0002666 inv=0.5520312 sub=0.0000000
17-03-25 13:24:59 [1] Step: 4200 Acc: 0.51062 0.82012 Cost: 1.33435 0.93627 0.26084 0.13724 Time: 0.00072
17-03-25 13:24:59 [1] Train Extra: lr=0.0002659 inv=0.5201562 sub=0.0000000
17-03-25 13:26:12 [1] Step: 4300 Acc: 0.54031 0.81816 Cost: 1.37156 0.97221 0.26376 0.13559 Time: 0.00070
17-03-25 13:26:12 [1] Train Extra: lr=0.0002651 inv=0.4826563 sub=0.0000000
17-03-25 13:27:23 [1] Step: 4400 Acc: 0.54375 0.82293 Cost: 1.12873 0.81513 0.17942 0.13418 Time: 0.00068
17-03-25 13:27:23 [1] Train Extra: lr=0.0002643 inv=0.4870313 sub=0.0000000
17-03-25 13:28:43 [1] Step: 4500 Acc: 0.52687 0.82540 Cost: 1.28448 0.85781 0.29397 0.13271 Time: 0.00073
17-03-25 13:28:43 [1] Train Extra: lr=0.0002636 inv=0.4910937 sub=0.0000000
17-03-25 13:29:35 [1] Step: 4500 Eval acc: 0.56383 0.83190 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 13:29:35 [1] Eval Extra: inv=0.4605234
17-03-25 13:29:35 [1] Checkpointing with new best dev accuracy of 0.563825
17-03-25 13:30:55 [1] Step: 4600 Acc: 0.54719 0.83723 Cost: 1.28876 0.86712 0.29043 0.13121 Time: 0.00075
17-03-25 13:30:55 [1] Train Extra: lr=0.0002628 inv=0.5028125 sub=0.0000000
17-03-25 13:32:12 [1] Step: 4700 Acc: 0.53531 0.83128 Cost: 1.40526 0.93857 0.33691 0.12978 Time: 0.00072
17-03-25 13:32:12 [1] Train Extra: lr=0.0002621 inv=0.4978125 sub=0.0000000
17-03-25 13:33:33 [1] Step: 4800 Acc: 0.55188 0.81845 Cost: 1.05536 0.72330 0.20349 0.12857 Time: 0.00071
17-03-25 13:33:33 [1] Train Extra: lr=0.0002613 inv=0.5167188 sub=0.0000000
17-03-25 13:34:54 [1] Step: 4900 Acc: 0.50813 0.82556 Cost: 1.31892 0.93434 0.25745 0.12713 Time: 0.00072
17-03-25 13:34:54 [1] Train Extra: lr=0.0002606 inv=0.5117188 sub=0.0000000
17-03-25 13:36:26 [1] Step: 5000 Acc: 0.52750 0.83769 Cost: 1.41080 0.99121 0.29372 0.12587 Time: 0.00076
17-03-25 13:36:26 [1] Train Extra: lr=0.0002598 inv=0.5098438 sub=0.0000000
17-03-25 13:37:18 [1] Step: 5000 Eval acc: 0.57586 0.82880 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 13:37:18 [1] Eval Extra: inv=0.4653269
17-03-25 13:37:18 [1] Checkpointing with new best dev accuracy of 0.575861
17-03-25 13:37:18 [1] Checkpointing.
17-03-25 13:38:32 [1] Step: 5100 Acc: 0.54125 0.83089 Cost: 1.37911 1.02012 0.23426 0.12473 Time: 0.00070
17-03-25 13:38:32 [1] Train Extra: lr=0.0002591 inv=0.4720313 sub=0.0000000
17-03-25 13:39:41 [1] Step: 5200 Acc: 0.54063 0.81876 Cost: 1.19172 0.84117 0.22704 0.12351 Time: 0.00066
17-03-25 13:39:41 [1] Train Extra: lr=0.0002583 inv=0.4775000 sub=0.0000000
17-03-25 13:41:05 [1] Step: 5300 Acc: 0.53500 0.82345 Cost: 1.41942 1.04003 0.25704 0.12235 Time: 0.00072
17-03-25 13:41:05 [1] Train Extra: lr=0.0002576 inv=0.5364062 sub=0.0000000
17-03-25 13:42:12 [1] Step: 5400 Acc: 0.54844 0.82811 Cost: 1.43800 1.00371 0.31306 0.12123 Time: 0.00068
17-03-25 13:42:12 [1] Train Extra: lr=0.0002568 inv=0.4640625 sub=0.0000000
17-03-25 13:43:29 [1] Step: 5500 Acc: 0.53469 0.82179 Cost: 1.58373 1.19146 0.27217 0.12010 Time: 0.00069
17-03-25 13:43:29 [1] Train Extra: lr=0.0002561 inv=0.4895312 sub=0.0000000
17-03-25 13:44:22 [1] Step: 5500 Eval acc: 0.58138 0.83132 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 13:44:22 [1] Eval Extra: inv=0.4731117
17-03-25 13:44:22 [1] Checkpointing with new best dev accuracy of 0.581383
17-03-25 13:45:47 [1] Step: 5600 Acc: 0.54438 0.82182 Cost: 1.19728 0.80368 0.27465 0.11895 Time: 0.00071
17-03-25 13:45:47 [1] Train Extra: lr=0.0002554 inv=0.5189063 sub=0.0000000
17-03-25 13:47:04 [1] Step: 5700 Acc: 0.54125 0.83132 Cost: 1.29913 0.98867 0.19260 0.11786 Time: 0.00070
17-03-25 13:47:04 [1] Train Extra: lr=0.0002546 inv=0.4851563 sub=0.0000000
17-03-25 13:48:25 [1] Step: 5800 Acc: 0.54281 0.82683 Cost: 1.51340 1.05847 0.33816 0.11677 Time: 0.00071
17-03-25 13:48:25 [1] Train Extra: lr=0.0002539 inv=0.5115625 sub=0.0000000
17-03-25 13:49:39 [1] Step: 5900 Acc: 0.53969 0.82252 Cost: 1.31163 1.07745 0.11846 0.11572 Time: 0.00069
17-03-25 13:49:39 [1] Train Extra: lr=0.0002532 inv=0.4742188 sub=0.0000000
17-03-25 13:50:58 [1] Step: 6000 Acc: 0.55750 0.82029 Cost: 1.26885 0.96800 0.18607 0.11478 Time: 0.00070
17-03-25 13:50:58 [1] Train Extra: lr=0.0002524 inv=0.5196875 sub=0.0000000
17-03-25 13:51:50 [1] Step: 6000 Eval acc: 0.57277 0.82847 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 13:51:50 [1] Eval Extra: inv=0.4921047
17-03-25 13:53:04 [1] Step: 6100 Acc: 0.55625 0.82875 Cost: 1.48307 1.08026 0.28900 0.11381 Time: 0.00069
17-03-25 13:53:04 [1] Train Extra: lr=0.0002517 inv=0.4790625 sub=0.0000000
17-03-25 13:54:12 [1] Step: 6200 Acc: 0.54906 0.82755 Cost: 1.26537 0.91958 0.23293 0.11285 Time: 0.00067
17-03-25 13:54:12 [1] Train Extra: lr=0.0002510 inv=0.4682812 sub=0.0000000
17-03-25 13:55:36 [1] Step: 6300 Acc: 0.55531 0.83015 Cost: 1.23105 0.89721 0.22187 0.11197 Time: 0.00073
17-03-25 13:55:36 [1] Train Extra: lr=0.0002503 inv=0.4721875 sub=0.0000000
17-03-25 13:56:45 [1] Step: 6400 Acc: 0.57500 0.83111 Cost: 1.27325 0.87487 0.28733 0.11106 Time: 0.00068
17-03-25 13:56:45 [1] Train Extra: lr=0.0002496 inv=0.4707812 sub=0.0000000
17-03-25 13:58:12 [1] Step: 6500 Acc: 0.55812 0.82992 Cost: 1.41617 1.08346 0.22257 0.11013 Time: 0.00073
17-03-25 13:58:12 [1] Train Extra: lr=0.0002488 inv=0.5106250 sub=0.0000000
17-03-25 13:59:03 [1] Step: 6500 Eval acc: 0.58392 0.83479 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 13:59:03 [1] Eval Extra: inv=0.4606890
17-03-25 14:00:19 [1] Step: 6600 Acc: 0.53844 0.82657 Cost: 1.33639 0.93708 0.29006 0.10925 Time: 0.00068
17-03-25 14:00:19 [1] Train Extra: lr=0.0002481 inv=0.4895312 sub=0.0000000
17-03-25 14:01:30 [1] Step: 6700 Acc: 0.56219 0.83188 Cost: 1.56817 1.18393 0.27578 0.10845 Time: 0.00070
17-03-25 14:01:30 [1] Train Extra: lr=0.0002474 inv=0.4667188 sub=0.0000000
17-03-25 14:02:46 [1] Step: 6800 Acc: 0.57406 0.82084 Cost: 1.28163 0.93774 0.23623 0.10765 Time: 0.00068
17-03-25 14:02:46 [1] Train Extra: lr=0.0002467 inv=0.5217188 sub=0.0000000
17-03-25 14:04:06 [1] Step: 6900 Acc: 0.55812 0.82445 Cost: 1.11439 0.73499 0.27265 0.10675 Time: 0.00070
17-03-25 14:04:06 [1] Train Extra: lr=0.0002460 inv=0.4726562 sub=0.0000000
17-03-25 14:05:20 [1] Step: 7000 Acc: 0.56875 0.82297 Cost: 1.15830 0.75822 0.29404 0.10604 Time: 0.00070
17-03-25 14:05:20 [1] Train Extra: lr=0.0002453 inv=0.4504687 sub=0.0000000
17-03-25 14:06:12 [1] Step: 7000 Eval acc: 0.59607 0.83462 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 14:06:12 [1] Eval Extra: inv=0.4767005
17-03-25 14:06:12 [1] Checkpointing with new best dev accuracy of 0.596069
17-03-25 14:07:32 [1] Step: 7100 Acc: 0.56688 0.83089 Cost: 0.91995 0.69856 0.11607 0.10532 Time: 0.00071
17-03-25 14:07:32 [1] Train Extra: lr=0.0002446 inv=0.4789062 sub=0.0000000
17-03-25 14:08:57 [1] Step: 7200 Acc: 0.55437 0.82526 Cost: 1.42132 0.95114 0.36559 0.10460 Time: 0.00072
17-03-25 14:08:57 [1] Train Extra: lr=0.0002439 inv=0.4917187 sub=0.0000000
17-03-25 14:10:05 [1] Step: 7300 Acc: 0.56250 0.82608 Cost: 1.40106 1.06529 0.23189 0.10388 Time: 0.00067
17-03-25 14:10:05 [1] Train Extra: lr=0.0002432 inv=0.4296875 sub=0.0000000
17-03-25 14:11:31 [1] Step: 7400 Acc: 0.57656 0.84186 Cost: 1.39486 1.00127 0.29035 0.10325 Time: 0.00074
17-03-25 14:11:31 [1] Train Extra: lr=0.0002425 inv=0.4718750 sub=0.0000000
17-03-25 14:12:41 [1] Step: 7500 Acc: 0.56969 0.82966 Cost: 0.93702 0.70615 0.12813 0.10274 Time: 0.00066
17-03-25 14:12:41 [1] Train Extra: lr=0.0002418 inv=0.4593750 sub=0.0000000
17-03-25 14:13:36 [1] Step: 7500 Eval acc: 0.59000 0.83091 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 14:13:36 [1] Eval Extra: inv=0.5249006
17-03-25 14:14:46 [1] Step: 7600 Acc: 0.56563 0.82514 Cost: 1.07348 0.75685 0.21447 0.10216 Time: 0.00067
17-03-25 14:14:46 [1] Train Extra: lr=0.0002411 inv=0.4562500 sub=0.0000000
17-03-25 14:16:10 [1] Step: 7700 Acc: 0.57531 0.82819 Cost: 1.19868 0.79518 0.30203 0.10147 Time: 0.00074
17-03-25 14:16:10 [1] Train Extra: lr=0.0002404 inv=0.4737500 sub=0.0000000
17-03-25 14:17:32 [1] Step: 7800 Acc: 0.57125 0.82875 Cost: 1.43753 1.06581 0.27078 0.10093 Time: 0.00070
17-03-25 14:17:32 [1] Train Extra: lr=0.0002397 inv=0.4689063 sub=0.0000000
17-03-25 14:18:52 [1] Step: 7900 Acc: 0.56531 0.82769 Cost: 1.20441 0.86838 0.23573 0.10030 Time: 0.00070
17-03-25 14:18:52 [1] Train Extra: lr=0.0002390 inv=0.4743750 sub=0.0000000
17-03-25 14:20:05 [1] Step: 8000 Acc: 0.57250 0.83001 Cost: 1.20348 0.81902 0.28472 0.09974 Time: 0.00068
17-03-25 14:20:05 [1] Train Extra: lr=0.0002383 inv=0.4612500 sub=0.0000000
17-03-25 14:20:56 [1] Step: 8000 Eval acc: 0.60071 0.83716 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 14:20:56 [1] Eval Extra: inv=0.4272306
17-03-25 14:20:56 [1] Checkpointing with new best dev accuracy of 0.600707
17-03-25 14:22:12 [1] Step: 8100 Acc: 0.57125 0.82654 Cost: 1.24109 0.99324 0.14869 0.09916 Time: 0.00067
17-03-25 14:22:12 [1] Train Extra: lr=0.0002376 inv=0.4381250 sub=0.0000000
17-03-25 14:23:30 [1] Step: 8200 Acc: 0.56781 0.82704 Cost: 1.28042 0.89977 0.28199 0.09866 Time: 0.00071
17-03-25 14:23:30 [1] Train Extra: lr=0.0002370 inv=0.4596875 sub=0.0000000
17-03-25 14:24:58 [1] Step: 8300 Acc: 0.57594 0.83302 Cost: 0.98129 0.73582 0.14726 0.09820 Time: 0.00073
17-03-25 14:24:58 [1] Train Extra: lr=0.0002363 inv=0.4573437 sub=0.0000000
17-03-25 14:26:26 [1] Step: 8400 Acc: 0.57531 0.83755 Cost: 0.96195 0.77773 0.08657 0.09765 Time: 0.00076
17-03-25 14:26:26 [1] Train Extra: lr=0.0002356 inv=0.4212500 sub=0.0000000
17-03-25 14:27:42 [1] Step: 8500 Acc: 0.58344 0.83287 Cost: 1.12668 0.84423 0.18532 0.09714 Time: 0.00070
17-03-25 14:27:42 [1] Train Extra: lr=0.0002349 inv=0.4382813 sub=0.0000000
17-03-25 14:28:34 [1] Step: 8500 Eval acc: 0.60049 0.83938 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 14:28:34 [1] Eval Extra: inv=0.4383834
17-03-25 14:30:01 [1] Step: 8600 Acc: 0.56469 0.83629 Cost: 1.20064 0.86644 0.23750 0.09669 Time: 0.00074
17-03-25 14:30:01 [1] Train Extra: lr=0.0002342 inv=0.4257812 sub=0.0000000
17-03-25 14:31:13 [1] Step: 8700 Acc: 0.56375 0.82893 Cost: 1.30507 0.98324 0.22562 0.09621 Time: 0.00068
17-03-25 14:31:13 [1] Train Extra: lr=0.0002336 inv=0.4700000 sub=0.0000000
17-03-25 14:32:34 [1] Step: 8800 Acc: 0.57281 0.83086 Cost: 1.18475 0.84177 0.24725 0.09573 Time: 0.00071
17-03-25 14:32:34 [1] Train Extra: lr=0.0002329 inv=0.4390625 sub=0.0000000
17-03-25 14:33:43 [1] Step: 8900 Acc: 0.58594 0.82617 Cost: 1.32555 0.94779 0.28243 0.09533 Time: 0.00067
17-03-25 14:33:43 [1] Train Extra: lr=0.0002322 inv=0.4560938 sub=0.0000000
17-03-25 14:34:57 [1] Step: 9000 Acc: 0.59750 0.82338 Cost: 1.15815 0.82514 0.23809 0.09492 Time: 0.00069
17-03-25 14:34:57 [1] Train Extra: lr=0.0002316 inv=0.4601562 sub=0.0000000
17-03-25 14:35:48 [1] Step: 9000 Eval acc: 0.60512 0.83511 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 14:35:48 [1] Eval Extra: inv=0.4458370
17-03-25 14:35:48 [1] Checkpointing with new best dev accuracy of 0.605124
17-03-25 14:37:04 [1] Step: 9100 Acc: 0.58531 0.83253 Cost: 1.13969 0.90025 0.14498 0.09446 Time: 0.00069
17-03-25 14:37:04 [1] Train Extra: lr=0.0002309 inv=0.4457813 sub=0.0000000
17-03-25 14:38:28 [1] Step: 9200 Acc: 0.56812 0.84133 Cost: 0.87891 0.68839 0.09655 0.09397 Time: 0.00075
17-03-25 14:38:28 [1] Train Extra: lr=0.0002302 inv=0.4429688 sub=0.0000000
17-03-25 14:39:44 [1] Step: 9300 Acc: 0.59156 0.83117 Cost: 1.21511 0.90635 0.21522 0.09354 Time: 0.00070
17-03-25 14:39:44 [1] Train Extra: lr=0.0002296 inv=0.4328125 sub=0.0000000
17-03-25 14:41:04 [1] Step: 9400 Acc: 0.57906 0.84077 Cost: 1.20663 0.90392 0.20947 0.09324 Time: 0.00072
17-03-25 14:41:04 [1] Train Extra: lr=0.0002289 inv=0.4365625 sub=0.0000000
17-03-25 14:42:24 [1] Step: 9500 Acc: 0.58406 0.83332 Cost: 1.26765 0.99507 0.17976 0.09282 Time: 0.00072
17-03-25 14:42:24 [1] Train Extra: lr=0.0002283 inv=0.4159375 sub=0.0000000
17-03-25 14:43:15 [1] Step: 9500 Eval acc: 0.60645 0.84093 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 14:43:15 [1] Eval Extra: inv=0.4302672
17-03-25 14:44:27 [1] Step: 9600 Acc: 0.58437 0.83004 Cost: 1.25321 0.91010 0.25068 0.09244 Time: 0.00068
17-03-25 14:44:27 [1] Train Extra: lr=0.0002276 inv=0.4234375 sub=0.0000000
17-03-25 14:45:40 [1] Step: 9700 Acc: 0.57531 0.82806 Cost: 1.14660 0.77704 0.27749 0.09208 Time: 0.00068
17-03-25 14:45:40 [1] Train Extra: lr=0.0002270 inv=0.4234375 sub=0.0000000
17-03-25 14:47:00 [1] Step: 9800 Acc: 0.58656 0.83704 Cost: 1.35758 0.99430 0.27152 0.09176 Time: 0.00071
17-03-25 14:47:00 [1] Train Extra: lr=0.0002263 inv=0.4193750 sub=0.0000000
17-03-25 14:48:15 [1] Step: 9900 Acc: 0.58375 0.82916 Cost: 1.17797 0.83512 0.25136 0.09148 Time: 0.00068
17-03-25 14:48:15 [1] Train Extra: lr=0.0002256 inv=0.4395312 sub=0.0000000
17-03-25 14:49:28 [1] Step: 10000 Acc: 0.58031 0.83275 Cost: 1.09271 0.86589 0.13566 0.09116 Time: 0.00069
17-03-25 14:49:28 [1] Train Extra: lr=0.0002250 inv=0.4281250 sub=0.0000000
17-03-25 14:50:20 [1] Step: 10000 Eval acc: 0.60976 0.83074 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 14:50:20 [1] Eval Extra: inv=0.4532906
17-03-25 14:50:20 [1] Checkpointing with new best dev accuracy of 0.609761
17-03-25 14:50:20 [1] Checkpointing.
17-03-25 14:51:48 [1] Step: 10100 Acc: 0.57375 0.83282 Cost: 1.34719 0.95563 0.30079 0.09077 Time: 0.00071
17-03-25 14:51:48 [1] Train Extra: lr=0.0002244 inv=0.4709375 sub=0.0000000
17-03-25 14:53:00 [1] Step: 10200 Acc: 0.59906 0.83190 Cost: 1.41283 1.06191 0.26041 0.09051 Time: 0.00070
17-03-25 14:53:00 [1] Train Extra: lr=0.0002237 inv=0.4237500 sub=0.0000000
17-03-25 14:54:20 [1] Step: 10300 Acc: 0.57594 0.83108 Cost: 1.49033 1.04147 0.35868 0.09018 Time: 0.00069
17-03-25 14:54:20 [1] Train Extra: lr=0.0002231 inv=0.4387500 sub=0.0000000
17-03-25 14:55:35 [1] Step: 10400 Acc: 0.59000 0.82648 Cost: 1.32482 0.94919 0.28588 0.08975 Time: 0.00068
17-03-25 14:55:35 [1] Train Extra: lr=0.0002224 inv=0.4426562 sub=0.0000000
17-03-25 14:56:45 [1] Step: 10500 Acc: 0.58906 0.83865 Cost: 1.18514 0.87316 0.22256 0.08941 Time: 0.00071
17-03-25 14:56:45 [1] Train Extra: lr=0.0002218 inv=0.4037500 sub=0.0000000
17-03-25 14:57:34 [1] Step: 10500 Eval acc: 0.60877 0.84358 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00014
17-03-25 14:57:34 [1] Eval Extra: inv=0.3812390
17-03-25 14:59:03 [1] Step: 10600 Acc: 0.57437 0.84170 Cost: 1.32841 0.95266 0.28675 0.08900 Time: 0.00076
17-03-25 14:59:03 [1] Train Extra: lr=0.0002211 inv=0.4220313 sub=0.0000000
17-03-25 15:00:17 [1] Step: 10700 Acc: 0.57750 0.82476 Cost: 1.16177 0.85485 0.21816 0.08877 Time: 0.00067
17-03-25 15:00:17 [1] Train Extra: lr=0.0002205 inv=0.4712500 sub=0.0000000
17-03-25 15:01:32 [1] Step: 10800 Acc: 0.58156 0.83272 Cost: 1.45574 0.98032 0.38695 0.08846 Time: 0.00068
17-03-25 15:01:32 [1] Train Extra: lr=0.0002199 inv=0.4371875 sub=0.0000000
17-03-25 15:02:37 [1] Step: 10900 Acc: 0.59250 0.83309 Cost: 1.12985 0.85524 0.18652 0.08810 Time: 0.00067
17-03-25 15:02:37 [1] Train Extra: lr=0.0002192 inv=0.3979687 sub=0.0000000
17-03-25 15:03:55 [1] Step: 11000 Acc: 0.58000 0.82674 Cost: 1.21822 0.86807 0.26236 0.08780 Time: 0.00068
17-03-25 15:03:55 [1] Train Extra: lr=0.0002186 inv=0.4179688 sub=0.0000000
17-03-25 15:04:47 [1] Step: 11000 Eval acc: 0.61871 0.83488 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 15:04:47 [1] Eval Extra: inv=0.4191696
17-03-25 15:04:47 [1] Checkpointing with new best dev accuracy of 0.618706
17-03-25 15:06:07 [1] Step: 11100 Acc: 0.59062 0.83795 Cost: 1.09128 0.75738 0.24635 0.08755 Time: 0.00073
17-03-25 15:06:07 [1] Train Extra: lr=0.0002180 inv=0.4214062 sub=0.0000000
17-03-25 15:07:21 [1] Step: 11200 Acc: 0.57250 0.83281 Cost: 1.42421 1.10584 0.23112 0.08725 Time: 0.00068
17-03-25 15:07:21 [1] Train Extra: lr=0.0002174 inv=0.4193750 sub=0.0000000
17-03-25 15:08:29 [1] Step: 11300 Acc: 0.59437 0.83095 Cost: 1.12518 0.76563 0.27251 0.08704 Time: 0.00067
17-03-25 15:08:29 [1] Train Extra: lr=0.0002167 inv=0.4028125 sub=0.0000000
17-03-25 15:09:41 [1] Step: 11400 Acc: 0.59469 0.83378 Cost: 1.29369 0.92014 0.28685 0.08670 Time: 0.00069
17-03-25 15:09:41 [1] Train Extra: lr=0.0002161 inv=0.4193750 sub=0.0000000
17-03-25 15:11:07 [1] Step: 11500 Acc: 0.58562 0.84260 Cost: 1.29932 0.89230 0.32058 0.08644 Time: 0.00073
17-03-25 15:11:07 [1] Train Extra: lr=0.0002155 inv=0.4259375 sub=0.0000000
17-03-25 15:11:58 [1] Step: 11500 Eval acc: 0.61484 0.84091 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 15:11:58 [1] Eval Extra: inv=0.4279483
17-03-25 15:13:06 [1] Step: 11600 Acc: 0.60562 0.83800 Cost: 1.16347 0.81654 0.26073 0.08621 Time: 0.00068
17-03-25 15:13:06 [1] Train Extra: lr=0.0002149 inv=0.3943750 sub=0.0000000
17-03-25 15:14:22 [1] Step: 11700 Acc: 0.59156 0.82888 Cost: 1.24191 0.87557 0.28031 0.08603 Time: 0.00068
17-03-25 15:14:22 [1] Train Extra: lr=0.0002143 inv=0.4315625 sub=0.0000000
17-03-25 15:15:41 [1] Step: 11800 Acc: 0.58813 0.83109 Cost: 1.21826 1.00518 0.12724 0.08584 Time: 0.00070
17-03-25 15:15:41 [1] Train Extra: lr=0.0002136 inv=0.4484375 sub=0.0000000
17-03-25 15:17:07 [1] Step: 11900 Acc: 0.58844 0.84062 Cost: 1.06757 0.83167 0.15037 0.08552 Time: 0.00073
17-03-25 15:17:07 [1] Train Extra: lr=0.0002130 inv=0.4379688 sub=0.0000000
17-03-25 15:18:20 [1] Step: 12000 Acc: 0.57875 0.83226 Cost: 1.14466 0.87095 0.18839 0.08532 Time: 0.00068
17-03-25 15:18:20 [1] Train Extra: lr=0.0002124 inv=0.4118750 sub=0.0000000
17-03-25 15:19:09 [1] Step: 12000 Eval acc: 0.62025 0.84254 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00014
17-03-25 15:19:09 [1] Eval Extra: inv=0.3885822
17-03-25 15:20:21 [1] Step: 12100 Acc: 0.60469 0.82920 Cost: 1.12867 0.90348 0.14005 0.08514 Time: 0.00067
17-03-25 15:20:21 [1] Train Extra: lr=0.0002118 inv=0.4364062 sub=0.0000000
17-03-25 15:21:42 [1] Step: 12200 Acc: 0.59594 0.82837 Cost: 1.06250 0.87622 0.10137 0.08490 Time: 0.00070
17-03-25 15:21:42 [1] Train Extra: lr=0.0002112 inv=0.4628125 sub=0.0000000
17-03-25 15:22:51 [1] Step: 12300 Acc: 0.61594 0.84394 Cost: 1.12950 0.83380 0.21096 0.08474 Time: 0.00069
17-03-25 15:22:51 [1] Train Extra: lr=0.0002106 inv=0.4075000 sub=0.0000000
17-03-25 15:24:10 [1] Step: 12400 Acc: 0.58437 0.83830 Cost: 1.17165 0.87296 0.21412 0.08457 Time: 0.00072
17-03-25 15:24:10 [1] Train Extra: lr=0.0002100 inv=0.4028125 sub=0.0000000
17-03-25 15:25:25 [1] Step: 12500 Acc: 0.57969 0.83277 Cost: 1.06882 0.78233 0.20221 0.08428 Time: 0.00068
17-03-25 15:25:25 [1] Train Extra: lr=0.0002094 inv=0.4315625 sub=0.0000000
17-03-25 15:26:16 [1] Step: 12500 Eval acc: 0.61440 0.84210 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 15:26:16 [1] Eval Extra: inv=0.4304329
17-03-25 15:27:34 [1] Step: 12600 Acc: 0.59437 0.83696 Cost: 1.22966 0.97511 0.17045 0.08410 Time: 0.00072
17-03-25 15:27:34 [1] Train Extra: lr=0.0002088 inv=0.4064063 sub=0.0000000
17-03-25 15:28:47 [1] Step: 12700 Acc: 0.59813 0.83535 Cost: 0.85378 0.61798 0.15197 0.08383 Time: 0.00068
17-03-25 15:28:47 [1] Train Extra: lr=0.0002082 inv=0.4120313 sub=0.0000000
17-03-25 15:30:06 [1] Step: 12800 Acc: 0.59656 0.84152 Cost: 1.32792 0.86621 0.37808 0.08362 Time: 0.00072
17-03-25 15:30:06 [1] Train Extra: lr=0.0002076 inv=0.4170313 sub=0.0000000
17-03-25 15:31:20 [1] Step: 12900 Acc: 0.59906 0.83175 Cost: 1.11603 0.80512 0.22746 0.08345 Time: 0.00068
17-03-25 15:31:20 [1] Train Extra: lr=0.0002070 inv=0.4257812 sub=0.0000000
17-03-25 15:32:38 [1] Step: 13000 Acc: 0.61031 0.84778 Cost: 1.31657 0.98797 0.24532 0.08328 Time: 0.00071
17-03-25 15:32:38 [1] Train Extra: lr=0.0002064 inv=0.3970313 sub=0.0000000
17-03-25 15:33:32 [1] Step: 13000 Eval acc: 0.62114 0.83131 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 15:33:32 [1] Eval Extra: inv=0.4674801
17-03-25 15:34:46 [1] Step: 13100 Acc: 0.59531 0.83543 Cost: 1.03306 0.83837 0.11164 0.08305 Time: 0.00068
17-03-25 15:34:46 [1] Train Extra: lr=0.0002058 inv=0.4292187 sub=0.0000000
17-03-25 15:36:00 [1] Step: 13200 Acc: 0.60469 0.83635 Cost: 1.22594 0.97516 0.16790 0.08288 Time: 0.00069
17-03-25 15:36:00 [1] Train Extra: lr=0.0002052 inv=0.4082812 sub=0.0000000
17-03-25 15:37:27 [1] Step: 13300 Acc: 0.59313 0.83587 Cost: 1.19738 0.83430 0.28042 0.08266 Time: 0.00072
17-03-25 15:37:27 [1] Train Extra: lr=0.0002046 inv=0.4542188 sub=0.0000000
17-03-25 15:38:55 [1] Step: 13400 Acc: 0.59031 0.84141 Cost: 1.06632 0.78383 0.19997 0.08252 Time: 0.00072
17-03-25 15:38:55 [1] Train Extra: lr=0.0002040 inv=0.4350000 sub=0.0000000
17-03-25 15:40:06 [1] Step: 13500 Acc: 0.61313 0.84208 Cost: 1.15747 0.94083 0.13425 0.08240 Time: 0.00071
17-03-25 15:40:06 [1] Train Extra: lr=0.0002034 inv=0.3942188 sub=0.0000000
17-03-25 15:40:58 [1] Step: 13500 Eval acc: 0.62268 0.83859 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 15:40:58 [1] Eval Extra: inv=0.4188936
17-03-25 15:40:58 [1] Checkpointing with new best dev accuracy of 0.622681
17-03-25 15:42:17 [1] Step: 13600 Acc: 0.60094 0.83286 Cost: 0.99493 0.65318 0.25957 0.08218 Time: 0.00069
17-03-25 15:42:17 [1] Train Extra: lr=0.0002029 inv=0.4410938 sub=0.0000000
17-03-25 15:43:30 [1] Step: 13700 Acc: 0.60062 0.82808 Cost: 1.28107 0.96744 0.23168 0.08195 Time: 0.00066
17-03-25 15:43:30 [1] Train Extra: lr=0.0002023 inv=0.4420312 sub=0.0000000
17-03-25 15:44:45 [1] Step: 13800 Acc: 0.59156 0.83504 Cost: 1.28758 0.94652 0.25929 0.08177 Time: 0.00068
17-03-25 15:44:45 [1] Train Extra: lr=0.0002017 inv=0.4056250 sub=0.0000000
17-03-25 15:46:11 [1] Step: 13900 Acc: 0.59469 0.83635 Cost: 1.30355 0.96019 0.26177 0.08159 Time: 0.00071
17-03-25 15:46:11 [1] Train Extra: lr=0.0002011 inv=0.4185937 sub=0.0000000
17-03-25 15:47:17 [1] Step: 14000 Acc: 0.61062 0.83814 Cost: 0.98005 0.72649 0.17207 0.08149 Time: 0.00067
17-03-25 15:47:17 [1] Train Extra: lr=0.0002005 inv=0.3948437 sub=0.0000000
17-03-25 15:48:08 [1] Step: 14000 Eval acc: 0.61760 0.84014 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 15:48:08 [1] Eval Extra: inv=0.4321996
17-03-25 15:49:23 [1] Step: 14100 Acc: 0.60250 0.83605 Cost: 1.27420 0.92950 0.26340 0.08130 Time: 0.00068
17-03-25 15:49:23 [1] Train Extra: lr=0.0002000 inv=0.4103125 sub=0.0000000
17-03-25 15:50:42 [1] Step: 14200 Acc: 0.60531 0.83741 Cost: 1.14838 0.81278 0.25445 0.08115 Time: 0.00071
17-03-25 15:50:42 [1] Train Extra: lr=0.0001994 inv=0.4232812 sub=0.0000000
17-03-25 15:52:00 [1] Step: 14300 Acc: 0.60406 0.83502 Cost: 1.18411 0.81691 0.28629 0.08091 Time: 0.00071
17-03-25 15:52:00 [1] Train Extra: lr=0.0001988 inv=0.4135937 sub=0.0000000
17-03-25 15:53:15 [1] Step: 14400 Acc: 0.60969 0.83606 Cost: 1.07576 0.78732 0.20765 0.08079 Time: 0.00068
17-03-25 15:53:15 [1] Train Extra: lr=0.0001982 inv=0.4226563 sub=0.0000000
17-03-25 15:54:29 [1] Step: 14500 Acc: 0.59844 0.84006 Cost: 1.03082 0.79179 0.15837 0.08065 Time: 0.00069
17-03-25 15:54:29 [1] Train Extra: lr=0.0001977 inv=0.3992188 sub=0.0000000
17-03-25 15:55:20 [1] Step: 14500 Eval acc: 0.62688 0.84472 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 15:55:20 [1] Eval Extra: inv=0.3791409
17-03-25 15:55:20 [1] Checkpointing with new best dev accuracy of 0.626877
17-03-25 15:56:33 [1] Step: 14600 Acc: 0.60500 0.82692 Cost: 1.15464 0.85432 0.21971 0.08061 Time: 0.00067
17-03-25 15:56:33 [1] Train Extra: lr=0.0001971 inv=0.4229688 sub=0.0000000
17-03-25 15:57:55 [1] Step: 14700 Acc: 0.58156 0.83207 Cost: 1.12778 0.83496 0.21234 0.08048 Time: 0.00070
17-03-25 15:57:55 [1] Train Extra: lr=0.0001965 inv=0.4179688 sub=0.0000000
17-03-25 15:59:06 [1] Step: 14800 Acc: 0.60688 0.83631 Cost: 1.14292 0.85910 0.20347 0.08035 Time: 0.00070
17-03-25 15:59:06 [1] Train Extra: lr=0.0001960 inv=0.3890625 sub=0.0000000
17-03-25 16:00:20 [1] Step: 14900 Acc: 0.59062 0.84205 Cost: 1.26698 0.91245 0.27428 0.08025 Time: 0.00068
17-03-25 16:00:20 [1] Train Extra: lr=0.0001954 inv=0.4007812 sub=0.0000000
17-03-25 16:01:45 [1] Step: 15000 Acc: 0.60750 0.83982 Cost: 1.25532 0.96422 0.21103 0.08006 Time: 0.00073
17-03-25 16:01:45 [1] Train Extra: lr=0.0001949 inv=0.4081250 sub=0.0000000
17-03-25 16:02:37 [1] Step: 15000 Eval acc: 0.62500 0.83795 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 16:02:37 [1] Eval Extra: inv=0.4710689
17-03-25 16:02:37 [1] Checkpointing.
17-03-25 16:03:54 [1] Step: 15100 Acc: 0.60281 0.83722 Cost: 1.43477 1.07628 0.27860 0.07989 Time: 0.00068
17-03-25 16:03:54 [1] Train Extra: lr=0.0001943 inv=0.4307813 sub=0.0000000
17-03-25 16:05:12 [1] Step: 15200 Acc: 0.59188 0.83424 Cost: 1.15102 0.89505 0.17630 0.07968 Time: 0.00072
17-03-25 16:05:12 [1] Train Extra: lr=0.0001937 inv=0.4156250 sub=0.0000000
17-03-25 16:06:28 [1] Step: 15300 Acc: 0.58625 0.83922 Cost: 1.43235 1.07495 0.27791 0.07950 Time: 0.00068
17-03-25 16:06:28 [1] Train Extra: lr=0.0001932 inv=0.4457813 sub=0.0000000
17-03-25 16:07:42 [1] Step: 15400 Acc: 0.61719 0.83856 Cost: 1.13355 0.78342 0.27067 0.07946 Time: 0.00068
17-03-25 16:07:42 [1] Train Extra: lr=0.0001926 inv=0.4237500 sub=0.0000000
17-03-25 16:08:59 [1] Step: 15500 Acc: 0.59813 0.84347 Cost: 0.92009 0.70300 0.13776 0.07933 Time: 0.00072
17-03-25 16:08:59 [1] Train Extra: lr=0.0001921 inv=0.4001562 sub=0.0000000
17-03-25 16:09:51 [1] Step: 15500 Eval acc: 0.62544 0.84485 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 16:09:51 [1] Eval Extra: inv=0.3725707
17-03-25 16:11:04 [1] Step: 15600 Acc: 0.60969 0.83254 Cost: 1.18595 0.88727 0.21948 0.07920 Time: 0.00067
17-03-25 16:11:04 [1] Train Extra: lr=0.0001915 inv=0.4392187 sub=0.0000000
17-03-25 16:12:13 [1] Step: 15700 Acc: 0.61531 0.83120 Cost: 1.20270 0.88721 0.23646 0.07903 Time: 0.00065
17-03-25 16:12:13 [1] Train Extra: lr=0.0001910 inv=0.4359375 sub=0.0000000
17-03-25 16:13:39 [1] Step: 15800 Acc: 0.58594 0.84376 Cost: 0.96134 0.77767 0.10477 0.07891 Time: 0.00073
17-03-25 16:13:39 [1] Train Extra: lr=0.0001904 inv=0.4039063 sub=0.0000000
17-03-25 16:14:58 [1] Step: 15900 Acc: 0.61313 0.83731 Cost: 1.26342 0.90049 0.28419 0.07875 Time: 0.00070
17-03-25 16:14:58 [1] Train Extra: lr=0.0001899 inv=0.4140625 sub=0.0000000
17-03-25 16:16:17 [1] Step: 16000 Acc: 0.60156 0.83102 Cost: 1.08575 0.75716 0.24994 0.07865 Time: 0.00068
17-03-25 16:16:17 [1] Train Extra: lr=0.0001893 inv=0.4306250 sub=0.0000000
17-03-25 16:17:09 [1] Step: 16000 Eval acc: 0.62114 0.84406 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 16:17:09 [1] Eval Extra: inv=0.3913980
17-03-25 16:18:28 [1] Step: 16100 Acc: 0.60813 0.83524 Cost: 1.27643 0.90582 0.29206 0.07856 Time: 0.00070
17-03-25 16:18:28 [1] Train Extra: lr=0.0001888 inv=0.4276563 sub=0.0000000
17-03-25 16:19:40 [1] Step: 16200 Acc: 0.60625 0.83632 Cost: 1.24308 0.91166 0.25289 0.07854 Time: 0.00068
17-03-25 16:19:40 [1] Train Extra: lr=0.0001882 inv=0.4193750 sub=0.0000000
17-03-25 16:20:55 [1] Step: 16300 Acc: 0.62469 0.82921 Cost: 0.91452 0.64540 0.19066 0.07846 Time: 0.00067
17-03-25 16:20:55 [1] Train Extra: lr=0.0001877 inv=0.4267187 sub=0.0000000
17-03-25 16:22:15 [1] Step: 16400 Acc: 0.58594 0.84182 Cost: 1.26027 0.93639 0.24548 0.07841 Time: 0.00073
17-03-25 16:22:15 [1] Train Extra: lr=0.0001872 inv=0.3865625 sub=0.0000000
17-03-25 16:23:28 [1] Step: 16500 Acc: 0.60562 0.83616 Cost: 1.13747 0.91647 0.14279 0.07821 Time: 0.00069
17-03-25 16:23:28 [1] Train Extra: lr=0.0001866 inv=0.4026562 sub=0.0000000
17-03-25 16:24:19 [1] Step: 16500 Eval acc: 0.62511 0.83945 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 16:24:19 [1] Eval Extra: inv=0.3864289
17-03-25 16:25:38 [1] Step: 16600 Acc: 0.62844 0.84045 Cost: 1.08830 0.86515 0.14505 0.07810 Time: 0.00071
17-03-25 16:25:38 [1] Train Extra: lr=0.0001861 inv=0.4012500 sub=0.0000000
17-03-25 16:26:58 [1] Step: 16700 Acc: 0.61687 0.84827 Cost: 1.52178 1.09599 0.34778 0.07801 Time: 0.00073
17-03-25 16:26:58 [1] Train Extra: lr=0.0001856 inv=0.4017188 sub=0.0000000
17-03-25 16:28:21 [1] Step: 16800 Acc: 0.61375 0.84152 Cost: 1.08395 0.75745 0.24857 0.07792 Time: 0.00072
17-03-25 16:28:21 [1] Train Extra: lr=0.0001850 inv=0.4618750 sub=0.0000000
17-03-25 16:29:53 [1] Step: 16900 Acc: 0.61562 0.83959 Cost: 0.85851 0.64273 0.13783 0.07795 Time: 0.00074
17-03-25 16:29:53 [1] Train Extra: lr=0.0001845 inv=0.4531250 sub=0.0000000
17-03-25 16:31:07 [1] Step: 17000 Acc: 0.62313 0.84116 Cost: 1.25287 0.85486 0.32009 0.07792 Time: 0.00069
17-03-25 16:31:07 [1] Train Extra: lr=0.0001840 inv=0.4193750 sub=0.0000000
17-03-25 16:31:59 [1] Step: 17000 Eval acc: 0.63571 0.84612 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 16:31:59 [1] Eval Extra: inv=0.4496466
17-03-25 16:31:59 [1] Checkpointing with new best dev accuracy of 0.635711
17-03-25 16:33:19 [1] Step: 17100 Acc: 0.61906 0.84137 Cost: 1.51941 1.08453 0.35699 0.07789 Time: 0.00071
17-03-25 16:33:19 [1] Train Extra: lr=0.0001834 inv=0.4226563 sub=0.0000000
17-03-25 16:34:32 [1] Step: 17200 Acc: 0.61844 0.84038 Cost: 1.00088 0.74989 0.17302 0.07798 Time: 0.00070
17-03-25 16:34:32 [1] Train Extra: lr=0.0001829 inv=0.4068750 sub=0.0000000
17-03-25 16:35:45 [1] Step: 17300 Acc: 0.62969 0.84025 Cost: 1.44709 1.10277 0.26628 0.07804 Time: 0.00068
17-03-25 16:35:45 [1] Train Extra: lr=0.0001824 inv=0.4368750 sub=0.0000000
17-03-25 16:37:00 [1] Step: 17400 Acc: 0.61875 0.83989 Cost: 1.35758 1.00691 0.27265 0.07802 Time: 0.00068
17-03-25 16:37:00 [1] Train Extra: lr=0.0001819 inv=0.3996875 sub=0.0000000
17-03-25 16:38:24 [1] Step: 17500 Acc: 0.62031 0.84282 Cost: 1.49083 1.07693 0.33580 0.07810 Time: 0.00073
17-03-25 16:38:24 [1] Train Extra: lr=0.0001813 inv=0.4292187 sub=0.0000000
17-03-25 16:39:16 [1] Step: 17500 Eval acc: 0.63052 0.84473 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 16:39:16 [1] Eval Extra: inv=0.3111749
17-03-25 16:40:29 [1] Step: 17600 Acc: 0.61250 0.84457 Cost: 1.34880 0.99519 0.27559 0.07801 Time: 0.00071
17-03-25 16:40:29 [1] Train Extra: lr=0.0001808 inv=0.3790625 sub=0.0000000
17-03-25 16:41:45 [1] Step: 17700 Acc: 0.61156 0.83306 Cost: 0.89540 0.66363 0.15388 0.07788 Time: 0.00068
17-03-25 16:41:45 [1] Train Extra: lr=0.0001803 inv=0.4287500 sub=0.0000000
17-03-25 16:43:05 [1] Step: 17800 Acc: 0.62219 0.84685 Cost: 1.41800 1.08747 0.25263 0.07791 Time: 0.00072
17-03-25 16:43:05 [1] Train Extra: lr=0.0001798 inv=0.4132812 sub=0.0000000
17-03-25 16:44:25 [1] Step: 17900 Acc: 0.61250 0.84174 Cost: 0.99743 0.70309 0.21645 0.07790 Time: 0.00071
17-03-25 16:44:25 [1] Train Extra: lr=0.0001793 inv=0.4095313 sub=0.0000000
17-03-25 16:45:52 [1] Step: 18000 Acc: 0.62906 0.83865 Cost: 0.89924 0.60101 0.22032 0.07791 Time: 0.00072
17-03-25 16:45:52 [1] Train Extra: lr=0.0001787 inv=0.4465625 sub=0.0000000
17-03-25 16:46:44 [1] Step: 18000 Eval acc: 0.63748 0.84785 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 16:46:44 [1] Eval Extra: inv=0.4236418
17-03-25 16:47:52 [1] Step: 18100 Acc: 0.63375 0.84133 Cost: 0.95782 0.74121 0.13874 0.07788 Time: 0.00068
17-03-25 16:47:52 [1] Train Extra: lr=0.0001782 inv=0.3753125 sub=0.0000000
17-03-25 16:49:17 [1] Step: 18200 Acc: 0.62562 0.83900 Cost: 1.28377 0.97372 0.23220 0.07784 Time: 0.00071
17-03-25 16:49:17 [1] Train Extra: lr=0.0001777 inv=0.4700000 sub=0.0000000
17-03-25 16:50:44 [1] Step: 18300 Acc: 0.61406 0.84691 Cost: 1.19305 0.92772 0.18758 0.07775 Time: 0.00074
17-03-25 16:50:44 [1] Train Extra: lr=0.0001772 inv=0.4078125 sub=0.0000000
17-03-25 16:52:02 [1] Step: 18400 Acc: 0.62656 0.84233 Cost: 1.25799 0.87329 0.30702 0.07769 Time: 0.00071
17-03-25 16:52:02 [1] Train Extra: lr=0.0001767 inv=0.4060937 sub=0.0000000
17-03-25 16:53:18 [1] Step: 18500 Acc: 0.62250 0.84979 Cost: 0.97314 0.72886 0.16660 0.07768 Time: 0.00070
17-03-25 16:53:18 [1] Train Extra: lr=0.0001762 inv=0.3976562 sub=0.0000000
17-03-25 16:54:09 [1] Step: 18500 Eval acc: 0.63328 0.83935 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 16:54:09 [1] Eval Extra: inv=0.4201634
17-03-25 16:55:27 [1] Step: 18600 Acc: 0.62906 0.84365 Cost: 1.22712 0.93469 0.21470 0.07772 Time: 0.00072
17-03-25 16:55:27 [1] Train Extra: lr=0.0001757 inv=0.4051562 sub=0.0000000
17-03-25 16:56:41 [1] Step: 18700 Acc: 0.61750 0.83664 Cost: 0.95138 0.77616 0.09744 0.07778 Time: 0.00069
17-03-25 16:56:41 [1] Train Extra: lr=0.0001752 inv=0.4120313 sub=0.0000000
17-03-25 16:58:01 [1] Step: 18800 Acc: 0.63313 0.84039 Cost: 1.11721 0.81430 0.22514 0.07777 Time: 0.00070
17-03-25 16:58:01 [1] Train Extra: lr=0.0001747 inv=0.4390625 sub=0.0000000
17-03-25 16:59:13 [1] Step: 18900 Acc: 0.62156 0.84398 Cost: 1.31584 0.90801 0.33009 0.07774 Time: 0.00070
17-03-25 16:59:13 [1] Train Extra: lr=0.0001742 inv=0.4059375 sub=0.0000000
17-03-25 17:00:27 [1] Step: 19000 Acc: 0.62500 0.83984 Cost: 1.28032 0.93890 0.26381 0.07762 Time: 0.00069
17-03-25 17:00:27 [1] Train Extra: lr=0.0001737 inv=0.4018750 sub=0.0000000
17-03-25 17:01:19 [1] Step: 19000 Eval acc: 0.64245 0.84652 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 17:01:19 [1] Eval Extra: inv=0.4048697
17-03-25 17:01:19 [1] Checkpointing with new best dev accuracy of 0.642447
17-03-25 17:02:26 [1] Step: 19100 Acc: 0.62125 0.84227 Cost: 1.03041 0.78368 0.16915 0.07759 Time: 0.00068
17-03-25 17:02:26 [1] Train Extra: lr=0.0001732 inv=0.3920312 sub=0.0000000
17-03-25 17:03:35 [1] Step: 19200 Acc: 0.62156 0.83716 Cost: 1.12416 0.77916 0.26744 0.07757 Time: 0.00066
17-03-25 17:03:35 [1] Train Extra: lr=0.0001727 inv=0.3956250 sub=0.0000000
17-03-25 17:04:56 [1] Step: 19300 Acc: 0.61313 0.83780 Cost: 1.10056 0.83309 0.19001 0.07746 Time: 0.00071
17-03-25 17:04:56 [1] Train Extra: lr=0.0001722 inv=0.3959375 sub=0.0000000
17-03-25 17:06:08 [1] Step: 19400 Acc: 0.63500 0.83552 Cost: 0.94872 0.65852 0.21276 0.07744 Time: 0.00068
17-03-25 17:06:08 [1] Train Extra: lr=0.0001717 inv=0.4196875 sub=0.0000000
17-03-25 17:07:23 [1] Step: 19500 Acc: 0.62562 0.83579 Cost: 1.45837 1.08561 0.29526 0.07750 Time: 0.00068
17-03-25 17:07:23 [1] Train Extra: lr=0.0001712 inv=0.4248438 sub=0.0000000
17-03-25 17:08:15 [1] Step: 19500 Eval acc: 0.63527 0.83890 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 17:08:15 [1] Eval Extra: inv=0.4404814
17-03-25 17:09:28 [1] Step: 19600 Acc: 0.62313 0.83853 Cost: 1.19117 0.90479 0.20888 0.07750 Time: 0.00068
17-03-25 17:09:28 [1] Train Extra: lr=0.0001707 inv=0.4346875 sub=0.0000000
17-03-25 17:10:42 [1] Step: 19700 Acc: 0.61313 0.83753 Cost: 1.13027 0.86574 0.18707 0.07746 Time: 0.00067
17-03-25 17:10:42 [1] Train Extra: lr=0.0001702 inv=0.4267187 sub=0.0000000
17-03-25 17:11:56 [1] Step: 19800 Acc: 0.63687 0.83701 Cost: 1.20297 0.81725 0.30826 0.07746 Time: 0.00068
17-03-25 17:11:56 [1] Train Extra: lr=0.0001697 inv=0.4210937 sub=0.0000000
17-03-25 17:13:14 [1] Step: 19900 Acc: 0.61750 0.84193 Cost: 0.89906 0.66419 0.15741 0.07746 Time: 0.00071
17-03-25 17:13:14 [1] Train Extra: lr=0.0001692 inv=0.4042188 sub=0.0000000
17-03-25 17:14:23 [1] Step: 20000 Acc: 0.62781 0.83931 Cost: 0.99548 0.72637 0.19169 0.07742 Time: 0.00066
17-03-25 17:14:23 [1] Train Extra: lr=0.0001687 inv=0.4218750 sub=0.0000000
17-03-25 17:15:15 [1] Step: 20000 Eval acc: 0.64554 0.84168 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 17:15:15 [1] Eval Extra: inv=0.4159673
17-03-25 17:15:15 [1] Checkpointing.
17-03-25 17:16:30 [1] Step: 20100 Acc: 0.62250 0.84411 Cost: 1.13013 0.79036 0.26236 0.07740 Time: 0.00069
17-03-25 17:16:30 [1] Train Extra: lr=0.0001683 inv=0.4064063 sub=0.0000000
17-03-25 17:17:43 [1] Step: 20200 Acc: 0.62438 0.83789 Cost: 1.02816 0.76562 0.18515 0.07739 Time: 0.00067
17-03-25 17:17:43 [1] Train Extra: lr=0.0001678 inv=0.4385938 sub=0.0000000
17-03-25 17:18:54 [1] Step: 20300 Acc: 0.64062 0.84275 Cost: 0.99488 0.77442 0.14302 0.07744 Time: 0.00070
17-03-25 17:18:54 [1] Train Extra: lr=0.0001673 inv=0.3909375 sub=0.0000000
17-03-25 17:20:14 [1] Step: 20400 Acc: 0.63000 0.83644 Cost: 1.19871 0.85517 0.26610 0.07743 Time: 0.00070
17-03-25 17:20:14 [1] Train Extra: lr=0.0001668 inv=0.4290625 sub=0.0000000
17-03-25 17:21:29 [1] Step: 20500 Acc: 0.61531 0.83784 Cost: 1.20403 0.91096 0.21566 0.07742 Time: 0.00067
17-03-25 17:21:29 [1] Train Extra: lr=0.0001663 inv=0.4485938 sub=0.0000000
17-03-25 17:22:21 [1] Step: 20500 Eval acc: 0.64808 0.84785 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 17:22:21 [1] Eval Extra: inv=0.4005080
17-03-25 17:22:21 [1] Checkpointing with new best dev accuracy of 0.648079
17-03-25 17:23:40 [1] Step: 20600 Acc: 0.62750 0.84456 Cost: 1.04768 0.81549 0.15480 0.07739 Time: 0.00071
17-03-25 17:23:40 [1] Train Extra: lr=0.0001659 inv=0.3995313 sub=0.0000000
17-03-25 17:24:59 [1] Step: 20700 Acc: 0.62938 0.84029 Cost: 1.01444 0.78789 0.14922 0.07733 Time: 0.00070
17-03-25 17:24:59 [1] Train Extra: lr=0.0001654 inv=0.4178125 sub=0.0000000
17-03-25 17:26:06 [1] Step: 20800 Acc: 0.60969 0.83755 Cost: 1.20069 0.87003 0.25335 0.07731 Time: 0.00067
17-03-25 17:26:06 [1] Train Extra: lr=0.0001649 inv=0.3920312 sub=0.0000000
17-03-25 17:27:27 [1] Step: 20900 Acc: 0.61750 0.83894 Cost: 0.94127 0.66732 0.19672 0.07722 Time: 0.00070
17-03-25 17:27:27 [1] Train Extra: lr=0.0001644 inv=0.4064063 sub=0.0000000
17-03-25 17:28:43 [1] Step: 21000 Acc: 0.62594 0.84539 Cost: 1.43994 1.05845 0.30430 0.07719 Time: 0.00069
17-03-25 17:28:43 [1] Train Extra: lr=0.0001640 inv=0.4071875 sub=0.0000000
17-03-25 17:29:38 [1] Step: 21000 Eval acc: 0.65150 0.84864 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 17:29:38 [1] Eval Extra: inv=0.3869810
17-03-25 17:30:53 [1] Step: 21100 Acc: 0.62094 0.84220 Cost: 1.17386 0.85322 0.24352 0.07712 Time: 0.00069
17-03-25 17:30:53 [1] Train Extra: lr=0.0001635 inv=0.4109375 sub=0.0000000
17-03-25 17:32:11 [1] Step: 21200 Acc: 0.62687 0.83702 Cost: 1.07310 0.80753 0.18846 0.07712 Time: 0.00070
17-03-25 17:32:11 [1] Train Extra: lr=0.0001630 inv=0.4573437 sub=0.0000000
17-03-25 17:33:20 [1] Step: 21300 Acc: 0.63187 0.84299 Cost: 1.21056 0.83359 0.29992 0.07704 Time: 0.00067
17-03-25 17:33:20 [1] Train Extra: lr=0.0001626 inv=0.3976562 sub=0.0000000
17-03-25 17:34:33 [1] Step: 21400 Acc: 0.61031 0.84132 Cost: 1.27198 0.90157 0.29343 0.07698 Time: 0.00068
17-03-25 17:34:33 [1] Train Extra: lr=0.0001621 inv=0.4232812 sub=0.0000000
17-03-25 17:36:00 [1] Step: 21500 Acc: 0.63687 0.85201 Cost: 1.24011 0.95858 0.20450 0.07702 Time: 0.00074
17-03-25 17:36:00 [1] Train Extra: lr=0.0001616 inv=0.4351563 sub=0.0000000
17-03-25 17:36:52 [1] Step: 21500 Eval acc: 0.64543 0.84372 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 17:36:52 [1] Eval Extra: inv=0.4082376
17-03-25 17:38:06 [1] Step: 21600 Acc: 0.63906 0.84810 Cost: 0.93692 0.70566 0.15424 0.07701 Time: 0.00071
17-03-25 17:38:06 [1] Train Extra: lr=0.0001612 inv=0.3875000 sub=0.0000000
17-03-25 17:39:25 [1] Step: 21700 Acc: 0.62313 0.83121 Cost: 1.23248 0.85994 0.29548 0.07706 Time: 0.00069
17-03-25 17:39:25 [1] Train Extra: lr=0.0001607 inv=0.4425000 sub=0.0000000
17-03-25 17:40:39 [1] Step: 21800 Acc: 0.62344 0.83512 Cost: 1.04868 0.74290 0.22881 0.07698 Time: 0.00068
17-03-25 17:40:39 [1] Train Extra: lr=0.0001602 inv=0.4362500 sub=0.0000000
17-03-25 17:41:53 [1] Step: 21900 Acc: 0.63344 0.84875 Cost: 1.19551 0.92365 0.19489 0.07698 Time: 0.00070
17-03-25 17:41:53 [1] Train Extra: lr=0.0001598 inv=0.3953125 sub=0.0000000
17-03-25 17:43:18 [1] Step: 22000 Acc: 0.61687 0.84804 Cost: 1.20139 0.95822 0.16615 0.07702 Time: 0.00074
17-03-25 17:43:18 [1] Train Extra: lr=0.0001593 inv=0.4304688 sub=0.0000000
17-03-25 17:44:10 [1] Step: 22000 Eval acc: 0.64587 0.84280 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 17:44:10 [1] Eval Extra: inv=0.4150839
17-03-25 17:45:30 [1] Step: 22100 Acc: 0.63062 0.84224 Cost: 0.97816 0.69312 0.20808 0.07696 Time: 0.00071
17-03-25 17:45:30 [1] Train Extra: lr=0.0001589 inv=0.4217187 sub=0.0000000
17-03-25 17:46:50 [1] Step: 22200 Acc: 0.63062 0.84055 Cost: 0.94920 0.73606 0.13623 0.07691 Time: 0.00070
17-03-25 17:46:50 [1] Train Extra: lr=0.0001584 inv=0.4165625 sub=0.0000000
17-03-25 17:48:04 [1] Step: 22300 Acc: 0.62906 0.83727 Cost: 1.18992 0.79347 0.31956 0.07689 Time: 0.00068
17-03-25 17:48:04 [1] Train Extra: lr=0.0001579 inv=0.4156250 sub=0.0000000
17-03-25 17:49:16 [1] Step: 22400 Acc: 0.61687 0.84068 Cost: 1.06391 0.72815 0.25888 0.07688 Time: 0.00069
17-03-25 17:49:16 [1] Train Extra: lr=0.0001575 inv=0.4198438 sub=0.0000000
17-03-25 17:50:38 [1] Step: 22500 Acc: 0.62438 0.84583 Cost: 1.16055 0.84808 0.23559 0.07687 Time: 0.00071
17-03-25 17:50:38 [1] Train Extra: lr=0.0001570 inv=0.4367187 sub=0.0000000
17-03-25 17:51:30 [1] Step: 22500 Eval acc: 0.64532 0.84713 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 17:51:30 [1] Eval Extra: inv=0.3958149
17-03-25 17:52:36 [1] Step: 22600 Acc: 0.62844 0.84439 Cost: 1.17953 0.85129 0.25144 0.07680 Time: 0.00068
17-03-25 17:52:36 [1] Train Extra: lr=0.0001566 inv=0.4217187 sub=0.0000000
17-03-25 17:53:50 [1] Step: 22700 Acc: 0.62062 0.83586 Cost: 1.02348 0.74078 0.20591 0.07679 Time: 0.00068
17-03-25 17:53:50 [1] Train Extra: lr=0.0001561 inv=0.4456250 sub=0.0000000
17-03-25 17:55:02 [1] Step: 22800 Acc: 0.63625 0.84649 Cost: 1.05743 0.92038 0.06033 0.07672 Time: 0.00070
17-03-25 17:55:02 [1] Train Extra: lr=0.0001557 inv=0.4048438 sub=0.0000000
17-03-25 17:56:13 [1] Step: 22900 Acc: 0.62562 0.84048 Cost: 1.48465 1.13090 0.27696 0.07678 Time: 0.00066
17-03-25 17:56:13 [1] Train Extra: lr=0.0001552 inv=0.4342187 sub=0.0000000
17-03-25 17:57:31 [1] Step: 23000 Acc: 0.61594 0.84609 Cost: 1.16054 0.90146 0.18237 0.07671 Time: 0.00072
17-03-25 17:57:31 [1] Train Extra: lr=0.0001548 inv=0.4267187 sub=0.0000000
17-03-25 17:58:23 [1] Step: 23000 Eval acc: 0.64764 0.84846 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 17:58:23 [1] Eval Extra: inv=0.3538538
17-03-25 17:59:36 [1] Step: 23100 Acc: 0.62906 0.84037 Cost: 1.11208 0.81164 0.22375 0.07668 Time: 0.00069
17-03-25 17:59:36 [1] Train Extra: lr=0.0001544 inv=0.4103125 sub=0.0000000
17-03-25 18:01:02 [1] Step: 23200 Acc: 0.62625 0.84001 Cost: 0.93323 0.84085 0.01572 0.07666 Time: 0.00072
17-03-25 18:01:02 [1] Train Extra: lr=0.0001539 inv=0.4410938 sub=0.0000000
17-03-25 18:02:18 [1] Step: 23300 Acc: 0.63344 0.84602 Cost: 1.12980 0.78374 0.26929 0.07677 Time: 0.00069
17-03-25 18:02:18 [1] Train Extra: lr=0.0001535 inv=0.4187500 sub=0.0000000
17-03-25 18:03:30 [1] Step: 23400 Acc: 0.63750 0.83590 Cost: 1.11474 0.79188 0.24606 0.07681 Time: 0.00068
17-03-25 18:03:30 [1] Train Extra: lr=0.0001530 inv=0.4225000 sub=0.0000000
17-03-25 18:04:50 [1] Step: 23500 Acc: 0.62687 0.83462 Cost: 1.17644 0.87507 0.22461 0.07676 Time: 0.00069
17-03-25 18:04:50 [1] Train Extra: lr=0.0001526 inv=0.4400000 sub=0.0000000
17-03-25 18:05:42 [1] Step: 23500 Eval acc: 0.65139 0.84950 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 18:05:42 [1] Eval Extra: inv=0.3912323
17-03-25 18:06:55 [1] Step: 23600 Acc: 0.63125 0.84034 Cost: 1.12805 0.79036 0.26091 0.07678 Time: 0.00069
17-03-25 18:06:55 [1] Train Extra: lr=0.0001521 inv=0.4090625 sub=0.0000000
17-03-25 18:08:16 [1] Step: 23700 Acc: 0.61344 0.84292 Cost: 1.17234 0.77596 0.31966 0.07672 Time: 0.00069
17-03-25 18:08:16 [1] Train Extra: lr=0.0001517 inv=0.4382813 sub=0.0000000
17-03-25 18:09:30 [1] Step: 23800 Acc: 0.63938 0.84747 Cost: 1.17953 1.08116 0.02164 0.07673 Time: 0.00070
17-03-25 18:09:30 [1] Train Extra: lr=0.0001513 inv=0.3926562 sub=0.0000000
17-03-25 18:10:49 [1] Step: 23900 Acc: 0.62687 0.84396 Cost: 0.89466 0.72708 0.09076 0.07683 Time: 0.00072
17-03-25 18:10:49 [1] Train Extra: lr=0.0001508 inv=0.3970313 sub=0.0000000
17-03-25 18:12:09 [1] Step: 24000 Acc: 0.62000 0.84509 Cost: 1.05497 0.74149 0.23667 0.07681 Time: 0.00071
17-03-25 18:12:09 [1] Train Extra: lr=0.0001504 inv=0.4407813 sub=0.0000000
17-03-25 18:13:01 [1] Step: 24000 Eval acc: 0.65216 0.84882 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 18:13:01 [1] Eval Extra: inv=0.4120473
17-03-25 18:13:01 [1] Checkpointing with new best dev accuracy of 0.652164
17-03-25 18:14:13 [1] Step: 24100 Acc: 0.63094 0.83516 Cost: 1.27826 0.99216 0.20931 0.07679 Time: 0.00068
17-03-25 18:14:13 [1] Train Extra: lr=0.0001500 inv=0.4504687 sub=0.0000000
17-03-25 18:15:33 [1] Step: 24200 Acc: 0.64000 0.84162 Cost: 1.11638 0.85522 0.18443 0.07673 Time: 0.00071
17-03-25 18:15:33 [1] Train Extra: lr=0.0001495 inv=0.4192187 sub=0.0000000
17-03-25 18:16:54 [1] Step: 24300 Acc: 0.61719 0.84188 Cost: 0.98979 0.78182 0.13125 0.07671 Time: 0.00070
17-03-25 18:16:54 [1] Train Extra: lr=0.0001491 inv=0.4293750 sub=0.0000000
17-03-25 18:18:14 [1] Step: 24400 Acc: 0.62500 0.84474 Cost: 1.15831 0.92665 0.15495 0.07671 Time: 0.00071
17-03-25 18:18:14 [1] Train Extra: lr=0.0001487 inv=0.4354688 sub=0.0000000
17-03-25 18:19:28 [1] Step: 24500 Acc: 0.63344 0.83922 Cost: 1.01852 0.73486 0.20692 0.07674 Time: 0.00068
17-03-25 18:19:28 [1] Train Extra: lr=0.0001483 inv=0.4245313 sub=0.0000000
17-03-25 18:20:19 [1] Step: 24500 Eval acc: 0.65051 0.84858 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 18:20:19 [1] Eval Extra: inv=0.4302120
17-03-25 18:21:43 [1] Step: 24600 Acc: 0.62969 0.84190 Cost: 1.01834 0.66280 0.27879 0.07675 Time: 0.00073
17-03-25 18:21:43 [1] Train Extra: lr=0.0001478 inv=0.4329688 sub=0.0000000
17-03-25 18:23:03 [1] Step: 24700 Acc: 0.63906 0.83872 Cost: 0.85525 0.64951 0.12895 0.07680 Time: 0.00070
17-03-25 18:23:03 [1] Train Extra: lr=0.0001474 inv=0.4375000 sub=0.0000000
17-03-25 18:24:23 [1] Step: 24800 Acc: 0.63000 0.83724 Cost: 1.04096 0.80579 0.15844 0.07674 Time: 0.00070
17-03-25 18:24:23 [1] Train Extra: lr=0.0001470 inv=0.4228125 sub=0.0000000
17-03-25 18:25:45 [1] Step: 24900 Acc: 0.62562 0.83437 Cost: 1.00546 0.73295 0.19578 0.07672 Time: 0.00068
17-03-25 18:25:45 [1] Train Extra: lr=0.0001466 inv=0.4520312 sub=0.0000000
17-03-25 18:26:58 [1] Step: 25000 Acc: 0.63031 0.84017 Cost: 1.00294 0.67794 0.24827 0.07673 Time: 0.00070
17-03-25 18:26:58 [1] Train Extra: lr=0.0001461 inv=0.4037500 sub=0.0000000
17-03-25 18:27:50 [1] Step: 25000 Eval acc: 0.64145 0.84235 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 18:27:50 [1] Eval Extra: inv=0.3966983
17-03-25 18:27:50 [1] Checkpointing.
17-03-25 18:29:06 [1] Step: 25100 Acc: 0.63219 0.83413 Cost: 1.05284 0.77112 0.20492 0.07679 Time: 0.00067
17-03-25 18:29:06 [1] Train Extra: lr=0.0001457 inv=0.4328125 sub=0.0000000
17-03-25 18:30:27 [1] Step: 25200 Acc: 0.62906 0.84142 Cost: 1.27031 0.89224 0.30128 0.07679 Time: 0.00072
17-03-25 18:30:27 [1] Train Extra: lr=0.0001453 inv=0.4434375 sub=0.0000000
17-03-25 18:31:47 [1] Step: 25300 Acc: 0.64344 0.84018 Cost: 1.02524 0.76344 0.18494 0.07686 Time: 0.00070
17-03-25 18:31:47 [1] Train Extra: lr=0.0001449 inv=0.4367187 sub=0.0000000
17-03-25 18:33:03 [1] Step: 25400 Acc: 0.64625 0.84179 Cost: 1.01608 0.71848 0.22068 0.07692 Time: 0.00069
17-03-25 18:33:03 [1] Train Extra: lr=0.0001445 inv=0.3984375 sub=0.0000000
17-03-25 18:34:14 [1] Step: 25500 Acc: 0.63187 0.84465 Cost: 0.95462 0.79972 0.07783 0.07706 Time: 0.00069
17-03-25 18:34:14 [1] Train Extra: lr=0.0001441 inv=0.4132812 sub=0.0000000
17-03-25 18:35:06 [1] Step: 25500 Eval acc: 0.64775 0.84752 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 18:35:06 [1] Eval Extra: inv=0.3906250
17-03-25 18:36:20 [1] Step: 25600 Acc: 0.63781 0.84799 Cost: 1.18731 0.89271 0.21758 0.07702 Time: 0.00070
17-03-25 18:36:20 [1] Train Extra: lr=0.0001436 inv=0.4050000 sub=0.0000000
17-03-25 18:37:36 [1] Step: 25700 Acc: 0.64000 0.84160 Cost: 1.17072 0.84958 0.24402 0.07712 Time: 0.00068
17-03-25 18:37:36 [1] Train Extra: lr=0.0001432 inv=0.4446875 sub=0.0000000
17-03-25 18:38:56 [1] Step: 25800 Acc: 0.63875 0.85379 Cost: 1.25123 0.80214 0.37189 0.07720 Time: 0.00074
17-03-25 18:38:56 [1] Train Extra: lr=0.0001428 inv=0.4129687 sub=0.0000000
17-03-25 18:40:20 [1] Step: 25900 Acc: 0.64469 0.84481 Cost: 1.02788 0.72522 0.22549 0.07717 Time: 0.00074
17-03-25 18:40:20 [1] Train Extra: lr=0.0001424 inv=0.4137500 sub=0.0000000
17-03-25 18:41:36 [1] Step: 26000 Acc: 0.63656 0.84643 Cost: 1.11565 0.79361 0.24478 0.07727 Time: 0.00069
17-03-25 18:41:36 [1] Train Extra: lr=0.0001420 inv=0.4310937 sub=0.0000000
17-03-25 18:42:31 [1] Step: 26000 Eval acc: 0.65459 0.84724 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 18:42:31 [1] Eval Extra: inv=0.4517999
17-03-25 18:43:53 [1] Step: 26100 Acc: 0.63219 0.84633 Cost: 1.27611 0.84240 0.35638 0.07733 Time: 0.00070
17-03-25 18:43:53 [1] Train Extra: lr=0.0001416 inv=0.4254688 sub=0.0000000
17-03-25 18:45:08 [1] Step: 26200 Acc: 0.65281 0.83519 Cost: 1.21405 0.93631 0.20029 0.07744 Time: 0.00069
17-03-25 18:45:08 [1] Train Extra: lr=0.0001412 inv=0.4443750 sub=0.0000000
17-03-25 18:46:25 [1] Step: 26300 Acc: 0.63125 0.84275 Cost: 1.39875 1.08091 0.24032 0.07751 Time: 0.00070
17-03-25 18:46:25 [1] Train Extra: lr=0.0001408 inv=0.4270312 sub=0.0000000
17-03-25 18:47:44 [1] Step: 26400 Acc: 0.64344 0.83600 Cost: 1.11993 0.82173 0.22061 0.07759 Time: 0.00071
17-03-25 18:47:44 [1] Train Extra: lr=0.0001404 inv=0.4156250 sub=0.0000000
17-03-25 18:48:53 [1] Step: 26500 Acc: 0.64312 0.84473 Cost: 1.36750 1.04679 0.24308 0.07763 Time: 0.00069
17-03-25 18:48:53 [1] Train Extra: lr=0.0001400 inv=0.3889063 sub=0.0000000
17-03-25 18:49:46 [1] Step: 26500 Eval acc: 0.64852 0.84275 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 18:49:46 [1] Eval Extra: inv=0.4458922
17-03-25 18:51:07 [1] Step: 26600 Acc: 0.64750 0.84265 Cost: 0.97342 0.69221 0.20350 0.07772 Time: 0.00071
17-03-25 18:51:07 [1] Train Extra: lr=0.0001396 inv=0.4051562 sub=0.0000000
17-03-25 18:52:21 [1] Step: 26700 Acc: 0.65531 0.83623 Cost: 1.17159 0.79015 0.30374 0.07770 Time: 0.00068
17-03-25 18:52:21 [1] Train Extra: lr=0.0001392 inv=0.4428125 sub=0.0000000
17-03-25 18:53:42 [1] Step: 26800 Acc: 0.65938 0.84634 Cost: 1.17714 0.89527 0.20414 0.07772 Time: 0.00072
17-03-25 18:53:42 [1] Train Extra: lr=0.0001388 inv=0.4162500 sub=0.0000000
17-03-25 18:54:52 [1] Step: 26900 Acc: 0.63687 0.84468 Cost: 1.00003 0.67707 0.24513 0.07784 Time: 0.00067
17-03-25 18:54:52 [1] Train Extra: lr=0.0001384 inv=0.4137500 sub=0.0000000
17-03-25 18:56:05 [1] Step: 27000 Acc: 0.65844 0.84297 Cost: 1.22254 0.89333 0.25131 0.07790 Time: 0.00070
17-03-25 18:56:05 [1] Train Extra: lr=0.0001380 inv=0.4045313 sub=0.0000000
17-03-25 18:56:57 [1] Step: 27000 Eval acc: 0.65382 0.85047 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 18:56:57 [1] Eval Extra: inv=0.3990724
17-03-25 18:58:16 [1] Step: 27100 Acc: 0.63906 0.84179 Cost: 0.94727 0.67046 0.19886 0.07795 Time: 0.00071
17-03-25 18:58:16 [1] Train Extra: lr=0.0001376 inv=0.4209375 sub=0.0000000
17-03-25 18:59:37 [1] Step: 27200 Acc: 0.63875 0.84145 Cost: 0.93727 0.59629 0.26298 0.07800 Time: 0.00071
17-03-25 18:59:37 [1] Train Extra: lr=0.0001372 inv=0.4175000 sub=0.0000000
17-03-25 19:00:53 [1] Step: 27300 Acc: 0.64938 0.83383 Cost: 1.01862 0.76930 0.17133 0.07799 Time: 0.00066
17-03-25 19:00:53 [1] Train Extra: lr=0.0001368 inv=0.4514063 sub=0.0000000
17-03-25 19:02:05 [1] Step: 27400 Acc: 0.65125 0.83749 Cost: 1.22146 0.87874 0.26470 0.07802 Time: 0.00069
17-03-25 19:02:05 [1] Train Extra: lr=0.0001364 inv=0.4093750 sub=0.0000000
17-03-25 19:03:18 [1] Step: 27500 Acc: 0.65563 0.84285 Cost: 1.06259 0.74824 0.23625 0.07810 Time: 0.00069
17-03-25 19:03:18 [1] Train Extra: lr=0.0001360 inv=0.4004687 sub=0.0000000
17-03-25 19:04:10 [1] Step: 27500 Eval acc: 0.65216 0.85039 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 19:04:10 [1] Eval Extra: inv=0.3968087
17-03-25 19:05:36 [1] Step: 27600 Acc: 0.62031 0.84715 Cost: 1.25004 0.86174 0.31021 0.07810 Time: 0.00073
17-03-25 19:05:36 [1] Train Extra: lr=0.0001356 inv=0.4504687 sub=0.0000000
17-03-25 19:06:51 [1] Step: 27700 Acc: 0.65094 0.83442 Cost: 0.95251 0.71041 0.16402 0.07807 Time: 0.00067
17-03-25 19:06:51 [1] Train Extra: lr=0.0001352 inv=0.4631250 sub=0.0000000
17-03-25 19:08:13 [1] Step: 27800 Acc: 0.65281 0.84991 Cost: 1.15581 0.80206 0.27563 0.07813 Time: 0.00073
17-03-25 19:08:13 [1] Train Extra: lr=0.0001348 inv=0.4153125 sub=0.0000000
17-03-25 19:09:26 [1] Step: 27900 Acc: 0.65844 0.83806 Cost: 1.32426 1.00901 0.23710 0.07815 Time: 0.00068
17-03-25 19:09:26 [1] Train Extra: lr=0.0001344 inv=0.4257812 sub=0.0000000
17-03-25 19:10:36 [1] Step: 28000 Acc: 0.64844 0.83973 Cost: 0.97982 0.63082 0.27075 0.07825 Time: 0.00066
17-03-25 19:10:36 [1] Train Extra: lr=0.0001341 inv=0.3945312 sub=0.0000000
17-03-25 19:11:31 [1] Step: 28000 Eval acc: 0.65283 0.84833 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 19:11:31 [1] Eval Extra: inv=0.4230897
17-03-25 19:12:46 [1] Step: 28100 Acc: 0.66063 0.84082 Cost: 1.13271 0.96509 0.08932 0.07830 Time: 0.00069
17-03-25 19:12:46 [1] Train Extra: lr=0.0001337 inv=0.4356250 sub=0.0000000
17-03-25 19:13:59 [1] Step: 28200 Acc: 0.65687 0.83765 Cost: 1.02366 0.69296 0.25237 0.07833 Time: 0.00068
17-03-25 19:13:59 [1] Train Extra: lr=0.0001333 inv=0.4262500 sub=0.0000000
17-03-25 19:15:19 [1] Step: 28300 Acc: 0.64094 0.84553 Cost: 1.25352 0.89484 0.28018 0.07850 Time: 0.00071
17-03-25 19:15:19 [1] Train Extra: lr=0.0001329 inv=0.4212500 sub=0.0000000
17-03-25 19:16:41 [1] Step: 28400 Acc: 0.66375 0.84311 Cost: 1.40529 1.07115 0.25561 0.07853 Time: 0.00071
17-03-25 19:16:41 [1] Train Extra: lr=0.0001325 inv=0.4250000 sub=0.0000000
17-03-25 19:17:54 [1] Step: 28500 Acc: 0.64750 0.84113 Cost: 1.24279 0.84096 0.32328 0.07855 Time: 0.00069
17-03-25 19:17:54 [1] Train Extra: lr=0.0001321 inv=0.4378125 sub=0.0000000
17-03-25 19:18:46 [1] Step: 28500 Eval acc: 0.65857 0.85059 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 19:18:46 [1] Eval Extra: inv=0.4651612
17-03-25 19:18:46 [1] Checkpointing with new best dev accuracy of 0.658569
17-03-25 19:20:02 [1] Step: 28600 Acc: 0.64344 0.84788 Cost: 0.74023 0.57317 0.08841 0.07865 Time: 0.00070
17-03-25 19:20:02 [1] Train Extra: lr=0.0001318 inv=0.4156250 sub=0.0000000
17-03-25 19:21:24 [1] Step: 28700 Acc: 0.64156 0.84365 Cost: 1.34116 1.00985 0.25272 0.07858 Time: 0.00072
17-03-25 19:21:24 [1] Train Extra: lr=0.0001314 inv=0.4351563 sub=0.0000000
17-03-25 19:22:39 [1] Step: 28800 Acc: 0.64250 0.84261 Cost: 0.97138 0.69812 0.19457 0.07869 Time: 0.00070
17-03-25 19:22:39 [1] Train Extra: lr=0.0001310 inv=0.4420312 sub=0.0000000
17-03-25 19:23:46 [1] Step: 28900 Acc: 0.64156 0.84342 Cost: 1.10987 0.75916 0.27202 0.07868 Time: 0.00067
17-03-25 19:23:46 [1] Train Extra: lr=0.0001306 inv=0.4293750 sub=0.0000000
17-03-25 19:25:14 [1] Step: 29000 Acc: 0.63625 0.84395 Cost: 0.95795 0.75957 0.11963 0.07875 Time: 0.00072
17-03-25 19:25:14 [1] Train Extra: lr=0.0001303 inv=0.4618750 sub=0.0000000
17-03-25 19:26:06 [1] Step: 29000 Eval acc: 0.65813 0.84950 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 19:26:06 [1] Eval Extra: inv=0.3981890
17-03-25 19:27:32 [1] Step: 29100 Acc: 0.65156 0.84172 Cost: 1.13652 0.88011 0.17757 0.07884 Time: 0.00073
17-03-25 19:27:32 [1] Train Extra: lr=0.0001299 inv=0.4542188 sub=0.0000000
17-03-25 19:28:40 [1] Step: 29200 Acc: 0.64750 0.84583 Cost: 1.00737 0.74517 0.18335 0.07885 Time: 0.00068
17-03-25 19:28:40 [1] Train Extra: lr=0.0001295 inv=0.3890625 sub=0.0000000
17-03-25 19:29:55 [1] Step: 29300 Acc: 0.63406 0.84700 Cost: 1.13091 0.85606 0.19609 0.07876 Time: 0.00068
17-03-25 19:29:55 [1] Train Extra: lr=0.0001291 inv=0.4329688 sub=0.0000000
17-03-25 19:31:16 [1] Step: 29400 Acc: 0.65344 0.85300 Cost: 1.28863 0.89417 0.31563 0.07882 Time: 0.00074
17-03-25 19:31:16 [1] Train Extra: lr=0.0001288 inv=0.3954687 sub=0.0000000
17-03-25 19:32:36 [1] Step: 29500 Acc: 0.63750 0.84017 Cost: 1.14379 0.79368 0.27125 0.07886 Time: 0.00070
17-03-25 19:32:36 [1] Train Extra: lr=0.0001284 inv=0.4368750 sub=0.0000000
17-03-25 19:33:29 [1] Step: 29500 Eval acc: 0.65835 0.85266 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 19:33:29 [1] Eval Extra: inv=0.3875331
17-03-25 19:34:52 [1] Step: 29600 Acc: 0.64375 0.85108 Cost: 0.80176 0.56650 0.15630 0.07895 Time: 0.00076
17-03-25 19:34:52 [1] Train Extra: lr=0.0001280 inv=0.4212500 sub=0.0000000
17-03-25 19:35:55 [1] Step: 29700 Acc: 0.65063 0.84486 Cost: 1.11456 0.81374 0.22187 0.07895 Time: 0.00066
17-03-25 19:35:55 [1] Train Extra: lr=0.0001277 inv=0.3748437 sub=0.0000000
17-03-25 19:37:06 [1] Step: 29800 Acc: 0.64719 0.84260 Cost: 1.07834 0.83530 0.16407 0.07897 Time: 0.00068
17-03-25 19:37:06 [1] Train Extra: lr=0.0001273 inv=0.4212500 sub=0.0000000
17-03-25 19:38:12 [1] Step: 29900 Acc: 0.64000 0.84708 Cost: 1.38343 1.12360 0.18087 0.07896 Time: 0.00068
17-03-25 19:38:12 [1] Train Extra: lr=0.0001269 inv=0.3975000 sub=0.0000000
17-03-25 19:39:38 [1] Step: 30000 Acc: 0.62906 0.85481 Cost: 1.07975 0.81149 0.18926 0.07900 Time: 0.00075
17-03-25 19:39:38 [1] Train Extra: lr=0.0001266 inv=0.4303125 sub=0.0000000
17-03-25 19:40:31 [1] Step: 30000 Eval acc: 0.66133 0.84797 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 19:40:31 [1] Eval Extra: inv=0.4238074
17-03-25 19:40:31 [1] Checkpointing.
17-03-25 19:41:52 [1] Step: 30100 Acc: 0.63219 0.82817 Cost: 1.39055 0.98242 0.32922 0.07891 Time: 0.00067
17-03-25 19:41:52 [1] Train Extra: lr=0.0001262 inv=0.4820313 sub=0.0000000
17-03-25 19:43:03 [1] Step: 30200 Acc: 0.64062 0.85120 Cost: 1.11861 0.78726 0.25247 0.07888 Time: 0.00072
17-03-25 19:43:03 [1] Train Extra: lr=0.0001258 inv=0.3787500 sub=0.0000000
17-03-25 19:44:18 [1] Step: 30300 Acc: 0.64687 0.84555 Cost: 0.84067 0.66713 0.09463 0.07891 Time: 0.00069
17-03-25 19:44:18 [1] Train Extra: lr=0.0001255 inv=0.4187500 sub=0.0000000
17-03-25 19:45:33 [1] Step: 30400 Acc: 0.64844 0.84395 Cost: 0.92934 0.62503 0.22532 0.07900 Time: 0.00069
17-03-25 19:45:33 [1] Train Extra: lr=0.0001251 inv=0.4134375 sub=0.0000000
17-03-25 19:46:53 [1] Step: 30500 Acc: 0.64750 0.83164 Cost: 0.86308 0.67688 0.10707 0.07913 Time: 0.00069
17-03-25 19:46:53 [1] Train Extra: lr=0.0001248 inv=0.4431250 sub=0.0000000
17-03-25 19:47:47 [1] Step: 30500 Eval acc: 0.66100 0.85050 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 19:47:47 [1] Eval Extra: inv=0.4095627
17-03-25 19:49:18 [1] Step: 30600 Acc: 0.64938 0.84372 Cost: 1.11013 0.72053 0.31050 0.07910 Time: 0.00077
17-03-25 19:49:18 [1] Train Extra: lr=0.0001244 inv=0.4343750 sub=0.0000000
17-03-25 19:50:47 [1] Step: 30700 Acc: 0.63313 0.84508 Cost: 1.13003 0.69601 0.35495 0.07908 Time: 0.00076
17-03-25 19:50:47 [1] Train Extra: lr=0.0001240 inv=0.4459375 sub=0.0000000
17-03-25 19:52:07 [1] Step: 30800 Acc: 0.64031 0.84155 Cost: 1.08020 0.75964 0.24146 0.07909 Time: 0.00070
17-03-25 19:52:07 [1] Train Extra: lr=0.0001237 inv=0.4648438 sub=0.0000000
17-03-25 19:53:28 [1] Step: 30900 Acc: 0.64031 0.84079 Cost: 1.16345 0.78807 0.29627 0.07912 Time: 0.00072
17-03-25 19:53:28 [1] Train Extra: lr=0.0001233 inv=0.4428125 sub=0.0000000
17-03-25 19:54:47 [1] Step: 31000 Acc: 0.63875 0.84482 Cost: 1.07083 0.72106 0.27059 0.07918 Time: 0.00072
17-03-25 19:54:47 [1] Train Extra: lr=0.0001230 inv=0.4282813 sub=0.0000000
17-03-25 19:55:42 [1] Step: 31000 Eval acc: 0.65448 0.84837 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 19:55:42 [1] Eval Extra: inv=0.3905698
17-03-25 19:57:08 [1] Step: 31100 Acc: 0.64438 0.84575 Cost: 0.98917 0.71730 0.19271 0.07916 Time: 0.00076
17-03-25 19:57:08 [1] Train Extra: lr=0.0001226 inv=0.4420312 sub=0.0000000
17-03-25 19:58:25 [1] Step: 31200 Acc: 0.63625 0.84675 Cost: 0.85762 0.61783 0.16057 0.07922 Time: 0.00071
17-03-25 19:58:25 [1] Train Extra: lr=0.0001223 inv=0.4271875 sub=0.0000000
17-03-25 19:59:34 [1] Step: 31300 Acc: 0.63562 0.84814 Cost: 1.51658 1.14309 0.29431 0.07918 Time: 0.00068
17-03-25 19:59:34 [1] Train Extra: lr=0.0001219 inv=0.3943750 sub=0.0000000
17-03-25 20:00:55 [1] Step: 31400 Acc: 0.65812 0.84515 Cost: 1.05522 0.71703 0.25901 0.07918 Time: 0.00071
17-03-25 20:00:55 [1] Train Extra: lr=0.0001216 inv=0.4504687 sub=0.0000000
17-03-25 20:02:20 [1] Step: 31500 Acc: 0.65781 0.85092 Cost: 1.04849 0.76077 0.20849 0.07922 Time: 0.00075
17-03-25 20:02:20 [1] Train Extra: lr=0.0001212 inv=0.4237500 sub=0.0000000
17-03-25 20:03:09 [1] Step: 31500 Eval acc: 0.66023 0.85353 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00014
17-03-25 20:03:09 [1] Eval Extra: inv=0.4239731
17-03-25 20:04:21 [1] Step: 31600 Acc: 0.65250 0.85160 Cost: 1.09289 0.82024 0.19338 0.07927 Time: 0.00071
17-03-25 20:04:21 [1] Train Extra: lr=0.0001209 inv=0.4046875 sub=0.0000000
17-03-25 20:05:38 [1] Step: 31700 Acc: 0.65000 0.84753 Cost: 1.29562 0.98675 0.22960 0.07928 Time: 0.00070
17-03-25 20:05:38 [1] Train Extra: lr=0.0001205 inv=0.4059375 sub=0.0000000
17-03-25 20:06:56 [1] Step: 31800 Acc: 0.62219 0.83977 Cost: 0.80003 0.61546 0.10528 0.07930 Time: 0.00071
17-03-25 20:06:56 [1] Train Extra: lr=0.0001202 inv=0.4296875 sub=0.0000000
17-03-25 20:08:06 [1] Step: 31900 Acc: 0.63875 0.84515 Cost: 1.17530 0.87715 0.21882 0.07933 Time: 0.00066
17-03-25 20:08:06 [1] Train Extra: lr=0.0001198 inv=0.4192187 sub=0.0000000
17-03-25 20:09:24 [1] Step: 32000 Acc: 0.64094 0.84743 Cost: 0.99858 0.77734 0.14184 0.07940 Time: 0.00072
17-03-25 20:09:24 [1] Train Extra: lr=0.0001195 inv=0.3975000 sub=0.0000000
17-03-25 20:10:18 [1] Step: 32000 Eval acc: 0.66221 0.84614 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 20:10:18 [1] Eval Extra: inv=0.4313715
17-03-25 20:10:18 [1] Checkpointing with new best dev accuracy of 0.662213
17-03-25 20:11:40 [1] Step: 32100 Acc: 0.64094 0.84617 Cost: 0.97457 0.73736 0.15776 0.07945 Time: 0.00071
17-03-25 20:11:40 [1] Train Extra: lr=0.0001191 inv=0.4496875 sub=0.0000000
17-03-25 20:13:07 [1] Step: 32200 Acc: 0.64781 0.85682 Cost: 1.02516 0.73718 0.20848 0.07950 Time: 0.00076
17-03-25 20:13:07 [1] Train Extra: lr=0.0001188 inv=0.4335938 sub=0.0000000
17-03-25 20:14:29 [1] Step: 32300 Acc: 0.63750 0.84334 Cost: 0.89002 0.72701 0.08349 0.07952 Time: 0.00073
17-03-25 20:14:29 [1] Train Extra: lr=0.0001185 inv=0.4254688 sub=0.0000000
17-03-25 20:15:52 [1] Step: 32400 Acc: 0.64750 0.85202 Cost: 0.80493 0.63755 0.08784 0.07954 Time: 0.00076
17-03-25 20:15:52 [1] Train Extra: lr=0.0001181 inv=0.3985938 sub=0.0000000
17-03-25 20:17:29 [1] Step: 32500 Acc: 0.63375 0.84875 Cost: 1.27863 0.83679 0.36235 0.07949 Time: 0.00078
17-03-25 20:17:29 [1] Train Extra: lr=0.0001178 inv=0.4337500 sub=0.0000000
17-03-25 20:18:25 [1] Step: 32500 Eval acc: 0.66299 0.85126 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 20:18:25 [1] Eval Extra: inv=0.4780808
17-03-25 20:19:35 [1] Step: 32600 Acc: 0.64438 0.84526 Cost: 1.12083 0.81122 0.23013 0.07948 Time: 0.00070
17-03-25 20:19:35 [1] Train Extra: lr=0.0001174 inv=0.4121875 sub=0.0000000
17-03-25 20:20:52 [1] Step: 32700 Acc: 0.63531 0.84560 Cost: 1.37892 1.06018 0.23931 0.07943 Time: 0.00073
17-03-25 20:20:52 [1] Train Extra: lr=0.0001171 inv=0.3964063 sub=0.0000000
17-03-25 20:22:16 [1] Step: 32800 Acc: 0.64625 0.83112 Cost: 1.01878 0.78035 0.15904 0.07940 Time: 0.00071
17-03-25 20:22:16 [1] Train Extra: lr=0.0001168 inv=0.4537500 sub=0.0000000
17-03-25 20:23:39 [1] Step: 32900 Acc: 0.64719 0.84522 Cost: 1.27821 0.92517 0.27366 0.07938 Time: 0.00074
17-03-25 20:23:39 [1] Train Extra: lr=0.0001164 inv=0.4525000 sub=0.0000000
17-03-25 20:24:56 [1] Step: 33000 Acc: 0.64031 0.84239 Cost: 0.84971 0.59953 0.17074 0.07944 Time: 0.00070
17-03-25 20:24:56 [1] Train Extra: lr=0.0001161 inv=0.4381250 sub=0.0000000
17-03-25 20:25:52 [1] Step: 33000 Eval acc: 0.65647 0.84582 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 20:25:52 [1] Eval Extra: inv=0.4402606
17-03-25 20:27:16 [1] Step: 33100 Acc: 0.63875 0.84577 Cost: 1.12883 0.80795 0.24147 0.07941 Time: 0.00074
17-03-25 20:27:16 [1] Train Extra: lr=0.0001158 inv=0.4267187 sub=0.0000000
17-03-25 20:28:35 [1] Step: 33200 Acc: 0.64250 0.84544 Cost: 1.24004 0.93633 0.22427 0.07944 Time: 0.00073
17-03-25 20:28:35 [1] Train Extra: lr=0.0001154 inv=0.4206250 sub=0.0000000
17-03-25 20:30:01 [1] Step: 33300 Acc: 0.64687 0.83503 Cost: 1.02061 0.71144 0.22973 0.07944 Time: 0.00072
17-03-25 20:30:01 [1] Train Extra: lr=0.0001151 inv=0.4557813 sub=0.0000000
17-03-25 20:31:19 [1] Step: 33400 Acc: 0.66156 0.85500 Cost: 0.91576 0.67144 0.16483 0.07948 Time: 0.00076
17-03-25 20:31:19 [1] Train Extra: lr=0.0001148 inv=0.4018750 sub=0.0000000
17-03-25 20:32:37 [1] Step: 33500 Acc: 0.64219 0.84288 Cost: 0.99300 0.70732 0.20615 0.07953 Time: 0.00070
17-03-25 20:32:37 [1] Train Extra: lr=0.0001144 inv=0.4539063 sub=0.0000000
17-03-25 20:33:34 [1] Step: 33500 Eval acc: 0.66199 0.84824 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 20:33:34 [1] Eval Extra: inv=0.4223719
17-03-25 20:34:53 [1] Step: 33600 Acc: 0.63719 0.84111 Cost: 0.86923 0.67340 0.11619 0.07965 Time: 0.00074
17-03-25 20:34:53 [1] Train Extra: lr=0.0001141 inv=0.4151563 sub=0.0000000
17-03-25 20:36:10 [1] Step: 33700 Acc: 0.66375 0.84604 Cost: 1.07236 0.75208 0.24052 0.07976 Time: 0.00073
17-03-25 20:36:10 [1] Train Extra: lr=0.0001138 inv=0.4167188 sub=0.0000000
17-03-25 20:37:30 [1] Step: 33800 Acc: 0.66312 0.84463 Cost: 1.00107 0.69419 0.22696 0.07992 Time: 0.00073
17-03-25 20:37:30 [1] Train Extra: lr=0.0001135 inv=0.4400000 sub=0.0000000
17-03-25 20:38:51 [1] Step: 33900 Acc: 0.67031 0.84463 Cost: 1.13431 0.83213 0.22224 0.07994 Time: 0.00071
17-03-25 20:38:51 [1] Train Extra: lr=0.0001131 inv=0.4450000 sub=0.0000000
17-03-25 20:40:12 [1] Step: 34000 Acc: 0.65969 0.85767 Cost: 0.94956 0.76972 0.09980 0.08004 Time: 0.00079
17-03-25 20:40:12 [1] Train Extra: lr=0.0001128 inv=0.3754688 sub=0.0000000
17-03-25 20:41:05 [1] Step: 34000 Eval acc: 0.66718 0.84868 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 20:41:05 [1] Eval Extra: inv=0.4184519
17-03-25 20:41:05 [1] Checkpointing with new best dev accuracy of 0.667182
17-03-25 20:42:28 [1] Step: 34100 Acc: 0.66219 0.84053 Cost: 0.99711 0.74793 0.16901 0.08017 Time: 0.00074
17-03-25 20:42:28 [1] Train Extra: lr=0.0001125 inv=0.4067188 sub=0.0000000
17-03-25 20:43:55 [1] Step: 34200 Acc: 0.66719 0.85366 Cost: 1.10482 0.76629 0.25837 0.08016 Time: 0.00078
17-03-25 20:43:55 [1] Train Extra: lr=0.0001122 inv=0.4231250 sub=0.0000000
17-03-25 20:45:12 [1] Step: 34300 Acc: 0.65750 0.84319 Cost: 1.15287 0.82401 0.24855 0.08031 Time: 0.00073
17-03-25 20:45:12 [1] Train Extra: lr=0.0001118 inv=0.4156250 sub=0.0000000
17-03-25 20:46:36 [1] Step: 34400 Acc: 0.66187 0.84766 Cost: 0.99645 0.71143 0.20459 0.08043 Time: 0.00075
17-03-25 20:46:36 [1] Train Extra: lr=0.0001115 inv=0.4156250 sub=0.0000000
17-03-25 20:47:59 [1] Step: 34500 Acc: 0.65344 0.85099 Cost: 1.35161 1.01604 0.25512 0.08045 Time: 0.00076
17-03-25 20:47:59 [1] Train Extra: lr=0.0001112 inv=0.4006250 sub=0.0000000
17-03-25 20:48:51 [1] Step: 34500 Eval acc: 0.66034 0.85125 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00015
17-03-25 20:48:51 [1] Eval Extra: inv=0.4196113
17-03-25 20:50:13 [1] Step: 34600 Acc: 0.65781 0.84777 Cost: 1.07596 0.84887 0.14660 0.08049 Time: 0.00076
17-03-25 20:50:13 [1] Train Extra: lr=0.0001109 inv=0.4131250 sub=0.0000000
17-03-25 20:51:31 [1] Step: 34700 Acc: 0.65781 0.83963 Cost: 0.81659 0.64053 0.09545 0.08061 Time: 0.00071
17-03-25 20:51:31 [1] Train Extra: lr=0.0001106 inv=0.4320312 sub=0.0000000
17-03-25 20:52:44 [1] Step: 34800 Acc: 0.66844 0.84601 Cost: 1.15875 0.80707 0.27100 0.08068 Time: 0.00069
17-03-25 20:52:44 [1] Train Extra: lr=0.0001102 inv=0.4150000 sub=0.0000000
17-03-25 20:54:12 [1] Step: 34900 Acc: 0.66875 0.83943 Cost: 1.08158 0.81033 0.19049 0.08076 Time: 0.00074
17-03-25 20:54:12 [1] Train Extra: lr=0.0001099 inv=0.4542188 sub=0.0000000
17-03-25 20:55:31 [1] Step: 35000 Acc: 0.65812 0.84595 Cost: 1.22298 0.87723 0.26495 0.08079 Time: 0.00073
17-03-25 20:55:31 [1] Train Extra: lr=0.0001096 inv=0.4035937 sub=0.0000000
17-03-25 20:56:28 [1] Step: 35000 Eval acc: 0.66453 0.85044 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 20:56:28 [1] Eval Extra: inv=0.4473830
17-03-25 20:56:28 [1] Checkpointing.
17-03-25 20:57:54 [1] Step: 35100 Acc: 0.64844 0.84043 Cost: 0.89845 0.62774 0.18983 0.08087 Time: 0.00075
17-03-25 20:57:54 [1] Train Extra: lr=0.0001093 inv=0.4539063 sub=0.0000000
17-03-25 20:59:17 [1] Step: 35200 Acc: 0.65812 0.83587 Cost: 1.18891 0.88633 0.22159 0.08099 Time: 0.00071
17-03-25 20:59:17 [1] Train Extra: lr=0.0001090 inv=0.4546875 sub=0.0000000
17-03-25 21:00:41 [1] Step: 35300 Acc: 0.64594 0.84056 Cost: 1.11124 0.76778 0.26244 0.08103 Time: 0.00074
17-03-25 21:00:41 [1] Train Extra: lr=0.0001087 inv=0.4257812 sub=0.0000000
17-03-25 21:01:59 [1] Step: 35400 Acc: 0.67406 0.84513 Cost: 0.97400 0.65719 0.23566 0.08115 Time: 0.00072
17-03-25 21:01:59 [1] Train Extra: lr=0.0001084 inv=0.4245313 sub=0.0000000
17-03-25 21:03:18 [1] Step: 35500 Acc: 0.65438 0.84826 Cost: 0.96814 0.73506 0.15187 0.08122 Time: 0.00073
17-03-25 21:03:18 [1] Train Extra: lr=0.0001080 inv=0.4129687 sub=0.0000000
17-03-25 21:04:14 [1] Step: 35500 Eval acc: 0.66508 0.85055 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 21:04:14 [1] Eval Extra: inv=0.4180654
17-03-25 21:05:33 [1] Step: 35600 Acc: 0.66094 0.84607 Cost: 1.09889 0.79560 0.22206 0.08123 Time: 0.00072
17-03-25 21:05:33 [1] Train Extra: lr=0.0001077 inv=0.4104687 sub=0.0000000
17-03-25 21:07:03 [1] Step: 35700 Acc: 0.65375 0.84824 Cost: 0.97943 0.78266 0.11550 0.08126 Time: 0.00078
17-03-25 21:07:03 [1] Train Extra: lr=0.0001074 inv=0.4251563 sub=0.0000000
17-03-25 21:08:27 [1] Step: 35800 Acc: 0.65156 0.84055 Cost: 1.34013 0.94096 0.31786 0.08131 Time: 0.00074
17-03-25 21:08:27 [1] Train Extra: lr=0.0001071 inv=0.4268750 sub=0.0000000
17-03-25 21:09:52 [1] Step: 35900 Acc: 0.65969 0.84685 Cost: 0.80901 0.59239 0.13518 0.08145 Time: 0.00073
17-03-25 21:09:52 [1] Train Extra: lr=0.0001068 inv=0.4304688 sub=0.0000000
17-03-25 21:11:16 [1] Step: 36000 Acc: 0.65250 0.83484 Cost: 0.93476 0.68602 0.16720 0.08155 Time: 0.00071
17-03-25 21:11:16 [1] Train Extra: lr=0.0001065 inv=0.4604687 sub=0.0000000
17-03-25 21:12:13 [1] Step: 36000 Eval acc: 0.65802 0.84871 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 21:12:13 [1] Eval Extra: inv=0.4073542
17-03-25 21:13:29 [1] Step: 36100 Acc: 0.65938 0.84263 Cost: 1.23279 0.85096 0.30025 0.08158 Time: 0.00073
17-03-25 21:13:29 [1] Train Extra: lr=0.0001062 inv=0.3984375 sub=0.0000000
17-03-25 21:14:59 [1] Step: 36200 Acc: 0.66094 0.85684 Cost: 0.94793 0.64722 0.21903 0.08169 Time: 0.00080
17-03-25 21:14:59 [1] Train Extra: lr=0.0001059 inv=0.4157812 sub=0.0000000
17-03-25 21:16:20 [1] Step: 36300 Acc: 0.66938 0.84581 Cost: 1.10284 0.74258 0.27847 0.08179 Time: 0.00074
17-03-25 21:16:20 [1] Train Extra: lr=0.0001056 inv=0.4179688 sub=0.0000000
17-03-25 21:17:49 [1] Step: 36400 Acc: 0.65094 0.83856 Cost: 1.04696 0.73437 0.23074 0.08185 Time: 0.00075
17-03-25 21:17:49 [1] Train Extra: lr=0.0001053 inv=0.4717188 sub=0.0000000
17-03-25 21:19:21 [1] Step: 36500 Acc: 0.65094 0.84870 Cost: 1.01680 0.67426 0.26063 0.08191 Time: 0.00079
17-03-25 21:19:21 [1] Train Extra: lr=0.0001050 inv=0.4373437 sub=0.0000000
17-03-25 21:20:17 [1] Step: 36500 Eval acc: 0.66542 0.84752 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 21:20:17 [1] Eval Extra: inv=0.3668838
17-03-25 21:21:36 [1] Step: 36600 Acc: 0.66531 0.84388 Cost: 1.36098 1.07433 0.20468 0.08197 Time: 0.00071
17-03-25 21:21:36 [1] Train Extra: lr=0.0001047 inv=0.4459375 sub=0.0000000
17-03-25 21:22:59 [1] Step: 36700 Acc: 0.65469 0.85790 Cost: 1.09228 0.83007 0.18025 0.08196 Time: 0.00078
17-03-25 21:22:59 [1] Train Extra: lr=0.0001044 inv=0.3890625 sub=0.0000000
17-03-25 21:24:12 [1] Step: 36800 Acc: 0.65594 0.84634 Cost: 1.03987 0.70731 0.25052 0.08203 Time: 0.00071
17-03-25 21:24:12 [1] Train Extra: lr=0.0001041 inv=0.3992188 sub=0.0000000
17-03-25 21:25:30 [1] Step: 36900 Acc: 0.65875 0.84248 Cost: 1.24312 0.94315 0.21787 0.08210 Time: 0.00072
17-03-25 21:25:30 [1] Train Extra: lr=0.0001038 inv=0.4046875 sub=0.0000000
17-03-25 21:26:51 [1] Step: 37000 Acc: 0.66156 0.84229 Cost: 1.04480 0.72927 0.23331 0.08222 Time: 0.00075
17-03-25 21:26:51 [1] Train Extra: lr=0.0001035 inv=0.4032812 sub=0.0000000
17-03-25 21:27:48 [1] Step: 37000 Eval acc: 0.66166 0.85223 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 21:27:48 [1] Eval Extra: inv=0.4062500
17-03-25 21:29:07 [1] Step: 37100 Acc: 0.64500 0.84325 Cost: 0.82058 0.64597 0.09226 0.08234 Time: 0.00073
17-03-25 21:29:07 [1] Train Extra: lr=0.0001032 inv=0.4284375 sub=0.0000000
17-03-25 21:30:33 [1] Step: 37200 Acc: 0.67250 0.85392 Cost: 1.12313 0.76888 0.27187 0.08237 Time: 0.00076
17-03-25 21:30:33 [1] Train Extra: lr=0.0001029 inv=0.4131250 sub=0.0000000
17-03-25 21:31:46 [1] Step: 37300 Acc: 0.64719 0.84832 Cost: 1.00409 0.75617 0.16547 0.08245 Time: 0.00070
17-03-25 21:31:46 [1] Train Extra: lr=0.0001026 inv=0.3998437 sub=0.0000000
17-03-25 21:33:03 [1] Step: 37400 Acc: 0.65781 0.84022 Cost: 1.22312 0.99660 0.14405 0.08247 Time: 0.00070
17-03-25 21:33:03 [1] Train Extra: lr=0.0001023 inv=0.4373437 sub=0.0000000
17-03-25 21:34:18 [1] Step: 37500 Acc: 0.66063 0.84310 Cost: 1.19217 0.81688 0.29281 0.08248 Time: 0.00073
17-03-25 21:34:18 [1] Train Extra: lr=0.0001020 inv=0.4220313 sub=0.0000000
17-03-25 21:35:12 [1] Step: 37500 Eval acc: 0.66464 0.85175 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 21:35:12 [1] Eval Extra: inv=0.4370031
17-03-25 21:36:29 [1] Step: 37600 Acc: 0.64219 0.85176 Cost: 1.21445 0.88167 0.25022 0.08256 Time: 0.00074
17-03-25 21:36:29 [1] Train Extra: lr=0.0001017 inv=0.4031250 sub=0.0000000
17-03-25 21:37:59 [1] Step: 37700 Acc: 0.65531 0.83785 Cost: 1.05781 0.87607 0.09913 0.08261 Time: 0.00074
17-03-25 21:37:59 [1] Train Extra: lr=0.0001014 inv=0.4456250 sub=0.0000000
17-03-25 21:39:17 [1] Step: 37800 Acc: 0.65125 0.84254 Cost: 0.98009 0.72216 0.17536 0.08257 Time: 0.00072
17-03-25 21:39:17 [1] Train Extra: lr=0.0001011 inv=0.4185937 sub=0.0000000
17-03-25 21:40:41 [1] Step: 37900 Acc: 0.65875 0.84657 Cost: 1.31020 0.93953 0.28803 0.08265 Time: 0.00075
17-03-25 21:40:41 [1] Train Extra: lr=0.0001008 inv=0.4404688 sub=0.0000000
17-03-25 21:41:55 [1] Step: 38000 Acc: 0.65906 0.83958 Cost: 1.04253 0.83741 0.12240 0.08273 Time: 0.00068
17-03-25 21:41:55 [1] Train Extra: lr=0.0001005 inv=0.4406250 sub=0.0000000
17-03-25 21:42:51 [1] Step: 38000 Eval acc: 0.67083 0.84607 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 21:42:51 [1] Eval Extra: inv=0.4295495
17-03-25 21:42:51 [1] Checkpointing with new best dev accuracy of 0.670826
17-03-25 21:44:13 [1] Step: 38100 Acc: 0.65844 0.83812 Cost: 0.91760 0.69010 0.14479 0.08270 Time: 0.00072
17-03-25 21:44:13 [1] Train Extra: lr=0.0001003 inv=0.4410938 sub=0.0000000
17-03-25 21:45:35 [1] Step: 38200 Acc: 0.65812 0.83526 Cost: 0.98698 0.69724 0.20698 0.08276 Time: 0.00073
17-03-25 21:45:35 [1] Train Extra: lr=0.0001000 inv=0.4389062 sub=0.0000000
17-03-25 21:47:01 [1] Step: 38300 Acc: 0.64844 0.84643 Cost: 1.02342 0.78532 0.15527 0.08283 Time: 0.00074
17-03-25 21:47:01 [1] Train Extra: lr=0.0000997 inv=0.4403125 sub=0.0000000
17-03-25 21:48:18 [1] Step: 38400 Acc: 0.64375 0.84022 Cost: 1.00401 0.77082 0.15026 0.08293 Time: 0.00071
17-03-25 21:48:18 [1] Train Extra: lr=0.0000994 inv=0.4250000 sub=0.0000000
17-03-25 21:49:37 [1] Step: 38500 Acc: 0.66094 0.84027 Cost: 1.16837 0.89382 0.19158 0.08297 Time: 0.00072
17-03-25 21:49:37 [1] Train Extra: lr=0.0000991 inv=0.4009375 sub=0.0000000
17-03-25 21:50:33 [1] Step: 38500 Eval acc: 0.66453 0.84959 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 21:50:33 [1] Eval Extra: inv=0.4054770
17-03-25 21:51:58 [1] Step: 38600 Acc: 0.65563 0.85069 Cost: 0.93811 0.75129 0.10382 0.08300 Time: 0.00075
17-03-25 21:51:58 [1] Train Extra: lr=0.0000988 inv=0.4165625 sub=0.0000000
17-03-25 21:53:18 [1] Step: 38700 Acc: 0.65719 0.84010 Cost: 1.04455 0.74046 0.22106 0.08303 Time: 0.00069
17-03-25 21:53:18 [1] Train Extra: lr=0.0000985 inv=0.4503125 sub=0.0000000
17-03-25 21:54:40 [1] Step: 38800 Acc: 0.64625 0.84623 Cost: 0.90192 0.60505 0.21380 0.08307 Time: 0.00074
17-03-25 21:54:40 [1] Train Extra: lr=0.0000983 inv=0.4314062 sub=0.0000000
17-03-25 21:55:57 [1] Step: 38900 Acc: 0.66750 0.85043 Cost: 1.05040 0.86892 0.09836 0.08312 Time: 0.00074
17-03-25 21:55:57 [1] Train Extra: lr=0.0000980 inv=0.4126563 sub=0.0000000
17-03-25 21:57:17 [1] Step: 39000 Acc: 0.65281 0.84158 Cost: 1.37510 0.99890 0.29304 0.08316 Time: 0.00072
17-03-25 21:57:17 [1] Train Extra: lr=0.0000977 inv=0.4407813 sub=0.0000000
17-03-25 21:58:14 [1] Step: 39000 Eval acc: 0.66696 0.85278 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 21:58:14 [1] Eval Extra: inv=0.4177341
17-03-25 21:59:39 [1] Step: 39100 Acc: 0.65687 0.84961 Cost: 1.31214 1.03542 0.19357 0.08316 Time: 0.00076
17-03-25 21:59:39 [1] Train Extra: lr=0.0000974 inv=0.4135937 sub=0.0000000
17-03-25 22:01:04 [1] Step: 39200 Acc: 0.66938 0.84691 Cost: 1.20991 0.91050 0.21623 0.08318 Time: 0.00077
17-03-25 22:01:04 [1] Train Extra: lr=0.0000971 inv=0.4004687 sub=0.0000000
17-03-25 22:02:16 [1] Step: 39300 Acc: 0.65969 0.84883 Cost: 0.90985 0.65364 0.17302 0.08319 Time: 0.00071
17-03-25 22:02:16 [1] Train Extra: lr=0.0000969 inv=0.3918750 sub=0.0000000
17-03-25 22:03:39 [1] Step: 39400 Acc: 0.66531 0.84749 Cost: 1.10703 0.79534 0.22851 0.08318 Time: 0.00074
17-03-25 22:03:39 [1] Train Extra: lr=0.0000966 inv=0.4178125 sub=0.0000000
17-03-25 22:04:48 [1] Step: 39500 Acc: 0.66687 0.84518 Cost: 1.02487 0.81716 0.12441 0.08330 Time: 0.00069
17-03-25 22:04:48 [1] Train Extra: lr=0.0000963 inv=0.4139062 sub=0.0000000
17-03-25 22:05:45 [1] Step: 39500 Eval acc: 0.67016 0.84524 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 22:05:45 [1] Eval Extra: inv=0.4541740
17-03-25 22:07:10 [1] Step: 39600 Acc: 0.66031 0.84793 Cost: 1.18693 0.79335 0.31022 0.08337 Time: 0.00078
17-03-25 22:07:10 [1] Train Extra: lr=0.0000960 inv=0.4084375 sub=0.0000000
17-03-25 22:08:24 [1] Step: 39700 Acc: 0.64969 0.85232 Cost: 0.77096 0.58550 0.10210 0.08336 Time: 0.00074
17-03-25 22:08:24 [1] Train Extra: lr=0.0000957 inv=0.3882813 sub=0.0000000
17-03-25 22:09:50 [1] Step: 39800 Acc: 0.64438 0.84393 Cost: 1.17367 0.76362 0.32670 0.08335 Time: 0.00075
17-03-25 22:09:50 [1] Train Extra: lr=0.0000955 inv=0.4156250 sub=0.0000000
17-03-25 22:11:14 [1] Step: 39900 Acc: 0.64906 0.85599 Cost: 1.04839 0.73348 0.23152 0.08339 Time: 0.00077
17-03-25 22:11:14 [1] Train Extra: lr=0.0000952 inv=0.4062500 sub=0.0000000
17-03-25 22:12:38 [1] Step: 40000 Acc: 0.65438 0.85274 Cost: 1.26070 0.81632 0.36098 0.08340 Time: 0.00077
17-03-25 22:12:38 [1] Train Extra: lr=0.0000949 inv=0.4054687 sub=0.0000000
17-03-25 22:13:33 [1] Step: 40000 Eval acc: 0.66376 0.85663 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 22:13:33 [1] Eval Extra: inv=0.3473940
17-03-25 22:13:33 [1] Checkpointing.
17-03-25 22:14:49 [1] Step: 40100 Acc: 0.65375 0.84625 Cost: 1.02092 0.77770 0.15979 0.08344 Time: 0.00074
17-03-25 22:14:49 [1] Train Extra: lr=0.0000946 inv=0.4051562 sub=0.0000000
17-03-25 22:16:08 [1] Step: 40200 Acc: 0.66625 0.84622 Cost: 0.96090 0.71929 0.15813 0.08348 Time: 0.00072
17-03-25 22:16:08 [1] Train Extra: lr=0.0000944 inv=0.4253125 sub=0.0000000
17-03-25 22:17:33 [1] Step: 40300 Acc: 0.67000 0.83460 Cost: 1.07892 0.72947 0.26597 0.08348 Time: 0.00072
17-03-25 22:17:33 [1] Train Extra: lr=0.0000941 inv=0.4570312 sub=0.0000000
17-03-25 22:18:57 [1] Step: 40400 Acc: 0.63938 0.85092 Cost: 1.07970 0.82059 0.17561 0.08351 Time: 0.00075
17-03-25 22:18:57 [1] Train Extra: lr=0.0000938 inv=0.4081250 sub=0.0000000
17-03-25 22:20:19 [1] Step: 40500 Acc: 0.65469 0.84216 Cost: 1.30212 0.91771 0.30086 0.08355 Time: 0.00073
17-03-25 22:20:19 [1] Train Extra: lr=0.0000936 inv=0.4381250 sub=0.0000000
17-03-25 22:21:15 [1] Step: 40500 Eval acc: 0.66928 0.85428 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 22:21:15 [1] Eval Extra: inv=0.4057531
17-03-25 22:22:30 [1] Step: 40600 Acc: 0.65312 0.84486 Cost: 0.89827 0.69070 0.12399 0.08358 Time: 0.00071
17-03-25 22:22:30 [1] Train Extra: lr=0.0000933 inv=0.4229688 sub=0.0000000
17-03-25 22:23:54 [1] Step: 40700 Acc: 0.65594 0.85523 Cost: 1.06494 0.82851 0.15282 0.08360 Time: 0.00078
17-03-25 22:23:54 [1] Train Extra: lr=0.0000930 inv=0.4192187 sub=0.0000000
17-03-25 22:25:05 [1] Step: 40800 Acc: 0.65281 0.84835 Cost: 1.17150 0.77777 0.31004 0.08368 Time: 0.00073
17-03-25 22:25:05 [1] Train Extra: lr=0.0000928 inv=0.4023438 sub=0.0000000
17-03-25 22:26:23 [1] Step: 40900 Acc: 0.66844 0.84058 Cost: 1.02832 0.64302 0.30159 0.08372 Time: 0.00073
17-03-25 22:26:23 [1] Train Extra: lr=0.0000925 inv=0.4290625 sub=0.0000000
17-03-25 22:27:45 [1] Step: 41000 Acc: 0.64594 0.84558 Cost: 1.03403 0.70981 0.24048 0.08374 Time: 0.00073
17-03-25 22:27:45 [1] Train Extra: lr=0.0000922 inv=0.4250000 sub=0.0000000
17-03-25 22:28:42 [1] Step: 41000 Eval acc: 0.66409 0.84968 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 22:28:42 [1] Eval Extra: inv=0.4292182
17-03-25 22:30:03 [1] Step: 41100 Acc: 0.67281 0.84811 Cost: 1.09074 0.80079 0.20616 0.08379 Time: 0.00072
17-03-25 22:30:03 [1] Train Extra: lr=0.0000920 inv=0.4310937 sub=0.0000000
17-03-25 22:31:25 [1] Step: 41200 Acc: 0.64969 0.84111 Cost: 1.27769 0.92960 0.26425 0.08384 Time: 0.00074
17-03-25 22:31:25 [1] Train Extra: lr=0.0000917 inv=0.4334375 sub=0.0000000
17-03-25 22:32:43 [1] Step: 41300 Acc: 0.64406 0.84752 Cost: 1.19400 0.85752 0.25265 0.08383 Time: 0.00074
17-03-25 22:32:43 [1] Train Extra: lr=0.0000914 inv=0.4118750 sub=0.0000000
17-03-25 22:33:56 [1] Step: 41400 Acc: 0.65844 0.84513 Cost: 0.97492 0.76102 0.13008 0.08381 Time: 0.00071
17-03-25 22:33:56 [1] Train Extra: lr=0.0000912 inv=0.4112500 sub=0.0000000
17-03-25 22:35:20 [1] Step: 41500 Acc: 0.66469 0.85163 Cost: 1.25591 0.94724 0.22485 0.08382 Time: 0.00075
17-03-25 22:35:20 [1] Train Extra: lr=0.0000909 inv=0.4259375 sub=0.0000000
17-03-25 22:36:17 [1] Step: 41500 Eval acc: 0.66829 0.85018 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 22:36:17 [1] Eval Extra: inv=0.3815150
17-03-25 22:37:36 [1] Step: 41600 Acc: 0.65625 0.85206 Cost: 1.05874 0.86165 0.11325 0.08383 Time: 0.00075
17-03-25 22:37:36 [1] Train Extra: lr=0.0000907 inv=0.4118750 sub=0.0000000
17-03-25 22:38:59 [1] Step: 41700 Acc: 0.65531 0.84591 Cost: 1.21401 0.83638 0.29385 0.08378 Time: 0.00073
17-03-25 22:38:59 [1] Train Extra: lr=0.0000904 inv=0.4476562 sub=0.0000000
17-03-25 22:40:32 [1] Step: 41800 Acc: 0.65594 0.85524 Cost: 1.11970 0.81990 0.21598 0.08383 Time: 0.00078
17-03-25 22:40:32 [1] Train Extra: lr=0.0000901 inv=0.4353125 sub=0.0000000
17-03-25 22:42:03 [1] Step: 41900 Acc: 0.64062 0.84734 Cost: 1.05028 0.71607 0.25037 0.08385 Time: 0.00075
17-03-25 22:42:03 [1] Train Extra: lr=0.0000899 inv=0.4606250 sub=0.0000000
17-03-25 22:43:34 [1] Step: 42000 Acc: 0.66781 0.84401 Cost: 1.19143 0.78885 0.31863 0.08395 Time: 0.00079
17-03-25 22:43:34 [1] Train Extra: lr=0.0000896 inv=0.4356250 sub=0.0000000
17-03-25 22:44:32 [1] Step: 42000 Eval acc: 0.66508 0.85462 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 22:44:32 [1] Eval Extra: inv=0.4085689
17-03-25 22:45:52 [1] Step: 42100 Acc: 0.69281 0.84710 Cost: 0.95708 0.59600 0.27695 0.08414 Time: 0.00074
17-03-25 22:45:52 [1] Train Extra: lr=0.0000894 inv=0.4264062 sub=0.0000000
17-03-25 22:47:12 [1] Step: 42200 Acc: 0.66531 0.84841 Cost: 1.04107 0.69371 0.26315 0.08421 Time: 0.00074
17-03-25 22:47:12 [1] Train Extra: lr=0.0000891 inv=0.4156250 sub=0.0000000
17-03-25 22:48:45 [1] Step: 42300 Acc: 0.66406 0.84575 Cost: 1.31147 0.94006 0.28709 0.08432 Time: 0.00078
17-03-25 22:48:45 [1] Train Extra: lr=0.0000888 inv=0.4290625 sub=0.0000000
17-03-25 22:50:10 [1] Step: 42400 Acc: 0.66312 0.83455 Cost: 1.07765 0.71266 0.28055 0.08444 Time: 0.00072
17-03-25 22:50:10 [1] Train Extra: lr=0.0000886 inv=0.4459375 sub=0.0000000
17-03-25 22:51:29 [1] Step: 42500 Acc: 0.67500 0.84935 Cost: 0.78860 0.55052 0.15355 0.08453 Time: 0.00074
17-03-25 22:51:29 [1] Train Extra: lr=0.0000883 inv=0.4109375 sub=0.0000000
17-03-25 22:52:26 [1] Step: 42500 Eval acc: 0.66630 0.85161 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 22:52:26 [1] Eval Extra: inv=0.4141453
17-03-25 22:53:33 [1] Step: 42600 Acc: 0.69344 0.85265 Cost: 1.00485 0.69428 0.22595 0.08462 Time: 0.00069
17-03-25 22:53:33 [1] Train Extra: lr=0.0000881 inv=0.3821875 sub=0.0000000
17-03-25 22:55:02 [1] Step: 42700 Acc: 0.67000 0.85119 Cost: 1.08987 0.75341 0.25173 0.08472 Time: 0.00078
17-03-25 22:55:02 [1] Train Extra: lr=0.0000878 inv=0.4226563 sub=0.0000000
17-03-25 22:56:28 [1] Step: 42800 Acc: 0.67719 0.84260 Cost: 1.28720 0.93869 0.26375 0.08475 Time: 0.00073
17-03-25 22:56:28 [1] Train Extra: lr=0.0000876 inv=0.4417187 sub=0.0000000
17-03-25 22:57:59 [1] Step: 42900 Acc: 0.67375 0.84270 Cost: 1.31367 0.96025 0.26855 0.08487 Time: 0.00075
17-03-25 22:57:59 [1] Train Extra: lr=0.0000873 inv=0.4634375 sub=0.0000000
17-03-25 22:59:17 [1] Step: 43000 Acc: 0.65094 0.85083 Cost: 0.92182 0.67136 0.16558 0.08488 Time: 0.00073
17-03-25 22:59:17 [1] Train Extra: lr=0.0000871 inv=0.4417187 sub=0.0000000
17-03-25 23:00:13 [1] Step: 43000 Eval acc: 0.66718 0.85437 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 23:00:13 [1] Eval Extra: inv=0.4044280
17-03-25 23:01:43 [1] Step: 43100 Acc: 0.67281 0.84142 Cost: 1.13831 0.79687 0.25647 0.08497 Time: 0.00075
17-03-25 23:01:43 [1] Train Extra: lr=0.0000868 inv=0.4540625 sub=0.0000000
17-03-25 23:03:01 [1] Step: 43200 Acc: 0.66719 0.84347 Cost: 1.10008 0.76126 0.25377 0.08505 Time: 0.00074
17-03-25 23:03:01 [1] Train Extra: lr=0.0000866 inv=0.4179688 sub=0.0000000
17-03-25 23:04:15 [1] Step: 43300 Acc: 0.67312 0.84579 Cost: 1.02092 0.77520 0.16051 0.08522 Time: 0.00070
17-03-25 23:04:15 [1] Train Extra: lr=0.0000863 inv=0.4132812 sub=0.0000000
17-03-25 23:05:37 [1] Step: 43400 Acc: 0.66156 0.84785 Cost: 1.25393 0.86444 0.30420 0.08529 Time: 0.00075
17-03-25 23:05:37 [1] Train Extra: lr=0.0000861 inv=0.4232812 sub=0.0000000
17-03-25 23:06:53 [1] Step: 43500 Acc: 0.67625 0.85213 Cost: 1.58925 1.30029 0.20358 0.08537 Time: 0.00072
17-03-25 23:06:53 [1] Train Extra: lr=0.0000858 inv=0.4064063 sub=0.0000000
17-03-25 23:07:49 [1] Step: 43500 Eval acc: 0.66884 0.84436 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 23:07:49 [1] Eval Extra: inv=0.4849823
17-03-25 23:09:11 [1] Step: 43600 Acc: 0.66531 0.83683 Cost: 1.10227 0.84237 0.17450 0.08539 Time: 0.00073
17-03-25 23:09:11 [1] Train Extra: lr=0.0000856 inv=0.4564063 sub=0.0000000
17-03-25 23:10:37 [1] Step: 43700 Acc: 0.66687 0.85461 Cost: 1.27054 0.88766 0.29737 0.08551 Time: 0.00075
17-03-25 23:10:37 [1] Train Extra: lr=0.0000853 inv=0.4264062 sub=0.0000000
17-03-25 23:11:55 [1] Step: 43800 Acc: 0.66406 0.84694 Cost: 1.07984 0.73833 0.25588 0.08563 Time: 0.00072
17-03-25 23:11:55 [1] Train Extra: lr=0.0000851 inv=0.4373437 sub=0.0000000
17-03-25 23:13:18 [1] Step: 43900 Acc: 0.66438 0.84408 Cost: 0.99206 0.71967 0.18672 0.08567 Time: 0.00074
17-03-25 23:13:18 [1] Train Extra: lr=0.0000848 inv=0.4450000 sub=0.0000000
17-03-25 23:14:42 [1] Step: 44000 Acc: 0.67063 0.84818 Cost: 0.93856 0.64411 0.20873 0.08572 Time: 0.00075
17-03-25 23:14:42 [1] Train Extra: lr=0.0000846 inv=0.4354688 sub=0.0000000
17-03-25 23:15:40 [1] Step: 44000 Eval acc: 0.66807 0.85167 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 23:15:40 [1] Eval Extra: inv=0.3971400
17-03-25 23:17:01 [1] Step: 44100 Acc: 0.68500 0.85184 Cost: 1.22284 0.87646 0.26057 0.08581 Time: 0.00074
17-03-25 23:17:01 [1] Train Extra: lr=0.0000844 inv=0.4106250 sub=0.0000000
17-03-25 23:18:28 [1] Step: 44200 Acc: 0.67750 0.84563 Cost: 1.06784 0.75838 0.22350 0.08596 Time: 0.00075
17-03-25 23:18:28 [1] Train Extra: lr=0.0000841 inv=0.4376563 sub=0.0000000
17-03-25 23:19:45 [1] Step: 44300 Acc: 0.67937 0.85072 Cost: 1.03006 0.66118 0.28291 0.08597 Time: 0.00074
17-03-25 23:19:45 [1] Train Extra: lr=0.0000839 inv=0.3915625 sub=0.0000000
17-03-25 23:20:58 [1] Step: 44400 Acc: 0.67375 0.84742 Cost: 1.19213 0.97862 0.12742 0.08609 Time: 0.00070
17-03-25 23:20:58 [1] Train Extra: lr=0.0000836 inv=0.4293750 sub=0.0000000
17-03-25 23:22:17 [1] Step: 44500 Acc: 0.65281 0.85344 Cost: 1.09477 0.83344 0.17522 0.08612 Time: 0.00075
17-03-25 23:22:17 [1] Train Extra: lr=0.0000834 inv=0.3942188 sub=0.0000000
17-03-25 23:23:13 [1] Step: 44500 Eval acc: 0.66718 0.84536 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 23:23:13 [1] Eval Extra: inv=0.4659342
17-03-25 23:24:31 [1] Step: 44600 Acc: 0.67875 0.85711 Cost: 1.02312 0.80828 0.12873 0.08611 Time: 0.00077
17-03-25 23:24:31 [1] Train Extra: lr=0.0000832 inv=0.3928125 sub=0.0000000
17-03-25 23:25:38 [1] Step: 44700 Acc: 0.67031 0.84839 Cost: 1.15369 0.89068 0.17689 0.08612 Time: 0.00070
17-03-25 23:25:38 [1] Train Extra: lr=0.0000829 inv=0.3989063 sub=0.0000000
17-03-25 23:27:01 [1] Step: 44800 Acc: 0.67937 0.85025 Cost: 1.26361 0.99270 0.18475 0.08617 Time: 0.00077
17-03-25 23:27:01 [1] Train Extra: lr=0.0000827 inv=0.3990625 sub=0.0000000
17-03-25 23:28:22 [1] Step: 44900 Acc: 0.66750 0.83795 Cost: 0.85198 0.60627 0.15945 0.08627 Time: 0.00070
17-03-25 23:28:22 [1] Train Extra: lr=0.0000824 inv=0.4373437 sub=0.0000000
17-03-25 23:29:39 [1] Step: 45000 Acc: 0.68219 0.85278 Cost: 0.91474 0.65700 0.17141 0.08633 Time: 0.00074
17-03-25 23:29:39 [1] Train Extra: lr=0.0000822 inv=0.3823437 sub=0.0000000
17-03-25 23:30:37 [1] Step: 45000 Eval acc: 0.66354 0.84657 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 23:30:37 [1] Eval Extra: inv=0.4165746
17-03-25 23:30:37 [1] Checkpointing.
17-03-25 23:31:55 [1] Step: 45100 Acc: 0.66844 0.84490 Cost: 1.01853 0.73188 0.20026 0.08640 Time: 0.00073
17-03-25 23:31:55 [1] Train Extra: lr=0.0000820 inv=0.4196875 sub=0.0000000
17-03-25 23:33:21 [1] Step: 45200 Acc: 0.65812 0.84894 Cost: 1.30858 0.96512 0.25702 0.08645 Time: 0.00076
17-03-25 23:33:21 [1] Train Extra: lr=0.0000817 inv=0.4351563 sub=0.0000000
17-03-25 23:34:37 [1] Step: 45300 Acc: 0.65656 0.84906 Cost: 1.19067 0.90138 0.20279 0.08649 Time: 0.00073
17-03-25 23:34:37 [1] Train Extra: lr=0.0000815 inv=0.4109375 sub=0.0000000
17-03-25 23:35:59 [1] Step: 45400 Acc: 0.66938 0.84265 Cost: 0.88759 0.58931 0.21166 0.08661 Time: 0.00074
17-03-25 23:35:59 [1] Train Extra: lr=0.0000813 inv=0.4409375 sub=0.0000000
17-03-25 23:37:18 [1] Step: 45500 Acc: 0.67281 0.84835 Cost: 1.32741 0.92532 0.31543 0.08665 Time: 0.00074
17-03-25 23:37:18 [1] Train Extra: lr=0.0000810 inv=0.4201563 sub=0.0000000
17-03-25 23:38:14 [1] Step: 45500 Eval acc: 0.67005 0.85334 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 23:38:14 [1] Eval Extra: inv=0.4097836
17-03-25 23:39:39 [1] Step: 45600 Acc: 0.66031 0.83707 Cost: 0.85860 0.61402 0.15780 0.08678 Time: 0.00073
17-03-25 23:39:39 [1] Train Extra: lr=0.0000808 inv=0.4548437 sub=0.0000000
17-03-25 23:41:00 [1] Step: 45700 Acc: 0.67531 0.84552 Cost: 0.94410 0.60809 0.24922 0.08679 Time: 0.00072
17-03-25 23:41:00 [1] Train Extra: lr=0.0000806 inv=0.4495312 sub=0.0000000
17-03-25 23:42:19 [1] Step: 45800 Acc: 0.68750 0.84989 Cost: 1.17974 0.94697 0.14597 0.08681 Time: 0.00074
17-03-25 23:42:19 [1] Train Extra: lr=0.0000803 inv=0.4045313 sub=0.0000000
17-03-25 23:43:44 [1] Step: 45900 Acc: 0.66500 0.84530 Cost: 0.90645 0.65953 0.16000 0.08692 Time: 0.00076
17-03-25 23:43:44 [1] Train Extra: lr=0.0000801 inv=0.4225000 sub=0.0000000
17-03-25 23:45:07 [1] Step: 46000 Acc: 0.66281 0.84661 Cost: 0.90243 0.62066 0.19482 0.08695 Time: 0.00074
17-03-25 23:45:07 [1] Train Extra: lr=0.0000799 inv=0.4407813 sub=0.0000000
17-03-25 23:46:03 [1] Step: 46000 Eval acc: 0.66928 0.85162 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-25 23:46:03 [1] Eval Extra: inv=0.4096731
17-03-25 23:47:19 [1] Step: 46100 Acc: 0.67281 0.85450 Cost: 1.09099 0.85763 0.14635 0.08701 Time: 0.00076
17-03-25 23:47:19 [1] Train Extra: lr=0.0000796 inv=0.3820312 sub=0.0000000
17-03-25 23:48:37 [1] Step: 46200 Acc: 0.67469 0.84571 Cost: 1.04569 0.79676 0.16178 0.08715 Time: 0.00074
17-03-25 23:48:37 [1] Train Extra: lr=0.0000794 inv=0.4115625 sub=0.0000000
17-03-25 23:49:58 [1] Step: 46300 Acc: 0.65594 0.85080 Cost: 0.76377 0.51065 0.16587 0.08725 Time: 0.00074
17-03-25 23:49:58 [1] Train Extra: lr=0.0000792 inv=0.4229688 sub=0.0000000
17-03-25 23:51:29 [1] Step: 46400 Acc: 0.66281 0.84990 Cost: 1.07324 0.73657 0.24941 0.08726 Time: 0.00079
17-03-25 23:51:29 [1] Train Extra: lr=0.0000790 inv=0.4295312 sub=0.0000000
17-03-25 23:52:55 [1] Step: 46500 Acc: 0.67875 0.85438 Cost: 1.09298 0.79033 0.21538 0.08727 Time: 0.00078
17-03-25 23:52:55 [1] Train Extra: lr=0.0000787 inv=0.4156250 sub=0.0000000
17-03-25 23:53:53 [1] Step: 46500 Eval acc: 0.67204 0.85053 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-25 23:53:53 [1] Eval Extra: inv=0.4194457
17-03-25 23:55:06 [1] Step: 46600 Acc: 0.66906 0.85130 Cost: 1.03536 0.81561 0.13240 0.08735 Time: 0.00071
17-03-25 23:55:06 [1] Train Extra: lr=0.0000785 inv=0.3981250 sub=0.0000000
17-03-25 23:56:43 [1] Step: 46700 Acc: 0.67281 0.85631 Cost: 0.76623 0.52755 0.15126 0.08742 Time: 0.00081
17-03-25 23:56:43 [1] Train Extra: lr=0.0000783 inv=0.4204688 sub=0.0000000
17-03-25 23:58:04 [1] Step: 46800 Acc: 0.67031 0.84636 Cost: 1.01066 0.72842 0.19476 0.08749 Time: 0.00073
17-03-25 23:58:04 [1] Train Extra: lr=0.0000781 inv=0.4278125 sub=0.0000000
17-03-25 23:59:22 [1] Step: 46900 Acc: 0.66844 0.84495 Cost: 0.99637 0.68590 0.22293 0.08753 Time: 0.00073
17-03-25 23:59:22 [1] Train Extra: lr=0.0000778 inv=0.4184375 sub=0.0000000
17-03-26 00:00:45 [1] Step: 47000 Acc: 0.65187 0.85023 Cost: 0.98275 0.65273 0.24253 0.08749 Time: 0.00076
17-03-26 00:00:45 [1] Train Extra: lr=0.0000776 inv=0.4215625 sub=0.0000000
17-03-26 00:01:41 [1] Step: 47000 Eval acc: 0.66906 0.85418 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 00:01:41 [1] Eval Extra: inv=0.3984099
17-03-26 00:03:00 [1] Step: 47100 Acc: 0.66875 0.84787 Cost: 1.11944 0.72646 0.30542 0.08755 Time: 0.00074
17-03-26 00:03:00 [1] Train Extra: lr=0.0000774 inv=0.4210937 sub=0.0000000
17-03-26 00:04:28 [1] Step: 47200 Acc: 0.65312 0.84043 Cost: 1.18674 0.87457 0.22461 0.08756 Time: 0.00075
17-03-26 00:04:28 [1] Train Extra: lr=0.0000772 inv=0.4448437 sub=0.0000000
17-03-26 00:05:53 [1] Step: 47300 Acc: 0.66094 0.84559 Cost: 0.96820 0.71162 0.16903 0.08755 Time: 0.00075
17-03-26 00:05:53 [1] Train Extra: lr=0.0000769 inv=0.4154687 sub=0.0000000
17-03-26 00:07:11 [1] Step: 47400 Acc: 0.66719 0.85135 Cost: 1.19342 0.83805 0.26779 0.08758 Time: 0.00072
17-03-26 00:07:11 [1] Train Extra: lr=0.0000767 inv=0.4162500 sub=0.0000000
17-03-26 00:08:23 [1] Step: 47500 Acc: 0.65938 0.85056 Cost: 1.15599 0.76922 0.29909 0.08769 Time: 0.00071
17-03-26 00:08:23 [1] Train Extra: lr=0.0000765 inv=0.3932813 sub=0.0000000
17-03-26 00:09:20 [1] Step: 47500 Eval acc: 0.67281 0.85322 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 00:09:20 [1] Eval Extra: inv=0.3978578
17-03-26 00:10:51 [1] Step: 47600 Acc: 0.66719 0.84539 Cost: 1.04288 0.72779 0.22738 0.08772 Time: 0.00078
17-03-26 00:10:51 [1] Train Extra: lr=0.0000763 inv=0.4357813 sub=0.0000000
17-03-26 00:12:16 [1] Step: 47700 Acc: 0.65563 0.84633 Cost: 1.08927 0.76703 0.23446 0.08778 Time: 0.00076
17-03-26 00:12:16 [1] Train Extra: lr=0.0000761 inv=0.4340625 sub=0.0000000
17-03-26 00:13:37 [1] Step: 47800 Acc: 0.65469 0.84060 Cost: 1.15238 0.73104 0.33361 0.08773 Time: 0.00072
17-03-26 00:13:37 [1] Train Extra: lr=0.0000758 inv=0.4262500 sub=0.0000000
17-03-26 00:15:00 [1] Step: 47900 Acc: 0.66469 0.84260 Cost: 1.21581 0.93840 0.18969 0.08771 Time: 0.00076
17-03-26 00:15:00 [1] Train Extra: lr=0.0000756 inv=0.4312500 sub=0.0000000
17-03-26 00:16:20 [1] Step: 48000 Acc: 0.66344 0.84905 Cost: 0.97196 0.70294 0.18129 0.08774 Time: 0.00075
17-03-26 00:16:20 [1] Train Extra: lr=0.0000754 inv=0.4023438 sub=0.0000000
17-03-26 00:17:17 [1] Step: 48000 Eval acc: 0.67281 0.85555 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 00:17:17 [1] Eval Extra: inv=0.4287213
17-03-26 00:18:43 [1] Step: 48100 Acc: 0.65063 0.83908 Cost: 1.00698 0.71265 0.20644 0.08789 Time: 0.00074
17-03-26 00:18:43 [1] Train Extra: lr=0.0000752 inv=0.4545312 sub=0.0000000
17-03-26 00:20:11 [1] Step: 48200 Acc: 0.66438 0.85035 Cost: 1.10366 0.80457 0.21113 0.08795 Time: 0.00077
17-03-26 00:20:11 [1] Train Extra: lr=0.0000750 inv=0.4231250 sub=0.0000000
17-03-26 00:21:44 [1] Step: 48300 Acc: 0.66344 0.83915 Cost: 1.00218 0.69773 0.21649 0.08796 Time: 0.00075
17-03-26 00:21:44 [1] Train Extra: lr=0.0000748 inv=0.4525000 sub=0.0000000
17-03-26 00:23:02 [1] Step: 48400 Acc: 0.66594 0.85933 Cost: 1.01787 0.83658 0.09332 0.08797 Time: 0.00077
17-03-26 00:23:02 [1] Train Extra: lr=0.0000745 inv=0.3882813 sub=0.0000000
17-03-26 00:24:37 [1] Step: 48500 Acc: 0.65594 0.84822 Cost: 1.10367 0.73796 0.27766 0.08804 Time: 0.00079
17-03-26 00:24:37 [1] Train Extra: lr=0.0000743 inv=0.4537500 sub=0.0000000
17-03-26 00:25:35 [1] Step: 48500 Eval acc: 0.67392 0.85137 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 00:25:35 [1] Eval Extra: inv=0.4437390
17-03-26 00:27:02 [1] Step: 48600 Acc: 0.65375 0.84632 Cost: 1.46308 1.08038 0.29464 0.08807 Time: 0.00074
17-03-26 00:27:02 [1] Train Extra: lr=0.0000741 inv=0.4406250 sub=0.0000000
17-03-26 00:28:17 [1] Step: 48700 Acc: 0.66344 0.84688 Cost: 1.16598 0.87355 0.20434 0.08809 Time: 0.00073
17-03-26 00:28:17 [1] Train Extra: lr=0.0000739 inv=0.3998437 sub=0.0000000
17-03-26 00:29:43 [1] Step: 48800 Acc: 0.66969 0.84740 Cost: 1.14229 0.91262 0.14165 0.08803 Time: 0.00074
17-03-26 00:29:43 [1] Train Extra: lr=0.0000737 inv=0.4289062 sub=0.0000000
17-03-26 00:31:04 [1] Step: 48900 Acc: 0.66750 0.85301 Cost: 0.76669 0.57845 0.10018 0.08806 Time: 0.00076
17-03-26 00:31:04 [1] Train Extra: lr=0.0000735 inv=0.4376563 sub=0.0000000
17-03-26 00:32:31 [1] Step: 49000 Acc: 0.65969 0.84740 Cost: 1.12588 0.82730 0.21036 0.08821 Time: 0.00074
17-03-26 00:32:31 [1] Train Extra: lr=0.0000733 inv=0.4556250 sub=0.0000000
17-03-26 00:33:29 [1] Step: 49000 Eval acc: 0.67259 0.85460 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 00:33:29 [1] Eval Extra: inv=0.3809077
17-03-26 00:34:51 [1] Step: 49100 Acc: 0.67125 0.84064 Cost: 1.04995 0.68599 0.27574 0.08822 Time: 0.00075
17-03-26 00:34:51 [1] Train Extra: lr=0.0000731 inv=0.4142188 sub=0.0000000
17-03-26 00:36:03 [1] Step: 49200 Acc: 0.66687 0.84412 Cost: 1.14852 0.80534 0.25490 0.08828 Time: 0.00070
17-03-26 00:36:03 [1] Train Extra: lr=0.0000728 inv=0.4271875 sub=0.0000000
17-03-26 00:37:22 [1] Step: 49300 Acc: 0.66281 0.84458 Cost: 0.93342 0.75151 0.09366 0.08826 Time: 0.00072
17-03-26 00:37:22 [1] Train Extra: lr=0.0000726 inv=0.4487500 sub=0.0000000
17-03-26 00:38:41 [1] Step: 49400 Acc: 0.65812 0.84729 Cost: 1.10004 0.78811 0.22363 0.08831 Time: 0.00074
17-03-26 00:38:41 [1] Train Extra: lr=0.0000724 inv=0.4085937 sub=0.0000000
17-03-26 00:40:03 [1] Step: 49500 Acc: 0.67188 0.85157 Cost: 1.07904 0.74903 0.24167 0.08834 Time: 0.00076
17-03-26 00:40:03 [1] Train Extra: lr=0.0000722 inv=0.4145313 sub=0.0000000
17-03-26 00:40:58 [1] Step: 49500 Eval acc: 0.67856 0.84797 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-26 00:40:58 [1] Eval Extra: inv=0.4293838
17-03-26 00:40:58 [1] Checkpointing with new best dev accuracy of 0.678556
17-03-26 00:42:19 [1] Step: 49600 Acc: 0.66875 0.85065 Cost: 0.91674 0.70783 0.12046 0.08844 Time: 0.00073
17-03-26 00:42:19 [1] Train Extra: lr=0.0000720 inv=0.4217187 sub=0.0000000
17-03-26 00:43:37 [1] Step: 49700 Acc: 0.67031 0.84497 Cost: 1.20016 0.96946 0.14228 0.08843 Time: 0.00073
17-03-26 00:43:37 [1] Train Extra: lr=0.0000718 inv=0.4306250 sub=0.0000000
17-03-26 00:44:55 [1] Step: 49800 Acc: 0.66969 0.84968 Cost: 0.97069 0.58524 0.29702 0.08842 Time: 0.00073
17-03-26 00:44:55 [1] Train Extra: lr=0.0000716 inv=0.4084375 sub=0.0000000
17-03-26 00:46:24 [1] Step: 49900 Acc: 0.65844 0.85012 Cost: 1.18561 0.82864 0.26852 0.08845 Time: 0.00078
17-03-26 00:46:24 [1] Train Extra: lr=0.0000714 inv=0.4337500 sub=0.0000000
17-03-26 00:47:44 [1] Step: 50000 Acc: 0.68875 0.84404 Cost: 0.91180 0.54227 0.28097 0.08857 Time: 0.00074
17-03-26 00:47:44 [1] Train Extra: lr=0.0000712 inv=0.4215625 sub=0.0000000
17-03-26 00:48:42 [1] Step: 50000 Eval acc: 0.67259 0.85289 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 00:48:42 [1] Eval Extra: inv=0.4052562
17-03-26 00:48:42 [1] Checkpointing.
17-03-26 00:50:05 [1] Step: 50100 Acc: 0.66219 0.84067 Cost: 1.21949 0.83008 0.30083 0.08858 Time: 0.00074
17-03-26 00:50:05 [1] Train Extra: lr=0.0000710 inv=0.4317187 sub=0.0000000
17-03-26 00:51:24 [1] Step: 50200 Acc: 0.64938 0.85076 Cost: 1.14547 0.90150 0.15536 0.08862 Time: 0.00073
17-03-26 00:51:24 [1] Train Extra: lr=0.0000708 inv=0.4046875 sub=0.0000000
17-03-26 00:52:50 [1] Step: 50300 Acc: 0.67406 0.84726 Cost: 1.25737 0.91610 0.25264 0.08862 Time: 0.00075
17-03-26 00:52:50 [1] Train Extra: lr=0.0000706 inv=0.4279688 sub=0.0000000
17-03-26 00:54:05 [1] Step: 50400 Acc: 0.68969 0.84838 Cost: 0.90090 0.57974 0.23250 0.08865 Time: 0.00072
17-03-26 00:54:05 [1] Train Extra: lr=0.0000704 inv=0.3998437 sub=0.0000000
17-03-26 00:55:36 [1] Step: 50500 Acc: 0.68375 0.84073 Cost: 1.05656 0.74468 0.22309 0.08879 Time: 0.00075
17-03-26 00:55:36 [1] Train Extra: lr=0.0000702 inv=0.4343750 sub=0.0000000
17-03-26 00:56:34 [1] Step: 50500 Eval acc: 0.67105 0.85573 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 00:56:34 [1] Eval Extra: inv=0.3761595
17-03-26 00:58:00 [1] Step: 50600 Acc: 0.67250 0.84265 Cost: 0.93287 0.76757 0.07648 0.08882 Time: 0.00074
17-03-26 00:58:00 [1] Train Extra: lr=0.0000700 inv=0.4351563 sub=0.0000000
17-03-26 00:59:12 [1] Step: 50700 Acc: 0.69281 0.84802 Cost: 0.79453 0.53867 0.16691 0.08895 Time: 0.00072
17-03-26 00:59:12 [1] Train Extra: lr=0.0000698 inv=0.4089063 sub=0.0000000
17-03-26 01:00:31 [1] Step: 50800 Acc: 0.67594 0.84181 Cost: 0.80563 0.64195 0.07460 0.08908 Time: 0.00073
17-03-26 01:00:31 [1] Train Extra: lr=0.0000696 inv=0.4267187 sub=0.0000000
17-03-26 01:01:47 [1] Step: 50900 Acc: 0.68812 0.84178 Cost: 0.91608 0.56079 0.26617 0.08912 Time: 0.00071
17-03-26 01:01:47 [1] Train Extra: lr=0.0000694 inv=0.4204688 sub=0.0000000
17-03-26 01:03:16 [1] Step: 51000 Acc: 0.68844 0.85199 Cost: 1.03532 0.79109 0.15499 0.08924 Time: 0.00076
17-03-26 01:03:16 [1] Train Extra: lr=0.0000692 inv=0.4439063 sub=0.0000000
17-03-26 01:04:13 [1] Step: 51000 Eval acc: 0.67303 0.85618 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 01:04:13 [1] Eval Extra: inv=0.4047593
17-03-26 01:05:38 [1] Step: 51100 Acc: 0.68219 0.84197 Cost: 1.07931 0.66050 0.32946 0.08935 Time: 0.00073
17-03-26 01:05:38 [1] Train Extra: lr=0.0000690 inv=0.4437500 sub=0.0000000
17-03-26 01:07:07 [1] Step: 51200 Acc: 0.69125 0.84816 Cost: 0.80127 0.55645 0.15536 0.08946 Time: 0.00077
17-03-26 01:07:07 [1] Train Extra: lr=0.0000688 inv=0.4375000 sub=0.0000000
17-03-26 01:08:22 [1] Step: 51300 Acc: 0.68063 0.85021 Cost: 0.95337 0.70481 0.15903 0.08953 Time: 0.00072
17-03-26 01:08:22 [1] Train Extra: lr=0.0000686 inv=0.4121875 sub=0.0000000
17-03-26 01:09:46 [1] Step: 51400 Acc: 0.68875 0.85267 Cost: 0.96700 0.61635 0.26099 0.08966 Time: 0.00077
17-03-26 01:09:46 [1] Train Extra: lr=0.0000684 inv=0.4054687 sub=0.0000000
17-03-26 01:11:01 [1] Step: 51500 Acc: 0.69312 0.84732 Cost: 0.83603 0.64774 0.09856 0.08974 Time: 0.00072
17-03-26 01:11:01 [1] Train Extra: lr=0.0000682 inv=0.4165625 sub=0.0000000
17-03-26 01:11:58 [1] Step: 51500 Eval acc: 0.67171 0.84732 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 01:11:58 [1] Eval Extra: inv=0.4133171
17-03-26 01:13:16 [1] Step: 51600 Acc: 0.66938 0.84349 Cost: 0.91852 0.57506 0.25360 0.08986 Time: 0.00072
17-03-26 01:13:16 [1] Train Extra: lr=0.0000680 inv=0.4218750 sub=0.0000000
17-03-26 01:14:30 [1] Step: 51700 Acc: 0.68781 0.85743 Cost: 1.32740 0.92858 0.30892 0.08990 Time: 0.00075
17-03-26 01:14:30 [1] Train Extra: lr=0.0000678 inv=0.3976562 sub=0.0000000
17-03-26 01:15:55 [1] Step: 51800 Acc: 0.68437 0.85532 Cost: 1.16278 0.79675 0.27605 0.08998 Time: 0.00077
17-03-26 01:15:55 [1] Train Extra: lr=0.0000676 inv=0.4176563 sub=0.0000000
17-03-26 01:17:23 [1] Step: 51900 Acc: 0.67219 0.85512 Cost: 0.89559 0.50533 0.30022 0.09005 Time: 0.00076
17-03-26 01:17:23 [1] Train Extra: lr=0.0000674 inv=0.4325000 sub=0.0000000
17-03-26 01:18:41 [1] Step: 52000 Acc: 0.68563 0.85063 Cost: 0.90800 0.62130 0.19653 0.09017 Time: 0.00075
17-03-26 01:18:41 [1] Train Extra: lr=0.0000672 inv=0.4054687 sub=0.0000000
17-03-26 01:19:39 [1] Step: 52000 Eval acc: 0.67535 0.85345 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 01:19:39 [1] Eval Extra: inv=0.4365614
17-03-26 01:21:04 [1] Step: 52100 Acc: 0.67750 0.84726 Cost: 1.16951 0.86646 0.21288 0.09016 Time: 0.00074
17-03-26 01:21:04 [1] Train Extra: lr=0.0000670 inv=0.4248438 sub=0.0000000
17-03-26 01:22:23 [1] Step: 52200 Acc: 0.68250 0.84406 Cost: 0.96839 0.68724 0.19095 0.09020 Time: 0.00073
17-03-26 01:22:23 [1] Train Extra: lr=0.0000668 inv=0.4075000 sub=0.0000000
17-03-26 01:23:52 [1] Step: 52300 Acc: 0.67469 0.83904 Cost: 1.02850 0.77153 0.16669 0.09029 Time: 0.00075
17-03-26 01:23:52 [1] Train Extra: lr=0.0000666 inv=0.4595313 sub=0.0000000
17-03-26 01:25:11 [1] Step: 52400 Acc: 0.68094 0.84696 Cost: 0.76880 0.55877 0.11965 0.09038 Time: 0.00072
17-03-26 01:25:11 [1] Train Extra: lr=0.0000664 inv=0.4079687 sub=0.0000000
17-03-26 01:26:30 [1] Step: 52500 Acc: 0.66969 0.84530 Cost: 1.16767 0.84132 0.23586 0.09048 Time: 0.00072
17-03-26 01:26:30 [1] Train Extra: lr=0.0000663 inv=0.4317187 sub=0.0000000
17-03-26 01:27:30 [1] Step: 52500 Eval acc: 0.67878 0.84957 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00018
17-03-26 01:27:30 [1] Eval Extra: inv=0.4770318
17-03-26 01:28:51 [1] Step: 52600 Acc: 0.68219 0.85397 Cost: 1.00886 0.75609 0.16220 0.09057 Time: 0.00075
17-03-26 01:28:51 [1] Train Extra: lr=0.0000661 inv=0.3896875 sub=0.0000000
17-03-26 01:30:16 [1] Step: 52700 Acc: 0.67344 0.84819 Cost: 0.97582 0.75798 0.12722 0.09062 Time: 0.00075
17-03-26 01:30:16 [1] Train Extra: lr=0.0000659 inv=0.4271875 sub=0.0000000
17-03-26 01:31:35 [1] Step: 52800 Acc: 0.67344 0.83955 Cost: 1.24829 0.93018 0.22736 0.09075 Time: 0.00072
17-03-26 01:31:35 [1] Train Extra: lr=0.0000657 inv=0.4315625 sub=0.0000000
17-03-26 01:32:54 [1] Step: 52900 Acc: 0.67594 0.84859 Cost: 1.03128 0.72868 0.21185 0.09075 Time: 0.00073
17-03-26 01:32:54 [1] Train Extra: lr=0.0000655 inv=0.4071875 sub=0.0000000
17-03-26 01:34:23 [1] Step: 53000 Acc: 0.69594 0.84660 Cost: 1.15265 0.89573 0.16607 0.09086 Time: 0.00078
17-03-26 01:34:23 [1] Train Extra: lr=0.0000653 inv=0.4251563 sub=0.0000000
17-03-26 01:35:19 [1] Step: 53000 Eval acc: 0.67458 0.84620 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 01:35:19 [1] Eval Extra: inv=0.4329174
17-03-26 01:36:50 [1] Step: 53100 Acc: 0.68406 0.85156 Cost: 1.27005 0.86911 0.30997 0.09097 Time: 0.00077
17-03-26 01:36:50 [1] Train Extra: lr=0.0000651 inv=0.4339062 sub=0.0000000
17-03-26 01:38:18 [1] Step: 53200 Acc: 0.67750 0.84941 Cost: 1.16967 0.83586 0.24280 0.09101 Time: 0.00075
17-03-26 01:38:18 [1] Train Extra: lr=0.0000649 inv=0.4431250 sub=0.0000000
17-03-26 01:39:47 [1] Step: 53300 Acc: 0.68250 0.85409 Cost: 1.14978 0.85634 0.20234 0.09110 Time: 0.00079
17-03-26 01:39:47 [1] Train Extra: lr=0.0000647 inv=0.4307813 sub=0.0000000
17-03-26 01:41:13 [1] Step: 53400 Acc: 0.67563 0.84408 Cost: 1.27894 0.88721 0.30057 0.09116 Time: 0.00074
17-03-26 01:41:13 [1] Train Extra: lr=0.0000646 inv=0.4453125 sub=0.0000000
17-03-26 01:42:27 [1] Step: 53500 Acc: 0.67281 0.84940 Cost: 1.07179 0.70986 0.27077 0.09116 Time: 0.00069
17-03-26 01:42:27 [1] Train Extra: lr=0.0000644 inv=0.4065625 sub=0.0000000
17-03-26 01:43:23 [1] Step: 53500 Eval acc: 0.67370 0.85229 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 01:43:23 [1] Eval Extra: inv=0.4274514
17-03-26 01:44:47 [1] Step: 53600 Acc: 0.68250 0.84887 Cost: 0.88548 0.67611 0.11813 0.09124 Time: 0.00075
17-03-26 01:44:47 [1] Train Extra: lr=0.0000642 inv=0.4196875 sub=0.0000000
17-03-26 01:46:09 [1] Step: 53700 Acc: 0.68188 0.84050 Cost: 0.98621 0.64091 0.25397 0.09132 Time: 0.00073
17-03-26 01:46:09 [1] Train Extra: lr=0.0000640 inv=0.4300000 sub=0.0000000
17-03-26 01:47:22 [1] Step: 53800 Acc: 0.67625 0.85176 Cost: 0.78343 0.57542 0.11665 0.09136 Time: 0.00071
17-03-26 01:47:22 [1] Train Extra: lr=0.0000638 inv=0.4046875 sub=0.0000000
17-03-26 01:48:47 [1] Step: 53900 Acc: 0.69031 0.85401 Cost: 1.15925 0.91364 0.15421 0.09140 Time: 0.00076
17-03-26 01:48:47 [1] Train Extra: lr=0.0000636 inv=0.4100000 sub=0.0000000
17-03-26 01:50:14 [1] Step: 54000 Acc: 0.68812 0.84179 Cost: 1.22386 0.93255 0.19985 0.09146 Time: 0.00075
17-03-26 01:50:14 [1] Train Extra: lr=0.0000635 inv=0.4287500 sub=0.0000000
17-03-26 01:51:11 [1] Step: 54000 Eval acc: 0.67027 0.84825 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 01:51:11 [1] Eval Extra: inv=0.4435181
17-03-26 01:52:28 [1] Step: 54100 Acc: 0.66656 0.84992 Cost: 0.88118 0.60264 0.18696 0.09158 Time: 0.00075
17-03-26 01:52:28 [1] Train Extra: lr=0.0000633 inv=0.3768750 sub=0.0000000
17-03-26 01:53:55 [1] Step: 54200 Acc: 0.67563 0.85080 Cost: 1.20010 0.85253 0.25587 0.09170 Time: 0.00075
17-03-26 01:53:55 [1] Train Extra: lr=0.0000631 inv=0.4192187 sub=0.0000000
17-03-26 01:55:12 [1] Step: 54300 Acc: 0.68250 0.85122 Cost: 1.14331 0.85276 0.19873 0.09182 Time: 0.00075
17-03-26 01:55:12 [1] Train Extra: lr=0.0000629 inv=0.3773437 sub=0.0000000
17-03-26 01:56:32 [1] Step: 54400 Acc: 0.68406 0.84643 Cost: 0.85351 0.62590 0.13577 0.09183 Time: 0.00073
17-03-26 01:56:32 [1] Train Extra: lr=0.0000627 inv=0.3995313 sub=0.0000000
17-03-26 01:58:06 [1] Step: 54500 Acc: 0.67500 0.85282 Cost: 0.95903 0.77831 0.08885 0.09187 Time: 0.00079
17-03-26 01:58:06 [1] Train Extra: lr=0.0000625 inv=0.4389062 sub=0.0000000
17-03-26 01:59:03 [1] Step: 54500 Eval acc: 0.67613 0.85289 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 01:59:03 [1] Eval Extra: inv=0.4260711
17-03-26 02:00:27 [1] Step: 54600 Acc: 0.66781 0.84741 Cost: 0.98803 0.71926 0.17686 0.09192 Time: 0.00074
17-03-26 02:00:27 [1] Train Extra: lr=0.0000624 inv=0.4287500 sub=0.0000000
17-03-26 02:01:41 [1] Step: 54700 Acc: 0.67563 0.85233 Cost: 1.00173 0.64187 0.26788 0.09197 Time: 0.00072
17-03-26 02:01:41 [1] Train Extra: lr=0.0000622 inv=0.3796875 sub=0.0000000
17-03-26 02:02:58 [1] Step: 54800 Acc: 0.67437 0.85188 Cost: 1.06130 0.78511 0.18426 0.09192 Time: 0.00074
17-03-26 02:02:58 [1] Train Extra: lr=0.0000620 inv=0.4057812 sub=0.0000000
17-03-26 02:04:19 [1] Step: 54900 Acc: 0.67188 0.85288 Cost: 0.99471 0.78768 0.11504 0.09198 Time: 0.00074
17-03-26 02:04:19 [1] Train Extra: lr=0.0000618 inv=0.4228125 sub=0.0000000
17-03-26 02:05:44 [1] Step: 55000 Acc: 0.66906 0.84658 Cost: 0.92844 0.76228 0.07419 0.09197 Time: 0.00073
17-03-26 02:05:44 [1] Train Extra: lr=0.0000617 inv=0.4715625 sub=0.0000000
17-03-26 02:06:44 [1] Step: 55000 Eval acc: 0.67458 0.85031 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00018
17-03-26 02:06:44 [1] Eval Extra: inv=0.4050905
17-03-26 02:06:44 [1] Checkpointing.
17-03-26 02:08:03 [1] Step: 55100 Acc: 0.67156 0.84948 Cost: 1.10114 0.80376 0.20538 0.09200 Time: 0.00072
17-03-26 02:08:03 [1] Train Extra: lr=0.0000615 inv=0.4192187 sub=0.0000000
17-03-26 02:09:17 [1] Step: 55200 Acc: 0.69063 0.84815 Cost: 1.39035 0.94793 0.35035 0.09207 Time: 0.00070
17-03-26 02:09:17 [1] Train Extra: lr=0.0000613 inv=0.4106250 sub=0.0000000
17-03-26 02:10:36 [1] Step: 55300 Acc: 0.65906 0.84947 Cost: 1.21908 0.96873 0.15823 0.09212 Time: 0.00074
17-03-26 02:10:36 [1] Train Extra: lr=0.0000611 inv=0.4042188 sub=0.0000000
17-03-26 02:11:48 [1] Step: 55400 Acc: 0.67281 0.85258 Cost: 1.00989 0.77671 0.14102 0.09215 Time: 0.00072
17-03-26 02:11:48 [1] Train Extra: lr=0.0000609 inv=0.4045313 sub=0.0000000
17-03-26 02:13:18 [1] Step: 55500 Acc: 0.66281 0.84295 Cost: 1.12933 0.81010 0.22703 0.09220 Time: 0.00077
17-03-26 02:13:18 [1] Train Extra: lr=0.0000608 inv=0.4428125 sub=0.0000000
17-03-26 02:14:16 [1] Step: 55500 Eval acc: 0.66928 0.85473 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 02:14:16 [1] Eval Extra: inv=0.3918397
17-03-26 02:15:41 [1] Step: 55600 Acc: 0.66531 0.85328 Cost: 0.84598 0.67003 0.08375 0.09220 Time: 0.00077
17-03-26 02:15:41 [1] Train Extra: lr=0.0000606 inv=0.4376563 sub=0.0000000
17-03-26 02:17:09 [1] Step: 55700 Acc: 0.66563 0.84931 Cost: 1.00064 0.70045 0.20791 0.09228 Time: 0.00075
17-03-26 02:17:09 [1] Train Extra: lr=0.0000604 inv=0.4387500 sub=0.0000000
17-03-26 02:18:34 [1] Step: 55800 Acc: 0.67188 0.84540 Cost: 1.06043 0.77380 0.19429 0.09234 Time: 0.00075
17-03-26 02:18:34 [1] Train Extra: lr=0.0000603 inv=0.4237500 sub=0.0000000
17-03-26 02:19:46 [1] Step: 55900 Acc: 0.67969 0.84753 Cost: 0.83981 0.59869 0.14873 0.09238 Time: 0.00071
17-03-26 02:19:46 [1] Train Extra: lr=0.0000601 inv=0.3998437 sub=0.0000000
17-03-26 02:20:58 [1] Step: 56000 Acc: 0.65844 0.84873 Cost: 1.13888 0.74824 0.29816 0.09248 Time: 0.00070
17-03-26 02:20:58 [1] Train Extra: lr=0.0000599 inv=0.4173438 sub=0.0000000
17-03-26 02:21:56 [1] Step: 56000 Eval acc: 0.67259 0.85413 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 02:21:56 [1] Eval Extra: inv=0.4552231
17-03-26 02:23:07 [1] Step: 56100 Acc: 0.68531 0.84946 Cost: 1.09706 0.76223 0.24229 0.09254 Time: 0.00072
17-03-26 02:23:07 [1] Train Extra: lr=0.0000597 inv=0.4153125 sub=0.0000000
17-03-26 02:24:34 [1] Step: 56200 Acc: 0.67688 0.85428 Cost: 0.83475 0.51885 0.22340 0.09250 Time: 0.00077
17-03-26 02:24:34 [1] Train Extra: lr=0.0000596 inv=0.4406250 sub=0.0000000
17-03-26 02:25:52 [1] Step: 56300 Acc: 0.67094 0.84538 Cost: 1.29064 0.92630 0.27184 0.09250 Time: 0.00073
17-03-26 02:25:52 [1] Train Extra: lr=0.0000594 inv=0.4209375 sub=0.0000000
17-03-26 02:27:10 [1] Step: 56400 Acc: 0.69875 0.85127 Cost: 1.05175 0.73031 0.22883 0.09261 Time: 0.00074
17-03-26 02:27:10 [1] Train Extra: lr=0.0000592 inv=0.4098438 sub=0.0000000
17-03-26 02:28:33 [1] Step: 56500 Acc: 0.67437 0.84306 Cost: 1.07066 0.73167 0.24633 0.09266 Time: 0.00073
17-03-26 02:28:33 [1] Train Extra: lr=0.0000590 inv=0.4320312 sub=0.0000000
17-03-26 02:29:30 [1] Step: 56500 Eval acc: 0.67149 0.85203 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 02:29:30 [1] Eval Extra: inv=0.4132619
17-03-26 02:30:47 [1] Step: 56600 Acc: 0.66250 0.83981 Cost: 1.02367 0.64395 0.28702 0.09269 Time: 0.00071
17-03-26 02:30:47 [1] Train Extra: lr=0.0000589 inv=0.4271875 sub=0.0000000
17-03-26 02:32:10 [1] Step: 56700 Acc: 0.66687 0.85285 Cost: 1.27781 0.86700 0.31805 0.09276 Time: 0.00076
17-03-26 02:32:10 [1] Train Extra: lr=0.0000587 inv=0.4062500 sub=0.0000000
17-03-26 02:33:30 [1] Step: 56800 Acc: 0.68344 0.86047 Cost: 1.07183 0.75672 0.22232 0.09279 Time: 0.00076
17-03-26 02:33:30 [1] Train Extra: lr=0.0000585 inv=0.3934375 sub=0.0000000
17-03-26 02:34:53 [1] Step: 56900 Acc: 0.66219 0.84315 Cost: 1.16570 0.85908 0.21384 0.09278 Time: 0.00073
17-03-26 02:34:53 [1] Train Extra: lr=0.0000584 inv=0.4651562 sub=0.0000000
17-03-26 02:36:13 [1] Step: 57000 Acc: 0.67750 0.85007 Cost: 0.86983 0.53852 0.23846 0.09285 Time: 0.00073
17-03-26 02:36:13 [1] Train Extra: lr=0.0000582 inv=0.4103125 sub=0.0000000
17-03-26 02:37:09 [1] Step: 57000 Eval acc: 0.68231 0.85203 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 02:37:09 [1] Eval Extra: inv=0.3974161
17-03-26 02:37:09 [1] Checkpointing with new best dev accuracy of 0.682310
17-03-26 02:38:33 [1] Step: 57100 Acc: 0.67156 0.84584 Cost: 0.95574 0.70162 0.16121 0.09291 Time: 0.00074
17-03-26 02:38:33 [1] Train Extra: lr=0.0000580 inv=0.4182812 sub=0.0000000
17-03-26 02:39:51 [1] Step: 57200 Acc: 0.67625 0.85387 Cost: 1.27218 0.92666 0.25263 0.09288 Time: 0.00076
17-03-26 02:39:51 [1] Train Extra: lr=0.0000579 inv=0.3921875 sub=0.0000000
17-03-26 02:41:09 [1] Step: 57300 Acc: 0.68469 0.84355 Cost: 0.96720 0.74768 0.12661 0.09291 Time: 0.00072
17-03-26 02:41:09 [1] Train Extra: lr=0.0000577 inv=0.4293750 sub=0.0000000
17-03-26 02:42:34 [1] Step: 57400 Acc: 0.67531 0.85521 Cost: 1.40403 1.02591 0.28516 0.09295 Time: 0.00077
17-03-26 02:42:34 [1] Train Extra: lr=0.0000575 inv=0.4275000 sub=0.0000000
17-03-26 02:43:49 [1] Step: 57500 Acc: 0.68437 0.84638 Cost: 0.68776 0.47456 0.12021 0.09298 Time: 0.00070
17-03-26 02:43:49 [1] Train Extra: lr=0.0000574 inv=0.3987500 sub=0.0000000
17-03-26 02:44:46 [1] Step: 57500 Eval acc: 0.67668 0.84830 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 02:44:46 [1] Eval Extra: inv=0.4351811
17-03-26 02:46:11 [1] Step: 57600 Acc: 0.66781 0.84087 Cost: 1.20139 0.90932 0.19911 0.09296 Time: 0.00075
17-03-26 02:46:11 [1] Train Extra: lr=0.0000572 inv=0.4251563 sub=0.0000000
17-03-26 02:47:36 [1] Step: 57700 Acc: 0.66500 0.84868 Cost: 1.14236 0.78480 0.26460 0.09296 Time: 0.00075
17-03-26 02:47:36 [1] Train Extra: lr=0.0000570 inv=0.4595313 sub=0.0000000
17-03-26 02:49:01 [1] Step: 57800 Acc: 0.67937 0.84582 Cost: 0.64413 0.41948 0.13163 0.09302 Time: 0.00076
17-03-26 02:49:01 [1] Train Extra: lr=0.0000569 inv=0.4206250 sub=0.0000000
17-03-26 02:50:19 [1] Step: 57900 Acc: 0.68000 0.85181 Cost: 1.02106 0.67918 0.24886 0.09302 Time: 0.00073
17-03-26 02:50:19 [1] Train Extra: lr=0.0000567 inv=0.4193750 sub=0.0000000
17-03-26 02:51:38 [1] Step: 58000 Acc: 0.66812 0.84954 Cost: 0.93308 0.70327 0.13675 0.09307 Time: 0.00072
17-03-26 02:51:38 [1] Train Extra: lr=0.0000566 inv=0.4362500 sub=0.0000000
17-03-26 02:52:35 [1] Step: 58000 Eval acc: 0.67557 0.85076 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 02:52:35 [1] Eval Extra: inv=0.4278931
17-03-26 02:53:51 [1] Step: 58100 Acc: 0.68031 0.85081 Cost: 0.81428 0.53323 0.18797 0.09308 Time: 0.00073
17-03-26 02:53:51 [1] Train Extra: lr=0.0000564 inv=0.3948437 sub=0.0000000
17-03-26 02:55:15 [1] Step: 58200 Acc: 0.67469 0.84826 Cost: 1.04935 0.73419 0.22197 0.09319 Time: 0.00074
17-03-26 02:55:15 [1] Train Extra: lr=0.0000562 inv=0.4321875 sub=0.0000000
17-03-26 02:56:29 [1] Step: 58300 Acc: 0.68156 0.84917 Cost: 1.20697 0.87216 0.24164 0.09317 Time: 0.00072
17-03-26 02:56:29 [1] Train Extra: lr=0.0000561 inv=0.4007812 sub=0.0000000
17-03-26 02:57:54 [1] Step: 58400 Acc: 0.67781 0.85289 Cost: 1.01071 0.76919 0.14831 0.09321 Time: 0.00078
17-03-26 02:57:54 [1] Train Extra: lr=0.0000559 inv=0.4079687 sub=0.0000000
17-03-26 02:59:14 [1] Step: 58500 Acc: 0.67719 0.84710 Cost: 0.99357 0.71926 0.18108 0.09324 Time: 0.00071
17-03-26 02:59:14 [1] Train Extra: lr=0.0000557 inv=0.4421875 sub=0.0000000
17-03-26 03:00:14 [1] Step: 58500 Eval acc: 0.68242 0.85178 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00018
17-03-26 03:00:14 [1] Eval Extra: inv=0.4543949
17-03-26 03:01:47 [1] Step: 58600 Acc: 0.66187 0.84846 Cost: 1.01817 0.62202 0.30293 0.09322 Time: 0.00076
17-03-26 03:01:47 [1] Train Extra: lr=0.0000556 inv=0.4595313 sub=0.0000000
17-03-26 03:03:06 [1] Step: 58700 Acc: 0.67781 0.84085 Cost: 1.07322 0.71144 0.26850 0.09328 Time: 0.00071
17-03-26 03:03:06 [1] Train Extra: lr=0.0000554 inv=0.4337500 sub=0.0000000
17-03-26 03:04:33 [1] Step: 58800 Acc: 0.68500 0.84414 Cost: 1.42155 1.04953 0.27863 0.09339 Time: 0.00075
17-03-26 03:04:33 [1] Train Extra: lr=0.0000553 inv=0.4537500 sub=0.0000000
17-03-26 03:05:51 [1] Step: 58900 Acc: 0.70250 0.85441 Cost: 0.93527 0.58983 0.25193 0.09351 Time: 0.00075
17-03-26 03:05:51 [1] Train Extra: lr=0.0000551 inv=0.4075000 sub=0.0000000
17-03-26 03:07:16 [1] Step: 59000 Acc: 0.70656 0.83639 Cost: 1.02009 0.72073 0.20570 0.09366 Time: 0.00073
17-03-26 03:07:16 [1] Train Extra: lr=0.0000550 inv=0.4593750 sub=0.0000000
17-03-26 03:08:14 [1] Step: 59000 Eval acc: 0.67756 0.85481 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 03:08:14 [1] Eval Extra: inv=0.4026060
17-03-26 03:09:39 [1] Step: 59100 Acc: 0.67906 0.85317 Cost: 1.21642 0.88636 0.23622 0.09384 Time: 0.00076
17-03-26 03:09:39 [1] Train Extra: lr=0.0000548 inv=0.4196875 sub=0.0000000
17-03-26 03:11:03 [1] Step: 59200 Acc: 0.70688 0.85104 Cost: 1.06598 0.84262 0.12940 0.09396 Time: 0.00076
17-03-26 03:11:03 [1] Train Extra: lr=0.0000546 inv=0.4267187 sub=0.0000000
17-03-26 03:12:23 [1] Step: 59300 Acc: 0.69688 0.85623 Cost: 0.92578 0.66944 0.16226 0.09408 Time: 0.00076
17-03-26 03:12:23 [1] Train Extra: lr=0.0000545 inv=0.4107812 sub=0.0000000
17-03-26 03:13:52 [1] Step: 59400 Acc: 0.70750 0.84462 Cost: 1.18765 0.82548 0.26797 0.09420 Time: 0.00076
17-03-26 03:13:52 [1] Train Extra: lr=0.0000543 inv=0.4528125 sub=0.0000000
17-03-26 03:15:16 [1] Step: 59500 Acc: 0.69219 0.85398 Cost: 1.22087 0.88786 0.23873 0.09428 Time: 0.00077
17-03-26 03:15:16 [1] Train Extra: lr=0.0000542 inv=0.4203125 sub=0.0000000
17-03-26 03:16:11 [1] Step: 59500 Eval acc: 0.67856 0.85275 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-26 03:16:11 [1] Eval Extra: inv=0.4469413
17-03-26 03:17:24 [1] Step: 59600 Acc: 0.69719 0.84988 Cost: 1.14343 0.82667 0.22230 0.09446 Time: 0.00070
17-03-26 03:17:24 [1] Train Extra: lr=0.0000540 inv=0.4106250 sub=0.0000000
17-03-26 03:18:49 [1] Step: 59700 Acc: 0.69688 0.85261 Cost: 1.23717 0.86326 0.27940 0.09451 Time: 0.00076
17-03-26 03:18:49 [1] Train Extra: lr=0.0000539 inv=0.4239062 sub=0.0000000
17-03-26 03:20:13 [1] Step: 59800 Acc: 0.69563 0.84021 Cost: 0.95351 0.64413 0.21480 0.09458 Time: 0.00075
17-03-26 03:20:13 [1] Train Extra: lr=0.0000537 inv=0.4296875 sub=0.0000000
17-03-26 03:21:34 [1] Step: 59900 Acc: 0.70531 0.85098 Cost: 0.83089 0.60581 0.13036 0.09472 Time: 0.00075
17-03-26 03:21:34 [1] Train Extra: lr=0.0000535 inv=0.3959375 sub=0.0000000
17-03-26 03:23:05 [1] Step: 60000 Acc: 0.68937 0.85321 Cost: 1.09860 0.86259 0.14117 0.09484 Time: 0.00079
17-03-26 03:23:05 [1] Train Extra: lr=0.0000534 inv=0.4167188 sub=0.0000000
17-03-26 03:24:02 [1] Step: 60000 Eval acc: 0.67690 0.85032 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 03:24:02 [1] Eval Extra: inv=0.4142557
17-03-26 03:24:02 [1] Checkpointing.
17-03-26 03:25:22 [1] Step: 60100 Acc: 0.69344 0.84789 Cost: 0.93445 0.70547 0.13398 0.09500 Time: 0.00072
17-03-26 03:25:22 [1] Train Extra: lr=0.0000532 inv=0.4412500 sub=0.0000000
17-03-26 03:26:41 [1] Step: 60200 Acc: 0.69656 0.85010 Cost: 1.13592 0.90314 0.13768 0.09510 Time: 0.00074
17-03-26 03:26:41 [1] Train Extra: lr=0.0000531 inv=0.4040625 sub=0.0000000
17-03-26 03:27:59 [1] Step: 60300 Acc: 0.70906 0.84624 Cost: 1.01380 0.78813 0.13053 0.09515 Time: 0.00072
17-03-26 03:27:59 [1] Train Extra: lr=0.0000529 inv=0.4339062 sub=0.0000000
17-03-26 03:29:24 [1] Step: 60400 Acc: 0.68031 0.85622 Cost: 0.89272 0.60834 0.18915 0.09523 Time: 0.00078
17-03-26 03:29:24 [1] Train Extra: lr=0.0000528 inv=0.4260937 sub=0.0000000
17-03-26 03:30:48 [1] Step: 60500 Acc: 0.69281 0.84462 Cost: 0.72677 0.52424 0.10711 0.09541 Time: 0.00074
17-03-26 03:30:48 [1] Train Extra: lr=0.0000526 inv=0.4420312 sub=0.0000000
17-03-26 03:31:45 [1] Step: 60500 Eval acc: 0.67712 0.84621 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 03:31:45 [1] Eval Extra: inv=0.4531250
17-03-26 03:32:59 [1] Step: 60600 Acc: 0.69719 0.84908 Cost: 0.91875 0.65489 0.16843 0.09544 Time: 0.00072
17-03-26 03:32:59 [1] Train Extra: lr=0.0000525 inv=0.4142188 sub=0.0000000
17-03-26 03:34:32 [1] Step: 60700 Acc: 0.66687 0.85149 Cost: 1.19659 0.85983 0.24120 0.09556 Time: 0.00079
17-03-26 03:34:32 [1] Train Extra: lr=0.0000523 inv=0.4556250 sub=0.0000000
17-03-26 03:35:56 [1] Step: 60800 Acc: 0.69250 0.84412 Cost: 1.08044 0.67376 0.31103 0.09565 Time: 0.00074
17-03-26 03:35:56 [1] Train Extra: lr=0.0000522 inv=0.4520312 sub=0.0000000
17-03-26 03:37:16 [1] Step: 60900 Acc: 0.68937 0.84847 Cost: 1.35683 1.05703 0.20411 0.09569 Time: 0.00072
17-03-26 03:37:16 [1] Train Extra: lr=0.0000520 inv=0.4364062 sub=0.0000000
17-03-26 03:38:42 [1] Step: 61000 Acc: 0.69219 0.84298 Cost: 0.82688 0.54954 0.18164 0.09571 Time: 0.00075
17-03-26 03:38:42 [1] Train Extra: lr=0.0000519 inv=0.4503125 sub=0.0000000
17-03-26 03:39:40 [1] Step: 61000 Eval acc: 0.67337 0.85286 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 03:39:40 [1] Eval Extra: inv=0.4044280
17-03-26 03:41:04 [1] Step: 61100 Acc: 0.68812 0.84764 Cost: 1.18894 0.82227 0.27093 0.09573 Time: 0.00078
17-03-26 03:41:04 [1] Train Extra: lr=0.0000517 inv=0.4178125 sub=0.0000000
17-03-26 03:42:32 [1] Step: 61200 Acc: 0.68094 0.85443 Cost: 0.88011 0.59305 0.19123 0.09583 Time: 0.00076
17-03-26 03:42:32 [1] Train Extra: lr=0.0000516 inv=0.4489063 sub=0.0000000
17-03-26 03:43:47 [1] Step: 61300 Acc: 0.68656 0.84695 Cost: 0.75654 0.48818 0.17245 0.09590 Time: 0.00071
17-03-26 03:43:47 [1] Train Extra: lr=0.0000514 inv=0.3950000 sub=0.0000000
17-03-26 03:45:10 [1] Step: 61400 Acc: 0.69094 0.85012 Cost: 1.45965 1.07637 0.28733 0.09594 Time: 0.00076
17-03-26 03:45:10 [1] Train Extra: lr=0.0000513 inv=0.4290625 sub=0.0000000
17-03-26 03:46:36 [1] Step: 61500 Acc: 0.70312 0.84592 Cost: 1.24322 0.82182 0.32540 0.09600 Time: 0.00074
17-03-26 03:46:36 [1] Train Extra: lr=0.0000511 inv=0.4364062 sub=0.0000000
17-03-26 03:47:33 [1] Step: 61500 Eval acc: 0.67734 0.85594 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 03:47:33 [1] Eval Extra: inv=0.4369479
17-03-26 03:48:46 [1] Step: 61600 Acc: 0.67875 0.84487 Cost: 1.18054 0.84922 0.23523 0.09608 Time: 0.00071
17-03-26 03:48:46 [1] Train Extra: lr=0.0000510 inv=0.4276563 sub=0.0000000
17-03-26 03:50:03 [1] Step: 61700 Acc: 0.68437 0.84537 Cost: 1.18870 0.79669 0.29593 0.09607 Time: 0.00072
17-03-26 03:50:03 [1] Train Extra: lr=0.0000508 inv=0.4182812 sub=0.0000000
17-03-26 03:51:34 [1] Step: 61800 Acc: 0.68125 0.84613 Cost: 0.97513 0.71616 0.16284 0.09613 Time: 0.00076
17-03-26 03:51:34 [1] Train Extra: lr=0.0000507 inv=0.4342187 sub=0.0000000
17-03-26 03:52:41 [1] Step: 61900 Acc: 0.67906 0.85328 Cost: 0.95767 0.68657 0.17483 0.09627 Time: 0.00069
17-03-26 03:52:41 [1] Train Extra: lr=0.0000506 inv=0.3590625 sub=0.0000000
17-03-26 03:54:03 [1] Step: 62000 Acc: 0.68312 0.85599 Cost: 1.05515 0.71397 0.24487 0.09631 Time: 0.00076
17-03-26 03:54:03 [1] Train Extra: lr=0.0000504 inv=0.4339062 sub=0.0000000
17-03-26 03:54:59 [1] Step: 62000 Eval acc: 0.67326 0.85106 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 03:54:59 [1] Eval Extra: inv=0.4327518
17-03-26 03:56:14 [1] Step: 62100 Acc: 0.68750 0.84452 Cost: 0.86969 0.55087 0.22251 0.09631 Time: 0.00071
17-03-26 03:56:14 [1] Train Extra: lr=0.0000503 inv=0.4325000 sub=0.0000000
17-03-26 03:57:31 [1] Step: 62200 Acc: 0.67969 0.84832 Cost: 1.25090 0.90801 0.24653 0.09635 Time: 0.00074
17-03-26 03:57:31 [1] Train Extra: lr=0.0000501 inv=0.3978125 sub=0.0000000
17-03-26 03:58:51 [1] Step: 62300 Acc: 0.70281 0.85271 Cost: 0.81611 0.63086 0.08882 0.09644 Time: 0.00074
17-03-26 03:58:51 [1] Train Extra: lr=0.0000500 inv=0.4062500 sub=0.0000000
17-03-26 04:00:10 [1] Step: 62400 Acc: 0.68563 0.85063 Cost: 1.27897 0.94702 0.23544 0.09650 Time: 0.00074
17-03-26 04:00:10 [1] Train Extra: lr=0.0000498 inv=0.4193750 sub=0.0000000
17-03-26 04:01:27 [1] Step: 62500 Acc: 0.67312 0.84921 Cost: 0.87433 0.65858 0.11920 0.09656 Time: 0.00074
17-03-26 04:01:27 [1] Train Extra: lr=0.0000497 inv=0.4121875 sub=0.0000000
17-03-26 04:02:24 [1] Step: 62500 Eval acc: 0.68143 0.84520 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 04:02:24 [1] Eval Extra: inv=0.4335799
17-03-26 04:03:49 [1] Step: 62600 Acc: 0.67969 0.84806 Cost: 1.09092 0.86108 0.13315 0.09669 Time: 0.00077
17-03-26 04:03:49 [1] Train Extra: lr=0.0000495 inv=0.4121875 sub=0.0000000
17-03-26 04:05:16 [1] Step: 62700 Acc: 0.68031 0.84802 Cost: 0.96640 0.66985 0.19984 0.09671 Time: 0.00075
17-03-26 04:05:16 [1] Train Extra: lr=0.0000494 inv=0.4403125 sub=0.0000000
17-03-26 04:06:42 [1] Step: 62800 Acc: 0.68969 0.84175 Cost: 0.95713 0.55618 0.30422 0.09672 Time: 0.00075
17-03-26 04:06:42 [1] Train Extra: lr=0.0000493 inv=0.4382813 sub=0.0000000
17-03-26 04:08:09 [1] Step: 62900 Acc: 0.68344 0.85359 Cost: 0.97885 0.71313 0.16890 0.09682 Time: 0.00077
17-03-26 04:08:09 [1] Train Extra: lr=0.0000491 inv=0.4351563 sub=0.0000000
17-03-26 04:09:35 [1] Step: 63000 Acc: 0.68594 0.84748 Cost: 0.99881 0.65325 0.24864 0.09692 Time: 0.00075
17-03-26 04:09:35 [1] Train Extra: lr=0.0000490 inv=0.4310937 sub=0.0000000
17-03-26 04:10:32 [1] Step: 63000 Eval acc: 0.68043 0.85240 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 04:10:32 [1] Eval Extra: inv=0.4100596
17-03-26 04:11:51 [1] Step: 63100 Acc: 0.68531 0.84500 Cost: 1.06268 0.71955 0.24622 0.09692 Time: 0.00073
17-03-26 04:11:51 [1] Train Extra: lr=0.0000488 inv=0.4162500 sub=0.0000000
17-03-26 04:13:11 [1] Step: 63200 Acc: 0.67812 0.84976 Cost: 1.04486 0.75652 0.19135 0.09700 Time: 0.00074
17-03-26 04:13:11 [1] Train Extra: lr=0.0000487 inv=0.4162500 sub=0.0000000
17-03-26 04:14:23 [1] Step: 63300 Acc: 0.69656 0.84648 Cost: 0.97656 0.57865 0.30090 0.09701 Time: 0.00072
17-03-26 04:14:23 [1] Train Extra: lr=0.0000486 inv=0.4140625 sub=0.0000000
17-03-26 04:15:42 [1] Step: 63400 Acc: 0.68750 0.84773 Cost: 1.03998 0.74994 0.19292 0.09712 Time: 0.00073
17-03-26 04:15:42 [1] Train Extra: lr=0.0000484 inv=0.4226563 sub=0.0000000
17-03-26 04:16:59 [1] Step: 63500 Acc: 0.69594 0.85376 Cost: 0.98297 0.76794 0.11786 0.09717 Time: 0.00076
17-03-26 04:16:59 [1] Train Extra: lr=0.0000483 inv=0.4045313 sub=0.0000000
17-03-26 04:17:57 [1] Step: 63500 Eval acc: 0.68264 0.85433 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 04:17:57 [1] Eval Extra: inv=0.4058635
17-03-26 04:19:16 [1] Step: 63600 Acc: 0.67500 0.84334 Cost: 1.11870 0.77420 0.24724 0.09726 Time: 0.00071
17-03-26 04:19:16 [1] Train Extra: lr=0.0000481 inv=0.4515625 sub=0.0000000
17-03-26 04:20:29 [1] Step: 63700 Acc: 0.69125 0.85384 Cost: 0.91998 0.71636 0.10639 0.09723 Time: 0.00072
17-03-26 04:20:29 [1] Train Extra: lr=0.0000480 inv=0.4004687 sub=0.0000000
17-03-26 04:21:47 [1] Step: 63800 Acc: 0.68250 0.84579 Cost: 1.06185 0.68298 0.28162 0.09726 Time: 0.00074
17-03-26 04:21:47 [1] Train Extra: lr=0.0000479 inv=0.4095313 sub=0.0000000
17-03-26 04:23:05 [1] Step: 63900 Acc: 0.68063 0.85149 Cost: 1.17434 0.80517 0.27190 0.09726 Time: 0.00073
17-03-26 04:23:05 [1] Train Extra: lr=0.0000477 inv=0.3942188 sub=0.0000000
17-03-26 04:24:25 [1] Step: 64000 Acc: 0.68875 0.84202 Cost: 1.12499 0.80275 0.22496 0.09728 Time: 0.00071
17-03-26 04:24:25 [1] Train Extra: lr=0.0000476 inv=0.4218750 sub=0.0000000
17-03-26 04:25:22 [1] Step: 64000 Eval acc: 0.68087 0.85239 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 04:25:22 [1] Eval Extra: inv=0.4049249
17-03-26 04:26:47 [1] Step: 64100 Acc: 0.67406 0.83962 Cost: 1.00618 0.72499 0.18387 0.09732 Time: 0.00073
17-03-26 04:26:47 [1] Train Extra: lr=0.0000475 inv=0.4595313 sub=0.0000000
17-03-26 04:28:28 [1] Step: 64200 Acc: 0.67906 0.85242 Cost: 1.00531 0.63418 0.27369 0.09744 Time: 0.00082
17-03-26 04:28:28 [1] Train Extra: lr=0.0000473 inv=0.4526562 sub=0.0000000
17-03-26 04:29:42 [1] Step: 64300 Acc: 0.68594 0.85105 Cost: 1.19712 0.89059 0.20906 0.09747 Time: 0.00072
17-03-26 04:29:42 [1] Train Extra: lr=0.0000472 inv=0.4070313 sub=0.0000000
17-03-26 04:31:05 [1] Step: 64400 Acc: 0.69219 0.85039 Cost: 1.39636 1.01518 0.28368 0.09750 Time: 0.00077
17-03-26 04:31:05 [1] Train Extra: lr=0.0000470 inv=0.4096875 sub=0.0000000
17-03-26 04:32:20 [1] Step: 64500 Acc: 0.69688 0.85012 Cost: 1.00986 0.83013 0.08215 0.09758 Time: 0.00072
17-03-26 04:32:20 [1] Train Extra: lr=0.0000469 inv=0.3940625 sub=0.0000000
17-03-26 04:33:17 [1] Step: 64500 Eval acc: 0.68165 0.85002 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 04:33:17 [1] Eval Extra: inv=0.4223719
17-03-26 04:34:38 [1] Step: 64600 Acc: 0.67031 0.84458 Cost: 1.05763 0.69682 0.26317 0.09764 Time: 0.00075
17-03-26 04:34:38 [1] Train Extra: lr=0.0000468 inv=0.4292187 sub=0.0000000
17-03-26 04:36:03 [1] Step: 64700 Acc: 0.67656 0.84243 Cost: 0.94802 0.66359 0.18674 0.09770 Time: 0.00074
17-03-26 04:36:03 [1] Train Extra: lr=0.0000466 inv=0.4318750 sub=0.0000000
17-03-26 04:37:31 [1] Step: 64800 Acc: 0.66531 0.84084 Cost: 0.98410 0.71398 0.17242 0.09770 Time: 0.00074
17-03-26 04:37:31 [1] Train Extra: lr=0.0000465 inv=0.4506250 sub=0.0000000
17-03-26 04:39:00 [1] Step: 64900 Acc: 0.68344 0.85154 Cost: 1.09203 0.80025 0.19405 0.09772 Time: 0.00079
17-03-26 04:39:00 [1] Train Extra: lr=0.0000464 inv=0.4303125 sub=0.0000000
17-03-26 04:40:33 [1] Step: 65000 Acc: 0.66875 0.85609 Cost: 1.05891 0.80943 0.15159 0.09789 Time: 0.00080
17-03-26 04:40:33 [1] Train Extra: lr=0.0000462 inv=0.4168750 sub=0.0000000
17-03-26 04:41:30 [1] Step: 65000 Eval acc: 0.68518 0.85347 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 04:41:30 [1] Eval Extra: inv=0.4139245
17-03-26 04:41:30 [1] Checkpointing.
17-03-26 04:42:55 [1] Step: 65100 Acc: 0.67500 0.84567 Cost: 0.94274 0.68124 0.16361 0.09790 Time: 0.00075
17-03-26 04:42:55 [1] Train Extra: lr=0.0000461 inv=0.4425000 sub=0.0000000
17-03-26 04:44:13 [1] Step: 65200 Acc: 0.68563 0.84764 Cost: 0.97113 0.65440 0.21883 0.09790 Time: 0.00073
17-03-26 04:44:13 [1] Train Extra: lr=0.0000460 inv=0.4279688 sub=0.0000000
17-03-26 04:45:43 [1] Step: 65300 Acc: 0.67969 0.85130 Cost: 1.05259 0.72847 0.22620 0.09793 Time: 0.00078
17-03-26 04:45:43 [1] Train Extra: lr=0.0000458 inv=0.4143750 sub=0.0000000
17-03-26 04:47:15 [1] Step: 65400 Acc: 0.65938 0.84015 Cost: 1.16931 0.92922 0.14216 0.09793 Time: 0.00075
17-03-26 04:47:15 [1] Train Extra: lr=0.0000457 inv=0.4512500 sub=0.0000000
17-03-26 04:48:36 [1] Step: 65500 Acc: 0.67188 0.84746 Cost: 1.20490 0.84601 0.26094 0.09794 Time: 0.00072
17-03-26 04:48:36 [1] Train Extra: lr=0.0000456 inv=0.4232812 sub=0.0000000
17-03-26 04:49:33 [1] Step: 65500 Eval acc: 0.67955 0.85600 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 04:49:33 [1] Eval Extra: inv=0.3786992
17-03-26 04:50:51 [1] Step: 65600 Acc: 0.67406 0.85400 Cost: 1.31820 0.96579 0.25447 0.09793 Time: 0.00074
17-03-26 04:50:51 [1] Train Extra: lr=0.0000454 inv=0.4035937 sub=0.0000000
17-03-26 04:52:10 [1] Step: 65700 Acc: 0.67875 0.85476 Cost: 0.85184 0.66539 0.08846 0.09799 Time: 0.00076
17-03-26 04:52:10 [1] Train Extra: lr=0.0000453 inv=0.3859375 sub=0.0000000
17-03-26 04:53:35 [1] Step: 65800 Acc: 0.70250 0.85485 Cost: 0.95355 0.70626 0.14923 0.09805 Time: 0.00076
17-03-26 04:53:35 [1] Train Extra: lr=0.0000452 inv=0.4270312 sub=0.0000000
17-03-26 04:54:59 [1] Step: 65900 Acc: 0.67125 0.84315 Cost: 1.34555 0.98401 0.26342 0.09812 Time: 0.00074
17-03-26 04:54:59 [1] Train Extra: lr=0.0000451 inv=0.4125000 sub=0.0000000
17-03-26 04:56:24 [1] Step: 66000 Acc: 0.67563 0.85611 Cost: 0.71365 0.51625 0.09924 0.09816 Time: 0.00076
17-03-26 04:56:24 [1] Train Extra: lr=0.0000449 inv=0.4125000 sub=0.0000000
17-03-26 04:57:20 [1] Step: 66000 Eval acc: 0.67646 0.85215 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 04:57:20 [1] Eval Extra: inv=0.4134828
17-03-26 04:58:40 [1] Step: 66100 Acc: 0.69250 0.84846 Cost: 1.21804 0.95732 0.16258 0.09815 Time: 0.00074
17-03-26 04:58:40 [1] Train Extra: lr=0.0000448 inv=0.4001562 sub=0.0000000
17-03-26 05:00:05 [1] Step: 66200 Acc: 0.69719 0.85199 Cost: 0.95178 0.65850 0.19514 0.09814 Time: 0.00075
17-03-26 05:00:05 [1] Train Extra: lr=0.0000447 inv=0.4273438 sub=0.0000000
17-03-26 05:01:29 [1] Step: 66300 Acc: 0.67875 0.84679 Cost: 0.86939 0.64896 0.12219 0.09824 Time: 0.00074
17-03-26 05:01:29 [1] Train Extra: lr=0.0000445 inv=0.4415625 sub=0.0000000
17-03-26 05:02:47 [1] Step: 66400 Acc: 0.67750 0.84888 Cost: 1.16561 0.82614 0.24124 0.09824 Time: 0.00073
17-03-26 05:02:47 [1] Train Extra: lr=0.0000444 inv=0.4021875 sub=0.0000000
17-03-26 05:04:07 [1] Step: 66500 Acc: 0.69000 0.85314 Cost: 1.34012 1.04829 0.19361 0.09822 Time: 0.00074
17-03-26 05:04:07 [1] Train Extra: lr=0.0000443 inv=0.4245313 sub=0.0000000
17-03-26 05:05:03 [1] Step: 66500 Eval acc: 0.67635 0.84986 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 05:05:03 [1] Eval Extra: inv=0.4051458
17-03-26 05:06:15 [1] Step: 66600 Acc: 0.67437 0.85073 Cost: 1.28739 0.96503 0.22412 0.09824 Time: 0.00072
17-03-26 05:06:15 [1] Train Extra: lr=0.0000442 inv=0.4006250 sub=0.0000000
17-03-26 05:07:33 [1] Step: 66700 Acc: 0.67625 0.85091 Cost: 0.82808 0.54083 0.18901 0.09825 Time: 0.00073
17-03-26 05:07:33 [1] Train Extra: lr=0.0000440 inv=0.4100000 sub=0.0000000
17-03-26 05:08:54 [1] Step: 66800 Acc: 0.67937 0.84899 Cost: 0.90732 0.66839 0.14066 0.09828 Time: 0.00075
17-03-26 05:08:54 [1] Train Extra: lr=0.0000439 inv=0.4239062 sub=0.0000000
17-03-26 05:10:20 [1] Step: 66900 Acc: 0.67594 0.84405 Cost: 1.37569 1.03644 0.24095 0.09829 Time: 0.00073
17-03-26 05:10:20 [1] Train Extra: lr=0.0000438 inv=0.4381250 sub=0.0000000
17-03-26 05:11:40 [1] Step: 67000 Acc: 0.67812 0.84780 Cost: 1.02953 0.69865 0.23250 0.09838 Time: 0.00073
17-03-26 05:11:40 [1] Train Extra: lr=0.0000437 inv=0.4210937 sub=0.0000000
17-03-26 05:12:37 [1] Step: 67000 Eval acc: 0.68065 0.85192 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 05:12:37 [1] Eval Extra: inv=0.4140349
17-03-26 05:13:55 [1] Step: 67100 Acc: 0.67312 0.85156 Cost: 1.22445 0.90666 0.21941 0.09838 Time: 0.00074
17-03-26 05:13:55 [1] Train Extra: lr=0.0000435 inv=0.4101562 sub=0.0000000
17-03-26 05:15:27 [1] Step: 67200 Acc: 0.69437 0.85301 Cost: 1.13704 0.81034 0.22826 0.09844 Time: 0.00080
17-03-26 05:15:27 [1] Train Extra: lr=0.0000434 inv=0.4171875 sub=0.0000000
17-03-26 05:16:44 [1] Step: 67300 Acc: 0.70312 0.84881 Cost: 0.84087 0.54284 0.19944 0.09858 Time: 0.00073
17-03-26 05:16:44 [1] Train Extra: lr=0.0000433 inv=0.4168750 sub=0.0000000
17-03-26 05:18:02 [1] Step: 67400 Acc: 0.69281 0.85231 Cost: 0.92341 0.62397 0.20073 0.09871 Time: 0.00074
17-03-26 05:18:02 [1] Train Extra: lr=0.0000432 inv=0.4056250 sub=0.0000000
17-03-26 05:19:21 [1] Step: 67500 Acc: 0.71375 0.84334 Cost: 0.98143 0.76899 0.11369 0.09875 Time: 0.00072
17-03-26 05:19:21 [1] Train Extra: lr=0.0000430 inv=0.4148438 sub=0.0000000
17-03-26 05:20:18 [1] Step: 67500 Eval acc: 0.68695 0.85508 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 05:20:18 [1] Eval Extra: inv=0.3763803
17-03-26 05:20:18 [1] Checkpointing with new best dev accuracy of 0.686948
17-03-26 05:21:43 [1] Step: 67600 Acc: 0.70969 0.84911 Cost: 1.04198 0.64558 0.29755 0.09885 Time: 0.00074
17-03-26 05:21:43 [1] Train Extra: lr=0.0000429 inv=0.4107812 sub=0.0000000
17-03-26 05:23:02 [1] Step: 67700 Acc: 0.70406 0.84913 Cost: 0.74116 0.46248 0.17973 0.09895 Time: 0.00073
17-03-26 05:23:02 [1] Train Extra: lr=0.0000428 inv=0.4142188 sub=0.0000000
17-03-26 05:24:34 [1] Step: 67800 Acc: 0.69812 0.84520 Cost: 0.83960 0.54849 0.19201 0.09909 Time: 0.00076
17-03-26 05:24:34 [1] Train Extra: lr=0.0000427 inv=0.4459375 sub=0.0000000
17-03-26 05:26:02 [1] Step: 67900 Acc: 0.69063 0.84642 Cost: 0.95346 0.68860 0.16567 0.09918 Time: 0.00076
17-03-26 05:26:02 [1] Train Extra: lr=0.0000425 inv=0.4379688 sub=0.0000000
17-03-26 05:27:09 [1] Step: 68000 Acc: 0.71625 0.85204 Cost: 0.86426 0.64770 0.11729 0.09927 Time: 0.00070
17-03-26 05:27:09 [1] Train Extra: lr=0.0000424 inv=0.3837500 sub=0.0000000
17-03-26 05:28:05 [1] Step: 68000 Eval acc: 0.67524 0.85044 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 05:28:05 [1] Eval Extra: inv=0.4600817
17-03-26 05:29:27 [1] Step: 68100 Acc: 0.71188 0.84957 Cost: 0.91541 0.63889 0.17713 0.09939 Time: 0.00075
17-03-26 05:29:27 [1] Train Extra: lr=0.0000423 inv=0.4176563 sub=0.0000000
17-03-26 05:30:54 [1] Step: 68200 Acc: 0.71250 0.85500 Cost: 0.67252 0.39155 0.18144 0.09953 Time: 0.00077
17-03-26 05:30:54 [1] Train Extra: lr=0.0000422 inv=0.4239062 sub=0.0000000
17-03-26 05:32:19 [1] Step: 68300 Acc: 0.70844 0.84611 Cost: 0.88107 0.63399 0.14744 0.09964 Time: 0.00075
17-03-26 05:32:19 [1] Train Extra: lr=0.0000421 inv=0.4262500 sub=0.0000000
17-03-26 05:33:38 [1] Step: 68400 Acc: 0.69406 0.84428 Cost: 0.91453 0.63478 0.17998 0.09977 Time: 0.00074
17-03-26 05:33:38 [1] Train Extra: lr=0.0000419 inv=0.4145313 sub=0.0000000
17-03-26 05:34:52 [1] Step: 68500 Acc: 0.70437 0.84650 Cost: 1.10475 0.80009 0.20479 0.09987 Time: 0.00071
17-03-26 05:34:52 [1] Train Extra: lr=0.0000418 inv=0.3937500 sub=0.0000000
17-03-26 05:35:50 [1] Step: 68500 Eval acc: 0.67314 0.85113 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 05:35:50 [1] Eval Extra: inv=0.4027164
17-03-26 05:37:13 [1] Step: 68600 Acc: 0.70312 0.84366 Cost: 1.21656 0.82883 0.28780 0.09992 Time: 0.00076
17-03-26 05:37:13 [1] Train Extra: lr=0.0000417 inv=0.4187500 sub=0.0000000
17-03-26 05:38:42 [1] Step: 68700 Acc: 0.71281 0.84670 Cost: 1.05886 0.72272 0.23612 0.10001 Time: 0.00074
17-03-26 05:38:42 [1] Train Extra: lr=0.0000416 inv=0.4537500 sub=0.0000000
17-03-26 05:40:13 [1] Step: 68800 Acc: 0.68812 0.85710 Cost: 1.18649 0.88711 0.19928 0.10011 Time: 0.00079
17-03-26 05:40:13 [1] Train Extra: lr=0.0000415 inv=0.4373437 sub=0.0000000
17-03-26 05:41:38 [1] Step: 68900 Acc: 0.68781 0.84467 Cost: 1.07495 0.80125 0.17352 0.10017 Time: 0.00073
17-03-26 05:41:38 [1] Train Extra: lr=0.0000413 inv=0.4456250 sub=0.0000000
17-03-26 05:42:52 [1] Step: 69000 Acc: 0.70469 0.85159 Cost: 1.03499 0.70872 0.22601 0.10026 Time: 0.00073
17-03-26 05:42:52 [1] Train Extra: lr=0.0000412 inv=0.4057812 sub=0.0000000
17-03-26 05:43:50 [1] Step: 69000 Eval acc: 0.68220 0.85108 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 05:43:50 [1] Eval Extra: inv=0.4628423
17-03-26 05:45:15 [1] Step: 69100 Acc: 0.68781 0.84678 Cost: 1.01937 0.76514 0.15390 0.10033 Time: 0.00076
17-03-26 05:45:15 [1] Train Extra: lr=0.0000411 inv=0.4339062 sub=0.0000000
17-03-26 05:46:32 [1] Step: 69200 Acc: 0.69344 0.84908 Cost: 1.07784 0.71309 0.26429 0.10046 Time: 0.00074
17-03-26 05:46:32 [1] Train Extra: lr=0.0000410 inv=0.4268750 sub=0.0000000
17-03-26 05:47:51 [1] Step: 69300 Acc: 0.69688 0.84835 Cost: 1.11431 0.74843 0.26529 0.10059 Time: 0.00074
17-03-26 05:47:51 [1] Train Extra: lr=0.0000409 inv=0.4064063 sub=0.0000000
17-03-26 05:49:13 [1] Step: 69400 Acc: 0.68250 0.85593 Cost: 0.93211 0.53835 0.29312 0.10064 Time: 0.00076
17-03-26 05:49:13 [1] Train Extra: lr=0.0000407 inv=0.4045313 sub=0.0000000
17-03-26 05:50:44 [1] Step: 69500 Acc: 0.68375 0.84872 Cost: 0.65301 0.44378 0.10852 0.10071 Time: 0.00077
17-03-26 05:50:44 [1] Train Extra: lr=0.0000406 inv=0.4615625 sub=0.0000000
17-03-26 05:51:41 [1] Step: 69500 Eval acc: 0.67734 0.85264 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 05:51:41 [1] Eval Extra: inv=0.4001215
17-03-26 05:52:56 [1] Step: 69600 Acc: 0.69719 0.84993 Cost: 1.01124 0.67721 0.23328 0.10076 Time: 0.00073
17-03-26 05:52:56 [1] Train Extra: lr=0.0000405 inv=0.4154687 sub=0.0000000
17-03-26 05:54:15 [1] Step: 69700 Acc: 0.70813 0.85536 Cost: 1.13679 0.77678 0.25920 0.10081 Time: 0.00076
17-03-26 05:54:15 [1] Train Extra: lr=0.0000404 inv=0.3943750 sub=0.0000000
17-03-26 05:55:47 [1] Step: 69800 Acc: 0.70688 0.85011 Cost: 0.73084 0.44426 0.18564 0.10094 Time: 0.00078
17-03-26 05:55:47 [1] Train Extra: lr=0.0000403 inv=0.4418750 sub=0.0000000
17-03-26 05:57:19 [1] Step: 69900 Acc: 0.69563 0.84702 Cost: 1.15908 0.79935 0.25865 0.10108 Time: 0.00076
17-03-26 05:57:19 [1] Train Extra: lr=0.0000402 inv=0.4521875 sub=0.0000000
17-03-26 05:58:37 [1] Step: 70000 Acc: 0.69906 0.84820 Cost: 0.95549 0.64549 0.20886 0.10115 Time: 0.00072
17-03-26 05:58:37 [1] Train Extra: lr=0.0000400 inv=0.4259375 sub=0.0000000
17-03-26 05:59:34 [1] Step: 70000 Eval acc: 0.67568 0.84897 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 05:59:34 [1] Eval Extra: inv=0.4221511
17-03-26 05:59:34 [1] Checkpointing.
17-03-26 06:00:47 [1] Step: 70100 Acc: 0.69125 0.84482 Cost: 1.18875 0.84384 0.24364 0.10126 Time: 0.00072
17-03-26 06:00:47 [1] Train Extra: lr=0.0000399 inv=0.4423437 sub=0.0000000
17-03-26 06:02:13 [1] Step: 70200 Acc: 0.68563 0.85399 Cost: 1.28517 0.87639 0.30745 0.10133 Time: 0.00077
17-03-26 06:02:13 [1] Train Extra: lr=0.0000398 inv=0.4359375 sub=0.0000000
17-03-26 06:03:30 [1] Step: 70300 Acc: 0.69563 0.85167 Cost: 1.14515 0.85280 0.19092 0.10143 Time: 0.00074
17-03-26 06:03:30 [1] Train Extra: lr=0.0000397 inv=0.4170313 sub=0.0000000
17-03-26 06:04:45 [1] Step: 70400 Acc: 0.70625 0.85197 Cost: 1.00943 0.67833 0.22952 0.10158 Time: 0.00071
17-03-26 06:04:45 [1] Train Extra: lr=0.0000396 inv=0.4214062 sub=0.0000000
17-03-26 06:06:10 [1] Step: 70500 Acc: 0.68469 0.85367 Cost: 1.24770 0.90611 0.24000 0.10160 Time: 0.00078
17-03-26 06:06:10 [1] Train Extra: lr=0.0000395 inv=0.4195313 sub=0.0000000
17-03-26 06:07:09 [1] Step: 70500 Eval acc: 0.67900 0.85448 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 06:07:09 [1] Eval Extra: inv=0.4217646
17-03-26 06:08:29 [1] Step: 70600 Acc: 0.69563 0.84556 Cost: 1.23225 0.91716 0.21341 0.10167 Time: 0.00073
17-03-26 06:08:29 [1] Train Extra: lr=0.0000394 inv=0.4323438 sub=0.0000000
17-03-26 06:09:47 [1] Step: 70700 Acc: 0.69719 0.85032 Cost: 1.00845 0.69782 0.20893 0.10171 Time: 0.00075
17-03-26 06:09:47 [1] Train Extra: lr=0.0000392 inv=0.4028125 sub=0.0000000
17-03-26 06:11:01 [1] Step: 70800 Acc: 0.69500 0.85145 Cost: 0.82622 0.60098 0.12342 0.10182 Time: 0.00071
17-03-26 06:11:01 [1] Train Extra: lr=0.0000391 inv=0.4003125 sub=0.0000000
17-03-26 06:12:26 [1] Step: 70900 Acc: 0.66969 0.84472 Cost: 1.14827 0.75464 0.29174 0.10189 Time: 0.00075
17-03-26 06:12:26 [1] Train Extra: lr=0.0000390 inv=0.4248438 sub=0.0000000
17-03-26 06:13:39 [1] Step: 71000 Acc: 0.70688 0.85583 Cost: 0.86689 0.55753 0.20740 0.10197 Time: 0.00072
17-03-26 06:13:39 [1] Train Extra: lr=0.0000389 inv=0.4148438 sub=0.0000000
17-03-26 06:14:35 [1] Step: 71000 Eval acc: 0.67469 0.84375 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 06:14:35 [1] Eval Extra: inv=0.4492602
17-03-26 06:15:53 [1] Step: 71100 Acc: 0.69594 0.84424 Cost: 1.05822 0.74920 0.20705 0.10197 Time: 0.00074
17-03-26 06:15:53 [1] Train Extra: lr=0.0000388 inv=0.4087500 sub=0.0000000
17-03-26 06:17:20 [1] Step: 71200 Acc: 0.69281 0.84738 Cost: 1.13173 0.78810 0.24161 0.10202 Time: 0.00076
17-03-26 06:17:20 [1] Train Extra: lr=0.0000387 inv=0.4370312 sub=0.0000000
17-03-26 06:18:47 [1] Step: 71300 Acc: 0.67375 0.85573 Cost: 1.08779 0.72079 0.26494 0.10207 Time: 0.00078
17-03-26 06:18:47 [1] Train Extra: lr=0.0000386 inv=0.4004687 sub=0.0000000
17-03-26 06:20:18 [1] Step: 71400 Acc: 0.67094 0.84501 Cost: 0.90941 0.63958 0.16772 0.10212 Time: 0.00076
17-03-26 06:20:18 [1] Train Extra: lr=0.0000385 inv=0.4534375 sub=0.0000000
17-03-26 06:21:37 [1] Step: 71500 Acc: 0.70344 0.84783 Cost: 0.81444 0.59665 0.11566 0.10213 Time: 0.00075
17-03-26 06:21:37 [1] Train Extra: lr=0.0000384 inv=0.4242187 sub=0.0000000
17-03-26 06:22:35 [1] Step: 71500 Eval acc: 0.67999 0.84656 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 06:22:35 [1] Eval Extra: inv=0.4278379
17-03-26 06:23:48 [1] Step: 71600 Acc: 0.69375 0.85116 Cost: 0.87603 0.55882 0.21501 0.10220 Time: 0.00071
17-03-26 06:23:48 [1] Train Extra: lr=0.0000382 inv=0.4114062 sub=0.0000000
17-03-26 06:25:13 [1] Step: 71700 Acc: 0.68875 0.84751 Cost: 1.21981 0.86619 0.25129 0.10233 Time: 0.00076
17-03-26 06:25:13 [1] Train Extra: lr=0.0000381 inv=0.4134375 sub=0.0000000
17-03-26 06:26:33 [1] Step: 71800 Acc: 0.70594 0.85057 Cost: 0.95546 0.63893 0.21412 0.10241 Time: 0.00073
17-03-26 06:26:33 [1] Train Extra: lr=0.0000380 inv=0.4195313 sub=0.0000000
17-03-26 06:27:56 [1] Step: 71900 Acc: 0.69375 0.85161 Cost: 0.84686 0.63876 0.10563 0.10246 Time: 0.00075
17-03-26 06:27:56 [1] Train Extra: lr=0.0000379 inv=0.4176563 sub=0.0000000
17-03-26 06:29:18 [1] Step: 72000 Acc: 0.70219 0.84859 Cost: 1.05856 0.79266 0.16344 0.10246 Time: 0.00073
17-03-26 06:29:18 [1] Train Extra: lr=0.0000378 inv=0.4021875 sub=0.0000000
17-03-26 06:30:16 [1] Step: 72000 Eval acc: 0.68110 0.85074 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 06:30:16 [1] Eval Extra: inv=0.3975817
17-03-26 06:31:40 [1] Step: 72100 Acc: 0.68125 0.84587 Cost: 1.26155 0.95429 0.20479 0.10247 Time: 0.00075
17-03-26 06:31:40 [1] Train Extra: lr=0.0000377 inv=0.4173438 sub=0.0000000
17-03-26 06:32:55 [1] Step: 72200 Acc: 0.69937 0.85467 Cost: 0.89122 0.58148 0.20722 0.10252 Time: 0.00074
17-03-26 06:32:55 [1] Train Extra: lr=0.0000376 inv=0.3848437 sub=0.0000000
17-03-26 06:34:25 [1] Step: 72300 Acc: 0.68781 0.84450 Cost: 0.79912 0.50788 0.18866 0.10258 Time: 0.00076
17-03-26 06:34:25 [1] Train Extra: lr=0.0000375 inv=0.4417187 sub=0.0000000
17-03-26 06:36:04 [1] Step: 72400 Acc: 0.67937 0.85022 Cost: 0.86400 0.51125 0.25015 0.10261 Time: 0.00080
17-03-26 06:36:04 [1] Train Extra: lr=0.0000374 inv=0.4275000 sub=0.0000000
17-03-26 06:37:25 [1] Step: 72500 Acc: 0.69812 0.85642 Cost: 1.10511 0.88322 0.11929 0.10260 Time: 0.00076
17-03-26 06:37:25 [1] Train Extra: lr=0.0000373 inv=0.3848437 sub=0.0000000
17-03-26 06:38:22 [1] Step: 72500 Eval acc: 0.68319 0.84982 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 06:38:22 [1] Eval Extra: inv=0.4211020
17-03-26 06:39:47 [1] Step: 72600 Acc: 0.69000 0.84865 Cost: 1.00339 0.68142 0.21938 0.10259 Time: 0.00076
17-03-26 06:39:47 [1] Train Extra: lr=0.0000372 inv=0.4151563 sub=0.0000000
17-03-26 06:41:07 [1] Step: 72700 Acc: 0.68156 0.85221 Cost: 0.89719 0.59446 0.20004 0.10270 Time: 0.00076
17-03-26 06:41:07 [1] Train Extra: lr=0.0000371 inv=0.3917188 sub=0.0000000
17-03-26 06:42:39 [1] Step: 72800 Acc: 0.69125 0.84239 Cost: 1.12727 0.81118 0.21343 0.10266 Time: 0.00077
17-03-26 06:42:39 [1] Train Extra: lr=0.0000369 inv=0.4540625 sub=0.0000000
17-03-26 06:43:58 [1] Step: 72900 Acc: 0.68719 0.84707 Cost: 1.10761 0.74983 0.25512 0.10267 Time: 0.00073
17-03-26 06:43:58 [1] Train Extra: lr=0.0000368 inv=0.4448437 sub=0.0000000
17-03-26 06:45:17 [1] Step: 73000 Acc: 0.69688 0.84748 Cost: 1.02615 0.68214 0.24136 0.10265 Time: 0.00072
17-03-26 06:45:17 [1] Train Extra: lr=0.0000367 inv=0.4562500 sub=0.0000000
17-03-26 06:46:14 [1] Step: 73000 Eval acc: 0.68087 0.85101 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 06:46:14 [1] Eval Extra: inv=0.4192800
17-03-26 06:47:36 [1] Step: 73100 Acc: 0.69063 0.84432 Cost: 1.20340 0.89697 0.20379 0.10265 Time: 0.00073
17-03-26 06:47:36 [1] Train Extra: lr=0.0000366 inv=0.4470312 sub=0.0000000
17-03-26 06:48:59 [1] Step: 73200 Acc: 0.68625 0.85171 Cost: 0.96586 0.68919 0.17392 0.10274 Time: 0.00076
17-03-26 06:48:59 [1] Train Extra: lr=0.0000365 inv=0.4298438 sub=0.0000000
17-03-26 06:50:18 [1] Step: 73300 Acc: 0.70781 0.84492 Cost: 0.97673 0.65182 0.22213 0.10278 Time: 0.00073
17-03-26 06:50:18 [1] Train Extra: lr=0.0000364 inv=0.4092188 sub=0.0000000
17-03-26 06:51:44 [1] Step: 73400 Acc: 0.68250 0.85331 Cost: 1.17196 0.80753 0.26161 0.10282 Time: 0.00077
17-03-26 06:51:44 [1] Train Extra: lr=0.0000363 inv=0.4160937 sub=0.0000000
17-03-26 06:53:05 [1] Step: 73500 Acc: 0.68688 0.84496 Cost: 1.18856 0.84121 0.24448 0.10287 Time: 0.00072
17-03-26 06:53:05 [1] Train Extra: lr=0.0000362 inv=0.4307813 sub=0.0000000
17-03-26 06:54:03 [1] Step: 73500 Eval acc: 0.67856 0.85343 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 06:54:03 [1] Eval Extra: inv=0.3680985
17-03-26 06:55:29 [1] Step: 73600 Acc: 0.69875 0.84930 Cost: 0.97184 0.75682 0.11208 0.10294 Time: 0.00076
17-03-26 06:55:29 [1] Train Extra: lr=0.0000361 inv=0.4003125 sub=0.0000000
17-03-26 06:56:55 [1] Step: 73700 Acc: 0.67625 0.85556 Cost: 0.92691 0.61836 0.20550 0.10305 Time: 0.00076
17-03-26 06:56:55 [1] Train Extra: lr=0.0000360 inv=0.4275000 sub=0.0000000
17-03-26 06:58:12 [1] Step: 73800 Acc: 0.68844 0.85340 Cost: 1.02405 0.66101 0.25998 0.10306 Time: 0.00074
17-03-26 06:58:12 [1] Train Extra: lr=0.0000359 inv=0.3918750 sub=0.0000000
17-03-26 06:59:42 [1] Step: 73900 Acc: 0.68875 0.85126 Cost: 1.31282 0.93945 0.27033 0.10305 Time: 0.00078
17-03-26 06:59:42 [1] Train Extra: lr=0.0000358 inv=0.4434375 sub=0.0000000
17-03-26 07:01:03 [1] Step: 74000 Acc: 0.69969 0.84901 Cost: 0.95442 0.76995 0.08141 0.10305 Time: 0.00074
17-03-26 07:01:03 [1] Train Extra: lr=0.0000357 inv=0.4118750 sub=0.0000000
17-03-26 07:02:01 [1] Step: 74000 Eval acc: 0.68275 0.84878 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 07:02:01 [1] Eval Extra: inv=0.4107222
17-03-26 07:03:19 [1] Step: 74100 Acc: 0.68719 0.84664 Cost: 1.06635 0.66807 0.29516 0.10313 Time: 0.00073
17-03-26 07:03:19 [1] Train Extra: lr=0.0000356 inv=0.4151563 sub=0.0000000
17-03-26 07:04:44 [1] Step: 74200 Acc: 0.70125 0.85403 Cost: 1.09624 0.72345 0.26966 0.10313 Time: 0.00078
17-03-26 07:04:44 [1] Train Extra: lr=0.0000355 inv=0.4146875 sub=0.0000000
17-03-26 07:06:03 [1] Step: 74300 Acc: 0.68656 0.85450 Cost: 1.02527 0.72351 0.19862 0.10314 Time: 0.00077
17-03-26 07:06:03 [1] Train Extra: lr=0.0000354 inv=0.4012500 sub=0.0000000
17-03-26 07:07:24 [1] Step: 74400 Acc: 0.68188 0.84506 Cost: 1.21540 0.97259 0.13955 0.10325 Time: 0.00072
17-03-26 07:07:24 [1] Train Extra: lr=0.0000353 inv=0.4667188 sub=0.0000000
17-03-26 07:08:37 [1] Step: 74500 Acc: 0.69375 0.85057 Cost: 0.88728 0.60322 0.18074 0.10331 Time: 0.00073
17-03-26 07:08:37 [1] Train Extra: lr=0.0000352 inv=0.4056250 sub=0.0000000
17-03-26 07:09:34 [1] Step: 74500 Eval acc: 0.68595 0.84970 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 07:09:34 [1] Eval Extra: inv=0.4315923
17-03-26 07:11:09 [1] Step: 74600 Acc: 0.67781 0.85792 Cost: 0.97111 0.65780 0.20994 0.10337 Time: 0.00082
17-03-26 07:11:09 [1] Train Extra: lr=0.0000351 inv=0.4289062 sub=0.0000000
17-03-26 07:12:24 [1] Step: 74700 Acc: 0.70656 0.84436 Cost: 1.01894 0.73619 0.17940 0.10335 Time: 0.00069
17-03-26 07:12:24 [1] Train Extra: lr=0.0000350 inv=0.4306250 sub=0.0000000
17-03-26 07:13:48 [1] Step: 74800 Acc: 0.67344 0.84644 Cost: 1.11817 0.77587 0.23884 0.10346 Time: 0.00074
17-03-26 07:13:48 [1] Train Extra: lr=0.0000349 inv=0.4357813 sub=0.0000000
17-03-26 07:15:11 [1] Step: 74900 Acc: 0.69469 0.84780 Cost: 0.85767 0.55169 0.20251 0.10346 Time: 0.00075
17-03-26 07:15:11 [1] Train Extra: lr=0.0000348 inv=0.4487500 sub=0.0000000
17-03-26 07:16:39 [1] Step: 75000 Acc: 0.69781 0.85035 Cost: 1.00767 0.61314 0.29108 0.10346 Time: 0.00075
17-03-26 07:16:39 [1] Train Extra: lr=0.0000347 inv=0.4473437 sub=0.0000000
17-03-26 07:17:37 [1] Step: 75000 Eval acc: 0.68441 0.85708 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 07:17:37 [1] Eval Extra: inv=0.3902385
17-03-26 07:17:37 [1] Checkpointing.
17-03-26 07:18:57 [1] Step: 75100 Acc: 0.69312 0.84524 Cost: 0.85026 0.59672 0.14997 0.10357 Time: 0.00071
17-03-26 07:18:57 [1] Train Extra: lr=0.0000346 inv=0.4598437 sub=0.0000000
17-03-26 07:20:19 [1] Step: 75200 Acc: 0.70281 0.84278 Cost: 0.99925 0.71408 0.18162 0.10355 Time: 0.00074
17-03-26 07:20:19 [1] Train Extra: lr=0.0000345 inv=0.4560938 sub=0.0000000
17-03-26 07:21:43 [1] Step: 75300 Acc: 0.68719 0.84595 Cost: 1.02414 0.68261 0.23798 0.10355 Time: 0.00073
17-03-26 07:21:43 [1] Train Extra: lr=0.0000344 inv=0.4406250 sub=0.0000000
17-03-26 07:22:58 [1] Step: 75400 Acc: 0.68531 0.85247 Cost: 1.00351 0.72137 0.17854 0.10360 Time: 0.00072
17-03-26 07:22:58 [1] Train Extra: lr=0.0000343 inv=0.4026562 sub=0.0000000
17-03-26 07:24:15 [1] Step: 75500 Acc: 0.68625 0.84898 Cost: 1.14310 0.74642 0.29307 0.10361 Time: 0.00074
17-03-26 07:24:15 [1] Train Extra: lr=0.0000342 inv=0.4237500 sub=0.0000000
17-03-26 07:25:14 [1] Step: 75500 Eval acc: 0.68386 0.85355 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 07:25:14 [1] Eval Extra: inv=0.4375000
17-03-26 07:26:40 [1] Step: 75600 Acc: 0.70531 0.84850 Cost: 1.07906 0.73865 0.23672 0.10369 Time: 0.00076
17-03-26 07:26:40 [1] Train Extra: lr=0.0000341 inv=0.4490625 sub=0.0000000
17-03-26 07:28:01 [1] Step: 75700 Acc: 0.69281 0.84885 Cost: 1.05946 0.70105 0.25460 0.10381 Time: 0.00072
17-03-26 07:28:01 [1] Train Extra: lr=0.0000340 inv=0.4390625 sub=0.0000000
17-03-26 07:29:26 [1] Step: 75800 Acc: 0.71188 0.85214 Cost: 0.94843 0.68666 0.15790 0.10388 Time: 0.00076
17-03-26 07:29:26 [1] Train Extra: lr=0.0000339 inv=0.4098438 sub=0.0000000
17-03-26 07:30:45 [1] Step: 75900 Acc: 0.71062 0.85184 Cost: 0.88214 0.49744 0.28061 0.10409 Time: 0.00074
17-03-26 07:30:45 [1] Train Extra: lr=0.0000338 inv=0.4032812 sub=0.0000000
17-03-26 07:32:11 [1] Step: 76000 Acc: 0.71531 0.85040 Cost: 0.92764 0.69072 0.13272 0.10420 Time: 0.00077
17-03-26 07:32:11 [1] Train Extra: lr=0.0000337 inv=0.4140625 sub=0.0000000
17-03-26 07:33:08 [1] Step: 76000 Eval acc: 0.68595 0.85172 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 07:33:08 [1] Eval Extra: inv=0.4135932
17-03-26 07:34:29 [1] Step: 76100 Acc: 0.71375 0.85301 Cost: 1.07867 0.68043 0.29392 0.10432 Time: 0.00074
17-03-26 07:34:29 [1] Train Extra: lr=0.0000336 inv=0.4275000 sub=0.0000000
17-03-26 07:35:48 [1] Step: 76200 Acc: 0.71219 0.85001 Cost: 0.71725 0.47010 0.14270 0.10444 Time: 0.00075
17-03-26 07:35:48 [1] Train Extra: lr=0.0000335 inv=0.4096875 sub=0.0000000
17-03-26 07:37:04 [1] Step: 76300 Acc: 0.71875 0.84987 Cost: 0.88312 0.67377 0.10479 0.10456 Time: 0.00073
17-03-26 07:37:04 [1] Train Extra: lr=0.0000334 inv=0.3942188 sub=0.0000000
17-03-26 07:38:29 [1] Step: 76400 Acc: 0.70188 0.84781 Cost: 0.88498 0.63177 0.14860 0.10461 Time: 0.00078
17-03-26 07:38:29 [1] Train Extra: lr=0.0000333 inv=0.4287500 sub=0.0000000
17-03-26 07:39:53 [1] Step: 76500 Acc: 0.71969 0.85052 Cost: 0.80456 0.51525 0.18457 0.10474 Time: 0.00075
17-03-26 07:39:53 [1] Train Extra: lr=0.0000332 inv=0.3978125 sub=0.0000000
17-03-26 07:40:50 [1] Step: 76500 Eval acc: 0.68375 0.85451 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 07:40:50 [1] Eval Extra: inv=0.4133723
17-03-26 07:42:21 [1] Step: 76600 Acc: 0.69063 0.85002 Cost: 0.78625 0.49204 0.18935 0.10487 Time: 0.00078
17-03-26 07:42:21 [1] Train Extra: lr=0.0000331 inv=0.4662500 sub=0.0000000
17-03-26 07:43:41 [1] Step: 76700 Acc: 0.70875 0.84991 Cost: 1.04260 0.76917 0.16842 0.10500 Time: 0.00073
17-03-26 07:43:41 [1] Train Extra: lr=0.0000330 inv=0.4321875 sub=0.0000000
17-03-26 07:45:07 [1] Step: 76800 Acc: 0.71469 0.84821 Cost: 0.97134 0.65865 0.20764 0.10505 Time: 0.00076
17-03-26 07:45:07 [1] Train Extra: lr=0.0000329 inv=0.4421875 sub=0.0000000
17-03-26 07:46:41 [1] Step: 76900 Acc: 0.68781 0.85276 Cost: 1.14812 0.82347 0.21946 0.10519 Time: 0.00082
17-03-26 07:46:41 [1] Train Extra: lr=0.0000328 inv=0.4346875 sub=0.0000000
17-03-26 07:48:04 [1] Step: 77000 Acc: 0.71250 0.84489 Cost: 1.09739 0.73112 0.26099 0.10528 Time: 0.00073
17-03-26 07:48:04 [1] Train Extra: lr=0.0000327 inv=0.4590625 sub=0.0000000
17-03-26 07:49:03 [1] Step: 77000 Eval acc: 0.67878 0.85340 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 07:49:03 [1] Eval Extra: inv=0.4080720
17-03-26 07:50:21 [1] Step: 77100 Acc: 0.70906 0.85016 Cost: 0.92373 0.70608 0.11227 0.10538 Time: 0.00075
17-03-26 07:50:21 [1] Train Extra: lr=0.0000326 inv=0.4137500 sub=0.0000000
17-03-26 07:51:37 [1] Step: 77200 Acc: 0.71281 0.84828 Cost: 0.66890 0.45447 0.10897 0.10546 Time: 0.00073
17-03-26 07:51:37 [1] Train Extra: lr=0.0000326 inv=0.4004687 sub=0.0000000
17-03-26 07:52:55 [1] Step: 77300 Acc: 0.70625 0.84252 Cost: 1.24621 0.84827 0.29241 0.10553 Time: 0.00072
17-03-26 07:52:55 [1] Train Extra: lr=0.0000325 inv=0.4259375 sub=0.0000000
17-03-26 07:54:19 [1] Step: 77400 Acc: 0.70562 0.84682 Cost: 0.74487 0.52659 0.11265 0.10563 Time: 0.00075
17-03-26 07:54:19 [1] Train Extra: lr=0.0000324 inv=0.4526562 sub=0.0000000
17-03-26 07:55:35 [1] Step: 77500 Acc: 0.72000 0.85341 Cost: 0.93089 0.63404 0.19112 0.10573 Time: 0.00073
17-03-26 07:55:35 [1] Train Extra: lr=0.0000323 inv=0.3903125 sub=0.0000000
17-03-26 07:56:32 [1] Step: 77500 Eval acc: 0.67867 0.84771 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 07:56:32 [1] Eval Extra: inv=0.4296047
17-03-26 07:58:03 [1] Step: 77600 Acc: 0.70125 0.84290 Cost: 1.04719 0.76670 0.17471 0.10579 Time: 0.00078
17-03-26 07:58:03 [1] Train Extra: lr=0.0000322 inv=0.4662500 sub=0.0000000
17-03-26 07:59:24 [1] Step: 77700 Acc: 0.69812 0.84633 Cost: 0.82699 0.60585 0.11532 0.10582 Time: 0.00074
17-03-26 07:59:24 [1] Train Extra: lr=0.0000321 inv=0.4351563 sub=0.0000000
17-03-26 08:00:45 [1] Step: 77800 Acc: 0.70188 0.84472 Cost: 1.06336 0.73248 0.22495 0.10593 Time: 0.00074
17-03-26 08:00:45 [1] Train Extra: lr=0.0000320 inv=0.4193750 sub=0.0000000
17-03-26 08:02:10 [1] Step: 77900 Acc: 0.71250 0.84951 Cost: 1.18152 0.82282 0.25266 0.10605 Time: 0.00078
17-03-26 08:02:10 [1] Train Extra: lr=0.0000319 inv=0.4315625 sub=0.0000000
17-03-26 08:03:31 [1] Step: 78000 Acc: 0.69750 0.85057 Cost: 1.19247 0.83345 0.25291 0.10612 Time: 0.00074
17-03-26 08:03:31 [1] Train Extra: lr=0.0000318 inv=0.4401563 sub=0.0000000
17-03-26 08:04:29 [1] Step: 78000 Eval acc: 0.67922 0.84847 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 08:04:29 [1] Eval Extra: inv=0.4351811
17-03-26 08:05:50 [1] Step: 78100 Acc: 0.71375 0.84686 Cost: 1.21986 0.85973 0.25390 0.10622 Time: 0.00075
17-03-26 08:05:50 [1] Train Extra: lr=0.0000317 inv=0.4260937 sub=0.0000000
17-03-26 08:07:15 [1] Step: 78200 Acc: 0.70469 0.84987 Cost: 1.12132 0.72693 0.28810 0.10629 Time: 0.00077
17-03-26 08:07:15 [1] Train Extra: lr=0.0000316 inv=0.4312500 sub=0.0000000
17-03-26 08:08:38 [1] Step: 78300 Acc: 0.70281 0.84476 Cost: 1.02861 0.64331 0.27898 0.10632 Time: 0.00074
17-03-26 08:08:38 [1] Train Extra: lr=0.0000315 inv=0.4295312 sub=0.0000000
17-03-26 08:09:59 [1] Step: 78400 Acc: 0.70156 0.85152 Cost: 0.85984 0.61291 0.14052 0.10640 Time: 0.00075
17-03-26 08:09:59 [1] Train Extra: lr=0.0000314 inv=0.4246875 sub=0.0000000
17-03-26 08:11:13 [1] Step: 78500 Acc: 0.72031 0.85249 Cost: 1.15504 0.80675 0.24185 0.10645 Time: 0.00073
17-03-26 08:11:13 [1] Train Extra: lr=0.0000314 inv=0.4167188 sub=0.0000000
17-03-26 08:12:11 [1] Step: 78500 Eval acc: 0.68386 0.85263 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 08:12:11 [1] Eval Extra: inv=0.4079064
17-03-26 08:13:25 [1] Step: 78600 Acc: 0.68312 0.84833 Cost: 0.89365 0.61571 0.17137 0.10657 Time: 0.00072
17-03-26 08:13:25 [1] Train Extra: lr=0.0000313 inv=0.3925000 sub=0.0000000
17-03-26 08:14:43 [1] Step: 78700 Acc: 0.68906 0.84869 Cost: 0.94156 0.73029 0.10465 0.10663 Time: 0.00075
17-03-26 08:14:43 [1] Train Extra: lr=0.0000312 inv=0.3964063 sub=0.0000000
17-03-26 08:16:04 [1] Step: 78800 Acc: 0.69188 0.84684 Cost: 1.00252 0.71426 0.18149 0.10677 Time: 0.00073
17-03-26 08:16:04 [1] Train Extra: lr=0.0000311 inv=0.4181250 sub=0.0000000
17-03-26 08:17:31 [1] Step: 78900 Acc: 0.69125 0.84389 Cost: 1.12837 0.76517 0.25632 0.10689 Time: 0.00077
17-03-26 08:17:31 [1] Train Extra: lr=0.0000310 inv=0.4287500 sub=0.0000000
17-03-26 08:18:56 [1] Step: 79000 Acc: 0.69688 0.84523 Cost: 1.01800 0.72679 0.18427 0.10694 Time: 0.00076
17-03-26 08:18:56 [1] Train Extra: lr=0.0000309 inv=0.4301563 sub=0.0000000
17-03-26 08:19:54 [1] Step: 79000 Eval acc: 0.68419 0.85264 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 08:19:54 [1] Eval Extra: inv=0.4399293
17-03-26 08:21:19 [1] Step: 79100 Acc: 0.69500 0.85865 Cost: 1.14328 0.85570 0.18065 0.10693 Time: 0.00080
17-03-26 08:21:19 [1] Train Extra: lr=0.0000308 inv=0.3909375 sub=0.0000000
17-03-26 08:22:43 [1] Step: 79200 Acc: 0.69500 0.83753 Cost: 1.16222 0.76730 0.28795 0.10697 Time: 0.00072
17-03-26 08:22:43 [1] Train Extra: lr=0.0000307 inv=0.4493750 sub=0.0000000
17-03-26 08:24:08 [1] Step: 79300 Acc: 0.69812 0.84680 Cost: 1.09809 0.78533 0.20573 0.10703 Time: 0.00077
17-03-26 08:24:08 [1] Train Extra: lr=0.0000306 inv=0.4254688 sub=0.0000000
17-03-26 08:25:30 [1] Step: 79400 Acc: 0.69219 0.84647 Cost: 0.91785 0.65313 0.15768 0.10704 Time: 0.00075
17-03-26 08:25:30 [1] Train Extra: lr=0.0000306 inv=0.4282813 sub=0.0000000
17-03-26 08:27:09 [1] Step: 79500 Acc: 0.70000 0.85637 Cost: 1.58322 1.13854 0.33749 0.10719 Time: 0.00082
17-03-26 08:27:09 [1] Train Extra: lr=0.0000305 inv=0.4353125 sub=0.0000000
17-03-26 08:28:05 [1] Step: 79500 Eval acc: 0.67845 0.85661 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 08:28:05 [1] Eval Extra: inv=0.4662102
17-03-26 08:29:25 [1] Step: 79600 Acc: 0.71219 0.85093 Cost: 1.15055 0.77819 0.26508 0.10727 Time: 0.00074
17-03-26 08:29:25 [1] Train Extra: lr=0.0000304 inv=0.4300000 sub=0.0000000
17-03-26 08:30:51 [1] Step: 79700 Acc: 0.68781 0.84637 Cost: 1.03832 0.75876 0.17225 0.10731 Time: 0.00076
17-03-26 08:30:51 [1] Train Extra: lr=0.0000303 inv=0.4567188 sub=0.0000000
17-03-26 08:32:19 [1] Step: 79800 Acc: 0.68312 0.84992 Cost: 1.02028 0.71342 0.19947 0.10738 Time: 0.00075
17-03-26 08:32:19 [1] Train Extra: lr=0.0000302 inv=0.4503125 sub=0.0000000
17-03-26 08:33:45 [1] Step: 79900 Acc: 0.70375 0.85137 Cost: 0.94263 0.61790 0.21729 0.10744 Time: 0.00078
17-03-26 08:33:45 [1] Train Extra: lr=0.0000301 inv=0.4357813 sub=0.0000000
17-03-26 08:35:07 [1] Step: 80000 Acc: 0.69094 0.85294 Cost: 0.89968 0.65742 0.13469 0.10757 Time: 0.00075
17-03-26 08:35:07 [1] Train Extra: lr=0.0000300 inv=0.4209375 sub=0.0000000
17-03-26 08:36:08 [1] Step: 80000 Eval acc: 0.68419 0.85779 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00018
17-03-26 08:36:08 [1] Eval Extra: inv=0.3768220
17-03-26 08:36:08 [1] Checkpointing.
17-03-26 08:37:31 [1] Step: 80100 Acc: 0.70906 0.84762 Cost: 1.18177 0.77960 0.29458 0.10759 Time: 0.00074
17-03-26 08:37:31 [1] Train Extra: lr=0.0000299 inv=0.4146875 sub=0.0000000
17-03-26 08:38:54 [1] Step: 80200 Acc: 0.70719 0.84433 Cost: 0.95950 0.67661 0.17535 0.10754 Time: 0.00074
17-03-26 08:38:54 [1] Train Extra: lr=0.0000299 inv=0.4303125 sub=0.0000000
17-03-26 08:40:16 [1] Step: 80300 Acc: 0.69500 0.84871 Cost: 1.13712 0.76466 0.26485 0.10761 Time: 0.00074
17-03-26 08:40:16 [1] Train Extra: lr=0.0000298 inv=0.4298438 sub=0.0000000
17-03-26 08:41:41 [1] Step: 80400 Acc: 0.70719 0.84981 Cost: 0.83878 0.56861 0.16253 0.10764 Time: 0.00077
17-03-26 08:41:41 [1] Train Extra: lr=0.0000297 inv=0.4081250 sub=0.0000000
17-03-26 08:43:01 [1] Step: 80500 Acc: 0.69406 0.85537 Cost: 0.86848 0.54172 0.21909 0.10767 Time: 0.00076
17-03-26 08:43:01 [1] Train Extra: lr=0.0000296 inv=0.4051562 sub=0.0000000
17-03-26 08:43:58 [1] Step: 80500 Eval acc: 0.68352 0.85290 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 08:43:58 [1] Eval Extra: inv=0.4273962
17-03-26 08:45:23 [1] Step: 80600 Acc: 0.69781 0.84912 Cost: 0.97807 0.66427 0.20609 0.10770 Time: 0.00075
17-03-26 08:45:23 [1] Train Extra: lr=0.0000295 inv=0.4404688 sub=0.0000000
17-03-26 08:46:44 [1] Step: 80700 Acc: 0.69219 0.85822 Cost: 0.80889 0.51119 0.18994 0.10776 Time: 0.00077
17-03-26 08:46:44 [1] Train Extra: lr=0.0000294 inv=0.4189062 sub=0.0000000
17-03-26 08:48:05 [1] Step: 80800 Acc: 0.70031 0.85091 Cost: 0.90031 0.63551 0.15700 0.10781 Time: 0.00073
17-03-26 08:48:05 [1] Train Extra: lr=0.0000294 inv=0.4075000 sub=0.0000000
17-03-26 08:49:31 [1] Step: 80900 Acc: 0.67688 0.84614 Cost: 0.92408 0.59420 0.22196 0.10792 Time: 0.00074
17-03-26 08:49:31 [1] Train Extra: lr=0.0000293 inv=0.4448437 sub=0.0000000
17-03-26 08:50:56 [1] Step: 81000 Acc: 0.69688 0.85402 Cost: 0.99381 0.62347 0.26236 0.10799 Time: 0.00079
17-03-26 08:50:56 [1] Train Extra: lr=0.0000292 inv=0.3920312 sub=0.0000000
17-03-26 08:51:54 [1] Step: 81000 Eval acc: 0.68640 0.85457 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 08:51:54 [1] Eval Extra: inv=0.4232553
17-03-26 08:53:19 [1] Step: 81100 Acc: 0.69500 0.84824 Cost: 1.26756 0.89053 0.26895 0.10808 Time: 0.00078
17-03-26 08:53:19 [1] Train Extra: lr=0.0000291 inv=0.4260937 sub=0.0000000
17-03-26 08:54:28 [1] Step: 81200 Acc: 0.69656 0.85047 Cost: 0.83496 0.55407 0.17279 0.10809 Time: 0.00070
17-03-26 08:54:28 [1] Train Extra: lr=0.0000290 inv=0.4103125 sub=0.0000000
17-03-26 08:56:01 [1] Step: 81300 Acc: 0.70688 0.85886 Cost: 1.09301 0.74748 0.23743 0.10810 Time: 0.00082
17-03-26 08:56:01 [1] Train Extra: lr=0.0000289 inv=0.4271875 sub=0.0000000
17-03-26 08:57:34 [1] Step: 81400 Acc: 0.67188 0.85458 Cost: 1.19882 0.77827 0.31235 0.10820 Time: 0.00080
17-03-26 08:57:34 [1] Train Extra: lr=0.0000288 inv=0.4279688 sub=0.0000000
17-03-26 08:58:57 [1] Step: 81500 Acc: 0.68469 0.85027 Cost: 1.21737 0.80326 0.30587 0.10823 Time: 0.00076
17-03-26 08:58:57 [1] Train Extra: lr=0.0000288 inv=0.4306250 sub=0.0000000
17-03-26 08:59:54 [1] Step: 81500 Eval acc: 0.68617 0.85530 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 08:59:54 [1] Eval Extra: inv=0.4455610
17-03-26 09:01:14 [1] Step: 81600 Acc: 0.71313 0.84916 Cost: 0.86964 0.59316 0.16822 0.10825 Time: 0.00072
17-03-26 09:01:14 [1] Train Extra: lr=0.0000287 inv=0.4164062 sub=0.0000000
17-03-26 09:02:31 [1] Step: 81700 Acc: 0.70063 0.84955 Cost: 0.98840 0.62763 0.25249 0.10829 Time: 0.00073
17-03-26 09:02:31 [1] Train Extra: lr=0.0000286 inv=0.4339062 sub=0.0000000
17-03-26 09:03:56 [1] Step: 81800 Acc: 0.69656 0.84384 Cost: 1.03177 0.65980 0.26363 0.10834 Time: 0.00075
17-03-26 09:03:56 [1] Train Extra: lr=0.0000285 inv=0.4293750 sub=0.0000000
17-03-26 09:05:15 [1] Step: 81900 Acc: 0.70813 0.85402 Cost: 0.75533 0.54907 0.09795 0.10831 Time: 0.00076
17-03-26 09:05:15 [1] Train Extra: lr=0.0000284 inv=0.3987500 sub=0.0000000
17-03-26 09:06:48 [1] Step: 82000 Acc: 0.68719 0.85284 Cost: 0.83740 0.56955 0.15954 0.10832 Time: 0.00079
17-03-26 09:06:48 [1] Train Extra: lr=0.0000284 inv=0.4314062 sub=0.0000000
17-03-26 09:07:46 [1] Step: 82000 Eval acc: 0.68684 0.85145 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 09:07:46 [1] Eval Extra: inv=0.4394876
17-03-26 09:09:12 [1] Step: 82100 Acc: 0.69469 0.85311 Cost: 1.15141 0.89871 0.14432 0.10838 Time: 0.00076
17-03-26 09:09:12 [1] Train Extra: lr=0.0000283 inv=0.4268750 sub=0.0000000
17-03-26 09:10:26 [1] Step: 82200 Acc: 0.70562 0.84962 Cost: 0.94508 0.76031 0.07637 0.10840 Time: 0.00071
17-03-26 09:10:26 [1] Train Extra: lr=0.0000282 inv=0.4087500 sub=0.0000000
17-03-26 09:11:49 [1] Step: 82300 Acc: 0.70562 0.84478 Cost: 1.11603 0.71307 0.29455 0.10841 Time: 0.00075
17-03-26 09:11:49 [1] Train Extra: lr=0.0000281 inv=0.4300000 sub=0.0000000
17-03-26 09:13:03 [1] Step: 82400 Acc: 0.70406 0.84690 Cost: 1.27428 0.91873 0.24705 0.10850 Time: 0.00071
17-03-26 09:13:03 [1] Train Extra: lr=0.0000280 inv=0.3928125 sub=0.0000000
17-03-26 09:14:25 [1] Step: 82500 Acc: 0.70219 0.84778 Cost: 0.88592 0.63877 0.13855 0.10861 Time: 0.00075
17-03-26 09:14:25 [1] Train Extra: lr=0.0000279 inv=0.4326563 sub=0.0000000
17-03-26 09:15:23 [1] Step: 82500 Eval acc: 0.68463 0.85327 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 09:15:23 [1] Eval Extra: inv=0.3927231
17-03-26 09:16:45 [1] Step: 82600 Acc: 0.68719 0.85108 Cost: 0.97927 0.75975 0.11090 0.10861 Time: 0.00075
17-03-26 09:16:45 [1] Train Extra: lr=0.0000279 inv=0.4214062 sub=0.0000000
17-03-26 09:18:15 [1] Step: 82700 Acc: 0.69156 0.85093 Cost: 1.23625 0.92369 0.20398 0.10857 Time: 0.00079
17-03-26 09:18:15 [1] Train Extra: lr=0.0000278 inv=0.4296875 sub=0.0000000
17-03-26 09:19:34 [1] Step: 82800 Acc: 0.69031 0.84383 Cost: 1.04503 0.74337 0.19304 0.10862 Time: 0.00073
17-03-26 09:19:34 [1] Train Extra: lr=0.0000277 inv=0.4240625 sub=0.0000000
17-03-26 09:20:50 [1] Step: 82900 Acc: 0.69469 0.84422 Cost: 0.87122 0.54986 0.21271 0.10865 Time: 0.00069
17-03-26 09:20:50 [1] Train Extra: lr=0.0000276 inv=0.4428125 sub=0.0000000
17-03-26 09:22:13 [1] Step: 83000 Acc: 0.69875 0.85159 Cost: 0.75505 0.56636 0.08004 0.10864 Time: 0.00079
17-03-26 09:22:13 [1] Train Extra: lr=0.0000276 inv=0.4026562 sub=0.0000000
17-03-26 09:23:11 [1] Step: 83000 Eval acc: 0.68606 0.85419 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 09:23:11 [1] Eval Extra: inv=0.4192800
17-03-26 09:24:25 [1] Step: 83100 Acc: 0.68688 0.84559 Cost: 1.11934 0.78530 0.22537 0.10867 Time: 0.00071
17-03-26 09:24:25 [1] Train Extra: lr=0.0000275 inv=0.4076562 sub=0.0000000
17-03-26 09:25:45 [1] Step: 83200 Acc: 0.71062 0.84138 Cost: 1.21262 0.91973 0.18428 0.10861 Time: 0.00071
17-03-26 09:25:45 [1] Train Extra: lr=0.0000274 inv=0.4468750 sub=0.0000000
17-03-26 09:27:18 [1] Step: 83300 Acc: 0.69875 0.85483 Cost: 0.74705 0.52094 0.11749 0.10862 Time: 0.00081
17-03-26 09:27:18 [1] Train Extra: lr=0.0000273 inv=0.4118750 sub=0.0000000
17-03-26 09:28:56 [1] Step: 83400 Acc: 0.69594 0.85186 Cost: 1.18912 0.77363 0.30686 0.10863 Time: 0.00080
17-03-26 09:28:56 [1] Train Extra: lr=0.0000272 inv=0.4300000 sub=0.0000000
17-03-26 09:30:21 [1] Step: 83500 Acc: 0.69000 0.84910 Cost: 1.14854 0.82052 0.21932 0.10870 Time: 0.00075
17-03-26 09:30:21 [1] Train Extra: lr=0.0000272 inv=0.4310937 sub=0.0000000
17-03-26 09:31:18 [1] Step: 83500 Eval acc: 0.68154 0.84777 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 09:31:18 [1] Eval Extra: inv=0.4320892
17-03-26 09:32:39 [1] Step: 83600 Acc: 0.70813 0.84694 Cost: 1.05502 0.67197 0.27428 0.10878 Time: 0.00074
17-03-26 09:32:39 [1] Train Extra: lr=0.0000271 inv=0.4375000 sub=0.0000000
17-03-26 09:34:00 [1] Step: 83700 Acc: 0.71594 0.85159 Cost: 0.84557 0.59592 0.14080 0.10884 Time: 0.00076
17-03-26 09:34:00 [1] Train Extra: lr=0.0000270 inv=0.3873437 sub=0.0000000
17-03-26 09:35:26 [1] Step: 83800 Acc: 0.68469 0.85953 Cost: 1.01866 0.71020 0.19953 0.10893 Time: 0.00077
17-03-26 09:35:26 [1] Train Extra: lr=0.0000269 inv=0.4007812 sub=0.0000000
17-03-26 09:36:47 [1] Step: 83900 Acc: 0.69656 0.85442 Cost: 0.96274 0.64822 0.20551 0.10902 Time: 0.00076
17-03-26 09:36:47 [1] Train Extra: lr=0.0000268 inv=0.3964063 sub=0.0000000
17-03-26 09:38:14 [1] Step: 84000 Acc: 0.72250 0.84483 Cost: 1.01148 0.62963 0.27274 0.10911 Time: 0.00076
17-03-26 09:38:14 [1] Train Extra: lr=0.0000268 inv=0.4257812 sub=0.0000000
17-03-26 09:39:13 [1] Step: 84000 Eval acc: 0.68242 0.85486 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 09:39:13 [1] Eval Extra: inv=0.3945451
17-03-26 09:40:38 [1] Step: 84100 Acc: 0.72188 0.84735 Cost: 0.99675 0.65744 0.23001 0.10930 Time: 0.00075
17-03-26 09:40:38 [1] Train Extra: lr=0.0000267 inv=0.4201563 sub=0.0000000
17-03-26 09:42:02 [1] Step: 84200 Acc: 0.71281 0.84976 Cost: 0.92896 0.65956 0.15995 0.10945 Time: 0.00076
17-03-26 09:42:02 [1] Train Extra: lr=0.0000266 inv=0.4198438 sub=0.0000000
17-03-26 09:43:30 [1] Step: 84300 Acc: 0.72656 0.85151 Cost: 0.95881 0.64590 0.20342 0.10950 Time: 0.00076
17-03-26 09:43:30 [1] Train Extra: lr=0.0000265 inv=0.4437500 sub=0.0000000
17-03-26 09:44:49 [1] Step: 84400 Acc: 0.72750 0.84924 Cost: 0.65595 0.48007 0.06629 0.10959 Time: 0.00074
17-03-26 09:44:49 [1] Train Extra: lr=0.0000265 inv=0.4050000 sub=0.0000000
17-03-26 09:46:07 [1] Step: 84500 Acc: 0.71719 0.84439 Cost: 1.05719 0.67371 0.27374 0.10974 Time: 0.00072
17-03-26 09:46:07 [1] Train Extra: lr=0.0000264 inv=0.4357813 sub=0.0000000
17-03-26 09:47:03 [1] Step: 84500 Eval acc: 0.68198 0.85510 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 09:47:03 [1] Eval Extra: inv=0.4252981
17-03-26 09:48:29 [1] Step: 84600 Acc: 0.72250 0.85145 Cost: 0.97932 0.65326 0.21623 0.10983 Time: 0.00076
17-03-26 09:48:29 [1] Train Extra: lr=0.0000263 inv=0.4239062 sub=0.0000000
17-03-26 09:49:43 [1] Step: 84700 Acc: 0.71937 0.85074 Cost: 0.76434 0.45183 0.20249 0.11002 Time: 0.00072
17-03-26 09:49:43 [1] Train Extra: lr=0.0000262 inv=0.4153125 sub=0.0000000
17-03-26 09:51:08 [1] Step: 84800 Acc: 0.73062 0.84900 Cost: 0.96308 0.62425 0.22871 0.11012 Time: 0.00078
17-03-26 09:51:08 [1] Train Extra: lr=0.0000262 inv=0.4143750 sub=0.0000000
17-03-26 09:52:31 [1] Step: 84900 Acc: 0.71156 0.84967 Cost: 1.11947 0.73788 0.27141 0.11018 Time: 0.00073
17-03-26 09:52:31 [1] Train Extra: lr=0.0000261 inv=0.4273438 sub=0.0000000
17-03-26 09:53:48 [1] Step: 85000 Acc: 0.71437 0.84899 Cost: 1.09661 0.73773 0.24863 0.11025 Time: 0.00073
17-03-26 09:53:48 [1] Train Extra: lr=0.0000260 inv=0.4510938 sub=0.0000000
17-03-26 09:54:46 [1] Step: 85000 Eval acc: 0.68297 0.85295 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 09:54:46 [1] Eval Extra: inv=0.4112191
17-03-26 09:54:46 [1] Checkpointing.
17-03-26 09:56:05 [1] Step: 85100 Acc: 0.71156 0.84629 Cost: 0.87608 0.65275 0.11299 0.11033 Time: 0.00073
17-03-26 09:56:05 [1] Train Extra: lr=0.0000259 inv=0.4184375 sub=0.0000000
17-03-26 09:57:26 [1] Step: 85200 Acc: 0.71313 0.85635 Cost: 1.22315 0.84657 0.26604 0.11054 Time: 0.00076
17-03-26 09:57:26 [1] Train Extra: lr=0.0000259 inv=0.3837500 sub=0.0000000
17-03-26 09:58:50 [1] Step: 85300 Acc: 0.70969 0.84875 Cost: 1.15367 0.74424 0.29882 0.11061 Time: 0.00076
17-03-26 09:58:50 [1] Train Extra: lr=0.0000258 inv=0.4212500 sub=0.0000000
17-03-26 10:00:09 [1] Step: 85400 Acc: 0.70844 0.84950 Cost: 1.07045 0.81620 0.14354 0.11071 Time: 0.00074
17-03-26 10:00:09 [1] Train Extra: lr=0.0000257 inv=0.4215625 sub=0.0000000
17-03-26 10:01:34 [1] Step: 85500 Acc: 0.72062 0.85790 Cost: 0.74735 0.54153 0.09505 0.11077 Time: 0.00077
17-03-26 10:01:34 [1] Train Extra: lr=0.0000256 inv=0.4292187 sub=0.0000000
17-03-26 10:02:32 [1] Step: 85500 Eval acc: 0.68739 0.84982 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 10:02:32 [1] Eval Extra: inv=0.4147527
17-03-26 10:03:47 [1] Step: 85600 Acc: 0.71750 0.85213 Cost: 0.72843 0.42565 0.19187 0.11091 Time: 0.00071
17-03-26 10:03:47 [1] Train Extra: lr=0.0000256 inv=0.4121875 sub=0.0000000
17-03-26 10:05:11 [1] Step: 85700 Acc: 0.71094 0.84474 Cost: 1.01776 0.70821 0.19855 0.11100 Time: 0.00075
17-03-26 10:05:11 [1] Train Extra: lr=0.0000255 inv=0.4507813 sub=0.0000000
17-03-26 10:06:30 [1] Step: 85800 Acc: 0.71250 0.85531 Cost: 1.08113 0.83918 0.13089 0.11106 Time: 0.00077
17-03-26 10:06:30 [1] Train Extra: lr=0.0000254 inv=0.3979687 sub=0.0000000
17-03-26 10:07:53 [1] Step: 85900 Acc: 0.72875 0.84834 Cost: 1.34012 0.83977 0.38918 0.11118 Time: 0.00074
17-03-26 10:07:53 [1] Train Extra: lr=0.0000253 inv=0.4026562 sub=0.0000000
17-03-26 10:09:17 [1] Step: 86000 Acc: 0.70312 0.84811 Cost: 1.01289 0.67718 0.22443 0.11128 Time: 0.00076
17-03-26 10:09:17 [1] Train Extra: lr=0.0000253 inv=0.4343750 sub=0.0000000
17-03-26 10:10:15 [1] Step: 86000 Eval acc: 0.67745 0.84924 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 10:10:15 [1] Eval Extra: inv=0.4193905
17-03-26 10:11:29 [1] Step: 86100 Acc: 0.72375 0.85302 Cost: 0.93516 0.65525 0.16855 0.11136 Time: 0.00074
17-03-26 10:11:29 [1] Train Extra: lr=0.0000252 inv=0.3935938 sub=0.0000000
17-03-26 10:12:55 [1] Step: 86200 Acc: 0.71937 0.84939 Cost: 0.99926 0.72987 0.15794 0.11145 Time: 0.00077
17-03-26 10:12:55 [1] Train Extra: lr=0.0000251 inv=0.4009375 sub=0.0000000
17-03-26 10:14:20 [1] Step: 86300 Acc: 0.71594 0.85000 Cost: 1.06625 0.71625 0.23845 0.11154 Time: 0.00076
17-03-26 10:14:20 [1] Train Extra: lr=0.0000251 inv=0.4181250 sub=0.0000000
17-03-26 10:15:39 [1] Step: 86400 Acc: 0.71844 0.84865 Cost: 0.96780 0.56571 0.29045 0.11164 Time: 0.00074
17-03-26 10:15:39 [1] Train Extra: lr=0.0000250 inv=0.3987500 sub=0.0000000
17-03-26 10:16:58 [1] Step: 86500 Acc: 0.69437 0.84191 Cost: 0.93708 0.57560 0.24976 0.11172 Time: 0.00073
17-03-26 10:16:58 [1] Train Extra: lr=0.0000249 inv=0.4101562 sub=0.0000000
17-03-26 10:17:56 [1] Step: 86500 Eval acc: 0.68110 0.85205 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 10:17:56 [1] Eval Extra: inv=0.4129859
17-03-26 10:19:23 [1] Step: 86600 Acc: 0.69969 0.84795 Cost: 1.16826 0.89478 0.16175 0.11173 Time: 0.00076
17-03-26 10:19:23 [1] Train Extra: lr=0.0000248 inv=0.4409375 sub=0.0000000
17-03-26 10:20:54 [1] Step: 86700 Acc: 0.71188 0.84870 Cost: 1.05852 0.72647 0.22022 0.11183 Time: 0.00077
17-03-26 10:20:54 [1] Train Extra: lr=0.0000248 inv=0.4348437 sub=0.0000000
17-03-26 10:22:14 [1] Step: 86800 Acc: 0.69500 0.84850 Cost: 0.78936 0.52534 0.15212 0.11191 Time: 0.00072
17-03-26 10:22:14 [1] Train Extra: lr=0.0000247 inv=0.4293750 sub=0.0000000
17-03-26 10:23:38 [1] Step: 86900 Acc: 0.71750 0.85792 Cost: 1.01819 0.67083 0.23536 0.11200 Time: 0.00078
17-03-26 10:23:38 [1] Train Extra: lr=0.0000246 inv=0.4203125 sub=0.0000000
17-03-26 10:25:10 [1] Step: 87000 Acc: 0.70375 0.85507 Cost: 0.79008 0.51571 0.16229 0.11208 Time: 0.00079
17-03-26 10:25:10 [1] Train Extra: lr=0.0000246 inv=0.4056250 sub=0.0000000
17-03-26 10:26:09 [1] Step: 87000 Eval acc: 0.68275 0.85064 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 10:26:09 [1] Eval Extra: inv=0.4057531
17-03-26 10:27:36 [1] Step: 87100 Acc: 0.71313 0.84855 Cost: 0.85306 0.54663 0.19432 0.11212 Time: 0.00076
17-03-26 10:27:36 [1] Train Extra: lr=0.0000245 inv=0.4312500 sub=0.0000000
17-03-26 10:28:59 [1] Step: 87200 Acc: 0.70531 0.85232 Cost: 1.01663 0.65523 0.24921 0.11219 Time: 0.00077
17-03-26 10:28:59 [1] Train Extra: lr=0.0000244 inv=0.3921875 sub=0.0000000
17-03-26 10:30:17 [1] Step: 87300 Acc: 0.70469 0.84534 Cost: 1.11595 0.77171 0.23209 0.11216 Time: 0.00073
17-03-26 10:30:17 [1] Train Extra: lr=0.0000243 inv=0.4106250 sub=0.0000000
17-03-26 10:31:25 [1] Step: 87400 Acc: 0.72375 0.85445 Cost: 0.79195 0.52119 0.15855 0.11221 Time: 0.00068
17-03-26 10:31:25 [1] Train Extra: lr=0.0000243 inv=0.4039063 sub=0.0000000
17-03-26 10:32:43 [1] Step: 87500 Acc: 0.70031 0.84648 Cost: 1.04135 0.76795 0.16114 0.11226 Time: 0.00072
17-03-26 10:32:43 [1] Train Extra: lr=0.0000242 inv=0.4140625 sub=0.0000000
17-03-26 10:33:42 [1] Step: 87500 Eval acc: 0.68209 0.85162 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 10:33:42 [1] Eval Extra: inv=0.4538428
17-03-26 10:35:12 [1] Step: 87600 Acc: 0.70719 0.85138 Cost: 0.90494 0.67831 0.11433 0.11230 Time: 0.00077
17-03-26 10:35:12 [1] Train Extra: lr=0.0000241 inv=0.4543750 sub=0.0000000
17-03-26 10:36:36 [1] Step: 87700 Acc: 0.70375 0.85546 Cost: 0.97089 0.68348 0.17505 0.11236 Time: 0.00077
17-03-26 10:36:36 [1] Train Extra: lr=0.0000241 inv=0.4204688 sub=0.0000000
17-03-26 10:37:50 [1] Step: 87800 Acc: 0.71562 0.84849 Cost: 0.86987 0.64353 0.11390 0.11245 Time: 0.00072
17-03-26 10:37:50 [1] Train Extra: lr=0.0000240 inv=0.4023438 sub=0.0000000
17-03-26 10:39:10 [1] Step: 87900 Acc: 0.70375 0.84529 Cost: 0.93483 0.56150 0.26069 0.11263 Time: 0.00072
17-03-26 10:39:10 [1] Train Extra: lr=0.0000239 inv=0.4353125 sub=0.0000000
17-03-26 10:40:29 [1] Step: 88000 Acc: 0.70219 0.84371 Cost: 0.80507 0.46720 0.22514 0.11273 Time: 0.00073
17-03-26 10:40:29 [1] Train Extra: lr=0.0000239 inv=0.4103125 sub=0.0000000
17-03-26 10:41:26 [1] Step: 88000 Eval acc: 0.68110 0.85150 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 10:41:26 [1] Eval Extra: inv=0.4178445
17-03-26 10:42:44 [1] Step: 88100 Acc: 0.70281 0.86343 Cost: 0.81165 0.54445 0.15438 0.11281 Time: 0.00075
17-03-26 10:42:44 [1] Train Extra: lr=0.0000238 inv=0.4006250 sub=0.0000000
17-03-26 10:44:03 [1] Step: 88200 Acc: 0.70719 0.85278 Cost: 0.97609 0.65393 0.20935 0.11281 Time: 0.00075
17-03-26 10:44:03 [1] Train Extra: lr=0.0000237 inv=0.4160937 sub=0.0000000
17-03-26 10:45:27 [1] Step: 88300 Acc: 0.70594 0.84150 Cost: 0.89622 0.60384 0.17953 0.11285 Time: 0.00074
17-03-26 10:45:27 [1] Train Extra: lr=0.0000237 inv=0.4448437 sub=0.0000000
17-03-26 10:46:52 [1] Step: 88400 Acc: 0.69875 0.84731 Cost: 0.96144 0.63496 0.21367 0.11281 Time: 0.00076
17-03-26 10:46:52 [1] Train Extra: lr=0.0000236 inv=0.4376563 sub=0.0000000
17-03-26 10:48:21 [1] Step: 88500 Acc: 0.69500 0.85429 Cost: 1.17986 0.83921 0.22784 0.11281 Time: 0.00077
17-03-26 10:48:21 [1] Train Extra: lr=0.0000235 inv=0.4206250 sub=0.0000000
17-03-26 10:49:18 [1] Step: 88500 Eval acc: 0.68253 0.85341 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 10:49:18 [1] Eval Extra: inv=0.4179549
17-03-26 10:50:37 [1] Step: 88600 Acc: 0.71344 0.85260 Cost: 1.26296 0.79630 0.35374 0.11292 Time: 0.00075
17-03-26 10:50:37 [1] Train Extra: lr=0.0000235 inv=0.4048438 sub=0.0000000
17-03-26 10:52:04 [1] Step: 88700 Acc: 0.70813 0.84860 Cost: 0.94278 0.62733 0.20247 0.11298 Time: 0.00076
17-03-26 10:52:04 [1] Train Extra: lr=0.0000234 inv=0.4173438 sub=0.0000000
17-03-26 10:53:28 [1] Step: 88800 Acc: 0.68750 0.84872 Cost: 0.95571 0.66947 0.17323 0.11301 Time: 0.00075
17-03-26 10:53:28 [1] Train Extra: lr=0.0000233 inv=0.4129687 sub=0.0000000
17-03-26 10:54:52 [1] Step: 88900 Acc: 0.69812 0.84895 Cost: 1.13129 0.76536 0.25286 0.11307 Time: 0.00076
17-03-26 10:54:52 [1] Train Extra: lr=0.0000232 inv=0.4300000 sub=0.0000000
17-03-26 10:56:07 [1] Step: 89000 Acc: 0.71562 0.85273 Cost: 1.03618 0.64408 0.27903 0.11307 Time: 0.00073
17-03-26 10:56:07 [1] Train Extra: lr=0.0000232 inv=0.4295312 sub=0.0000000
17-03-26 10:57:05 [1] Step: 89000 Eval acc: 0.68187 0.85283 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 10:57:05 [1] Eval Extra: inv=0.4598609
17-03-26 10:58:23 [1] Step: 89100 Acc: 0.69094 0.84584 Cost: 0.95233 0.63337 0.20575 0.11321 Time: 0.00074
17-03-26 10:58:23 [1] Train Extra: lr=0.0000231 inv=0.4309375 sub=0.0000000
17-03-26 10:59:43 [1] Step: 89200 Acc: 0.69188 0.84696 Cost: 0.86288 0.52070 0.22891 0.11327 Time: 0.00073
17-03-26 10:59:43 [1] Train Extra: lr=0.0000230 inv=0.4373437 sub=0.0000000
17-03-26 11:00:56 [1] Step: 89300 Acc: 0.72313 0.85657 Cost: 0.85144 0.61385 0.12420 0.11339 Time: 0.00074
17-03-26 11:00:56 [1] Train Extra: lr=0.0000230 inv=0.3909375 sub=0.0000000
17-03-26 11:02:20 [1] Step: 89400 Acc: 0.68969 0.85109 Cost: 0.98784 0.57172 0.30271 0.11341 Time: 0.00078
17-03-26 11:02:20 [1] Train Extra: lr=0.0000229 inv=0.4206250 sub=0.0000000
17-03-26 11:03:47 [1] Step: 89500 Acc: 0.70813 0.83963 Cost: 0.84661 0.53048 0.20262 0.11351 Time: 0.00074
17-03-26 11:03:47 [1] Train Extra: lr=0.0000229 inv=0.4610938 sub=0.0000000
17-03-26 11:04:43 [1] Step: 89500 Eval acc: 0.68286 0.85521 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 11:04:43 [1] Eval Extra: inv=0.4361197
17-03-26 11:06:03 [1] Step: 89600 Acc: 0.70469 0.84896 Cost: 1.27901 0.92933 0.23612 0.11356 Time: 0.00073
17-03-26 11:06:03 [1] Train Extra: lr=0.0000228 inv=0.4289062 sub=0.0000000
17-03-26 11:07:34 [1] Step: 89700 Acc: 0.69375 0.84161 Cost: 1.32575 0.96257 0.24963 0.11355 Time: 0.00076
17-03-26 11:07:34 [1] Train Extra: lr=0.0000227 inv=0.4754687 sub=0.0000000
17-03-26 11:08:53 [1] Step: 89800 Acc: 0.69469 0.84609 Cost: 1.04761 0.68821 0.24586 0.11354 Time: 0.00072
17-03-26 11:08:53 [1] Train Extra: lr=0.0000227 inv=0.4392187 sub=0.0000000
17-03-26 11:10:22 [1] Step: 89900 Acc: 0.71656 0.85186 Cost: 0.85801 0.48174 0.26276 0.11351 Time: 0.00076
17-03-26 11:10:22 [1] Train Extra: lr=0.0000226 inv=0.4306250 sub=0.0000000
17-03-26 11:11:40 [1] Step: 90000 Acc: 0.71031 0.83923 Cost: 0.99669 0.71075 0.17245 0.11349 Time: 0.00072
17-03-26 11:11:40 [1] Train Extra: lr=0.0000225 inv=0.4454688 sub=0.0000000
17-03-26 11:12:38 [1] Step: 90000 Eval acc: 0.68441 0.84982 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 11:12:38 [1] Eval Extra: inv=0.4675905
17-03-26 11:12:38 [1] Checkpointing.
17-03-26 11:14:02 [1] Step: 90100 Acc: 0.70219 0.84799 Cost: 0.99906 0.71455 0.17097 0.11354 Time: 0.00077
17-03-26 11:14:02 [1] Train Extra: lr=0.0000225 inv=0.4460938 sub=0.0000000
17-03-26 11:15:29 [1] Step: 90200 Acc: 0.69063 0.84759 Cost: 1.04752 0.68657 0.24740 0.11355 Time: 0.00075
17-03-26 11:15:29 [1] Train Extra: lr=0.0000224 inv=0.4570312 sub=0.0000000
17-03-26 11:16:55 [1] Step: 90300 Acc: 0.69875 0.84785 Cost: 1.17156 0.80981 0.24817 0.11358 Time: 0.00075
17-03-26 11:16:55 [1] Train Extra: lr=0.0000223 inv=0.4581250 sub=0.0000000
17-03-26 11:18:23 [1] Step: 90400 Acc: 0.71156 0.84784 Cost: 0.96261 0.59615 0.25275 0.11371 Time: 0.00075
17-03-26 11:18:23 [1] Train Extra: lr=0.0000223 inv=0.4390625 sub=0.0000000
17-03-26 11:19:46 [1] Step: 90500 Acc: 0.70125 0.84906 Cost: 1.08227 0.74768 0.22076 0.11383 Time: 0.00077
17-03-26 11:19:46 [1] Train Extra: lr=0.0000222 inv=0.4243750 sub=0.0000000
17-03-26 11:20:44 [1] Step: 90500 Eval acc: 0.68629 0.85013 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 11:20:44 [1] Eval Extra: inv=0.4301016
17-03-26 11:22:07 [1] Step: 90600 Acc: 0.69844 0.84930 Cost: 1.06466 0.76637 0.18451 0.11378 Time: 0.00074
17-03-26 11:22:07 [1] Train Extra: lr=0.0000221 inv=0.4273438 sub=0.0000000
17-03-26 11:23:26 [1] Step: 90700 Acc: 0.69719 0.85203 Cost: 1.07622 0.77208 0.19035 0.11379 Time: 0.00076
17-03-26 11:23:26 [1] Train Extra: lr=0.0000221 inv=0.4120313 sub=0.0000000
17-03-26 11:24:49 [1] Step: 90800 Acc: 0.69719 0.85461 Cost: 1.00388 0.62657 0.26352 0.11379 Time: 0.00077
17-03-26 11:24:49 [1] Train Extra: lr=0.0000220 inv=0.4250000 sub=0.0000000
17-03-26 11:26:12 [1] Step: 90900 Acc: 0.69125 0.85967 Cost: 0.90685 0.60429 0.18873 0.11383 Time: 0.00076
17-03-26 11:26:12 [1] Train Extra: lr=0.0000219 inv=0.4057812 sub=0.0000000
17-03-26 11:27:40 [1] Step: 91000 Acc: 0.68688 0.84780 Cost: 0.98904 0.61291 0.26228 0.11385 Time: 0.00076
17-03-26 11:27:40 [1] Train Extra: lr=0.0000219 inv=0.4470312 sub=0.0000000
17-03-26 11:28:37 [1] Step: 91000 Eval acc: 0.68198 0.85722 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 11:28:37 [1] Eval Extra: inv=0.3917845
17-03-26 11:30:04 [1] Step: 91100 Acc: 0.70188 0.85626 Cost: 1.03079 0.82786 0.08910 0.11383 Time: 0.00077
17-03-26 11:30:04 [1] Train Extra: lr=0.0000218 inv=0.4334375 sub=0.0000000
17-03-26 11:31:36 [1] Step: 91200 Acc: 0.70250 0.85894 Cost: 1.37868 0.94458 0.32026 0.11384 Time: 0.00079
17-03-26 11:31:36 [1] Train Extra: lr=0.0000218 inv=0.4409375 sub=0.0000000
17-03-26 11:33:03 [1] Step: 91300 Acc: 0.69531 0.84914 Cost: 1.00101 0.60316 0.28393 0.11391 Time: 0.00075
17-03-26 11:33:03 [1] Train Extra: lr=0.0000217 inv=0.4392187 sub=0.0000000
17-03-26 11:34:23 [1] Step: 91400 Acc: 0.69219 0.85332 Cost: 1.08767 0.70155 0.27222 0.11390 Time: 0.00075
17-03-26 11:34:23 [1] Train Extra: lr=0.0000216 inv=0.4103125 sub=0.0000000
17-03-26 11:35:47 [1] Step: 91500 Acc: 0.71437 0.86119 Cost: 1.17474 0.75814 0.30270 0.11389 Time: 0.00081
17-03-26 11:35:47 [1] Train Extra: lr=0.0000216 inv=0.3953125 sub=0.0000000
17-03-26 11:36:42 [1] Step: 91500 Eval acc: 0.68860 0.85237 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-26 11:36:42 [1] Eval Extra: inv=0.4379417
17-03-26 11:38:11 [1] Step: 91600 Acc: 0.70250 0.85265 Cost: 0.96665 0.66462 0.18805 0.11398 Time: 0.00077
17-03-26 11:38:11 [1] Train Extra: lr=0.0000215 inv=0.4192187 sub=0.0000000
17-03-26 11:39:36 [1] Step: 91700 Acc: 0.70719 0.85076 Cost: 0.72470 0.47573 0.13495 0.11402 Time: 0.00075
17-03-26 11:39:36 [1] Train Extra: lr=0.0000215 inv=0.4246875 sub=0.0000000
17-03-26 11:40:49 [1] Step: 91800 Acc: 0.70437 0.84622 Cost: 0.85113 0.53520 0.20195 0.11398 Time: 0.00071
17-03-26 11:40:49 [1] Train Extra: lr=0.0000214 inv=0.3960938 sub=0.0000000
17-03-26 11:42:16 [1] Step: 91900 Acc: 0.69312 0.85347 Cost: 1.15725 0.78156 0.26168 0.11400 Time: 0.00076
17-03-26 11:42:16 [1] Train Extra: lr=0.0000213 inv=0.4220313 sub=0.0000000
17-03-26 11:43:42 [1] Step: 92000 Acc: 0.68906 0.84505 Cost: 1.46213 1.02260 0.32559 0.11394 Time: 0.00074
17-03-26 11:43:42 [1] Train Extra: lr=0.0000213 inv=0.4356250 sub=0.0000000
17-03-26 11:44:40 [1] Step: 92000 Eval acc: 0.68209 0.85657 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 11:44:40 [1] Eval Extra: inv=0.4027164
17-03-26 11:45:53 [1] Step: 92100 Acc: 0.70281 0.85027 Cost: 0.97927 0.60475 0.26053 0.11399 Time: 0.00072
17-03-26 11:45:53 [1] Train Extra: lr=0.0000212 inv=0.4128125 sub=0.0000000
17-03-26 11:47:12 [1] Step: 92200 Acc: 0.70312 0.85108 Cost: 1.19979 0.79706 0.28872 0.11400 Time: 0.00074
17-03-26 11:47:12 [1] Train Extra: lr=0.0000211 inv=0.4192187 sub=0.0000000
17-03-26 11:48:34 [1] Step: 92300 Acc: 0.71000 0.84783 Cost: 0.93438 0.66550 0.15478 0.11411 Time: 0.00076
17-03-26 11:48:34 [1] Train Extra: lr=0.0000211 inv=0.4156250 sub=0.0000000
17-03-26 11:49:54 [1] Step: 92400 Acc: 0.72625 0.84986 Cost: 1.03168 0.68592 0.23153 0.11423 Time: 0.00073
17-03-26 11:49:54 [1] Train Extra: lr=0.0000210 inv=0.4100000 sub=0.0000000
17-03-26 11:51:18 [1] Step: 92500 Acc: 0.74156 0.84447 Cost: 1.23142 0.91199 0.20504 0.11439 Time: 0.00075
17-03-26 11:51:18 [1] Train Extra: lr=0.0000210 inv=0.4364062 sub=0.0000000
17-03-26 11:52:16 [1] Step: 92500 Eval acc: 0.68485 0.85408 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 11:52:16 [1] Eval Extra: inv=0.3852142
17-03-26 11:53:36 [1] Step: 92600 Acc: 0.73031 0.84355 Cost: 0.88982 0.54241 0.23283 0.11459 Time: 0.00072
17-03-26 11:53:36 [1] Train Extra: lr=0.0000209 inv=0.4450000 sub=0.0000000
17-03-26 11:55:00 [1] Step: 92700 Acc: 0.74250 0.84431 Cost: 0.69150 0.48444 0.09237 0.11469 Time: 0.00075
17-03-26 11:55:00 [1] Train Extra: lr=0.0000208 inv=0.4412500 sub=0.0000000
17-03-26 11:56:15 [1] Step: 92800 Acc: 0.72000 0.85143 Cost: 1.13957 0.78834 0.23643 0.11480 Time: 0.00072
17-03-26 11:56:15 [1] Train Extra: lr=0.0000208 inv=0.4193750 sub=0.0000000
17-03-26 11:57:40 [1] Step: 92900 Acc: 0.72062 0.85171 Cost: 0.99603 0.67291 0.20824 0.11487 Time: 0.00078
17-03-26 11:57:40 [1] Train Extra: lr=0.0000207 inv=0.4389062 sub=0.0000000
17-03-26 11:58:59 [1] Step: 93000 Acc: 0.73469 0.84702 Cost: 1.02164 0.66461 0.24208 0.11496 Time: 0.00073
17-03-26 11:58:59 [1] Train Extra: lr=0.0000207 inv=0.4318750 sub=0.0000000
17-03-26 11:59:57 [1] Step: 93000 Eval acc: 0.68286 0.85280 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 11:59:57 [1] Eval Extra: inv=0.4252981
17-03-26 12:01:11 [1] Step: 93100 Acc: 0.71344 0.85395 Cost: 0.92470 0.65001 0.15968 0.11501 Time: 0.00071
17-03-26 12:01:11 [1] Train Extra: lr=0.0000206 inv=0.4067188 sub=0.0000000
17-03-26 12:02:24 [1] Step: 93200 Acc: 0.71375 0.85276 Cost: 0.81497 0.52742 0.17242 0.11512 Time: 0.00072
17-03-26 12:02:24 [1] Train Extra: lr=0.0000205 inv=0.4204688 sub=0.0000000
17-03-26 12:03:45 [1] Step: 93300 Acc: 0.70500 0.85530 Cost: 0.83765 0.57129 0.15115 0.11521 Time: 0.00076
17-03-26 12:03:45 [1] Train Extra: lr=0.0000205 inv=0.4134375 sub=0.0000000
17-03-26 12:05:01 [1] Step: 93400 Acc: 0.73813 0.85044 Cost: 1.12985 0.78541 0.22921 0.11523 Time: 0.00075
17-03-26 12:05:01 [1] Train Extra: lr=0.0000204 inv=0.4071875 sub=0.0000000
17-03-26 12:06:27 [1] Step: 93500 Acc: 0.72250 0.84336 Cost: 1.00302 0.73975 0.14786 0.11541 Time: 0.00075
17-03-26 12:06:27 [1] Train Extra: lr=0.0000204 inv=0.4385938 sub=0.0000000
17-03-26 12:07:25 [1] Step: 93500 Eval acc: 0.68507 0.85749 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 12:07:25 [1] Eval Extra: inv=0.4400950
17-03-26 12:08:50 [1] Step: 93600 Acc: 0.71500 0.84004 Cost: 0.85136 0.51348 0.22238 0.11550 Time: 0.00073
17-03-26 12:08:50 [1] Train Extra: lr=0.0000203 inv=0.4468750 sub=0.0000000
17-03-26 12:10:10 [1] Step: 93700 Acc: 0.72375 0.84540 Cost: 0.91460 0.57332 0.22569 0.11560 Time: 0.00074
17-03-26 12:10:10 [1] Train Extra: lr=0.0000203 inv=0.4039063 sub=0.0000000
17-03-26 12:11:30 [1] Step: 93800 Acc: 0.71750 0.85026 Cost: 0.78550 0.56592 0.10391 0.11567 Time: 0.00074
17-03-26 12:11:30 [1] Train Extra: lr=0.0000202 inv=0.4560938 sub=0.0000000
17-03-26 12:12:45 [1] Step: 93900 Acc: 0.72656 0.84931 Cost: 1.29740 0.92167 0.25994 0.11579 Time: 0.00072
17-03-26 12:12:45 [1] Train Extra: lr=0.0000201 inv=0.4231250 sub=0.0000000
17-03-26 12:14:09 [1] Step: 94000 Acc: 0.70625 0.84994 Cost: 1.08568 0.70355 0.26624 0.11589 Time: 0.00076
17-03-26 12:14:09 [1] Train Extra: lr=0.0000201 inv=0.4171875 sub=0.0000000
17-03-26 12:15:07 [1] Step: 94000 Eval acc: 0.68021 0.85103 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 12:15:07 [1] Eval Extra: inv=0.4323101
17-03-26 12:16:32 [1] Step: 94100 Acc: 0.71375 0.84620 Cost: 1.04526 0.78976 0.13955 0.11596 Time: 0.00076
17-03-26 12:16:32 [1] Train Extra: lr=0.0000200 inv=0.4242187 sub=0.0000000
17-03-26 12:17:49 [1] Step: 94200 Acc: 0.71781 0.85468 Cost: 1.01897 0.69613 0.20684 0.11600 Time: 0.00073
17-03-26 12:17:49 [1] Train Extra: lr=0.0000200 inv=0.3976562 sub=0.0000000
17-03-26 12:19:22 [1] Step: 94300 Acc: 0.71125 0.85240 Cost: 0.90087 0.54127 0.24353 0.11606 Time: 0.00079
17-03-26 12:19:22 [1] Train Extra: lr=0.0000199 inv=0.4446875 sub=0.0000000
17-03-26 12:20:41 [1] Step: 94400 Acc: 0.72937 0.84851 Cost: 1.02956 0.74147 0.17191 0.11618 Time: 0.00075
17-03-26 12:20:41 [1] Train Extra: lr=0.0000198 inv=0.4037500 sub=0.0000000
17-03-26 12:22:00 [1] Step: 94500 Acc: 0.72469 0.84673 Cost: 0.80228 0.56033 0.12571 0.11624 Time: 0.00072
17-03-26 12:22:00 [1] Train Extra: lr=0.0000198 inv=0.4267187 sub=0.0000000
17-03-26 12:22:57 [1] Step: 94500 Eval acc: 0.68706 0.85134 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 12:22:57 [1] Eval Extra: inv=0.4103357
17-03-26 12:24:21 [1] Step: 94600 Acc: 0.71031 0.84923 Cost: 1.06681 0.87455 0.07595 0.11631 Time: 0.00075
17-03-26 12:24:21 [1] Train Extra: lr=0.0000197 inv=0.4040625 sub=0.0000000
17-03-26 12:25:49 [1] Step: 94700 Acc: 0.71313 0.84818 Cost: 1.31267 0.90627 0.28999 0.11640 Time: 0.00076
17-03-26 12:25:49 [1] Train Extra: lr=0.0000197 inv=0.4248438 sub=0.0000000
17-03-26 12:27:12 [1] Step: 94800 Acc: 0.72437 0.84984 Cost: 0.92969 0.72062 0.09254 0.11653 Time: 0.00078
17-03-26 12:27:12 [1] Train Extra: lr=0.0000196 inv=0.4007812 sub=0.0000000
17-03-26 12:28:26 [1] Step: 94900 Acc: 0.72594 0.85609 Cost: 1.19047 0.79627 0.27759 0.11661 Time: 0.00074
17-03-26 12:28:26 [1] Train Extra: lr=0.0000196 inv=0.4050000 sub=0.0000000
17-03-26 12:29:51 [1] Step: 95000 Acc: 0.71219 0.84823 Cost: 1.22412 0.84653 0.26086 0.11673 Time: 0.00076
17-03-26 12:29:51 [1] Train Extra: lr=0.0000195 inv=0.4265625 sub=0.0000000
17-03-26 12:30:46 [1] Step: 95000 Eval acc: 0.68463 0.85549 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-26 12:30:46 [1] Eval Extra: inv=0.4149183
17-03-26 12:30:46 [1] Checkpointing.
17-03-26 12:32:04 [1] Step: 95100 Acc: 0.72375 0.84297 Cost: 0.93966 0.64123 0.18169 0.11674 Time: 0.00073
17-03-26 12:32:04 [1] Train Extra: lr=0.0000195 inv=0.4104687 sub=0.0000000
17-03-26 12:33:30 [1] Step: 95200 Acc: 0.71625 0.84076 Cost: 1.22881 0.77848 0.33355 0.11678 Time: 0.00074
17-03-26 12:33:30 [1] Train Extra: lr=0.0000194 inv=0.4359375 sub=0.0000000
17-03-26 12:34:49 [1] Step: 95300 Acc: 0.69750 0.85096 Cost: 0.88927 0.60305 0.16939 0.11683 Time: 0.00073
17-03-26 12:34:49 [1] Train Extra: lr=0.0000193 inv=0.4095313 sub=0.0000000
17-03-26 12:36:15 [1] Step: 95400 Acc: 0.71906 0.85359 Cost: 0.97702 0.57449 0.28560 0.11693 Time: 0.00077
17-03-26 12:36:15 [1] Train Extra: lr=0.0000193 inv=0.4268750 sub=0.0000000
17-03-26 12:37:41 [1] Step: 95500 Acc: 0.71688 0.85067 Cost: 0.87182 0.55570 0.19908 0.11704 Time: 0.00076
17-03-26 12:37:41 [1] Train Extra: lr=0.0000192 inv=0.4279688 sub=0.0000000
17-03-26 12:38:36 [1] Step: 95500 Eval acc: 0.68330 0.85593 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-26 12:38:36 [1] Eval Extra: inv=0.4071886
17-03-26 12:39:55 [1] Step: 95600 Acc: 0.71375 0.85258 Cost: 0.95546 0.61835 0.22000 0.11711 Time: 0.00075
17-03-26 12:39:55 [1] Train Extra: lr=0.0000192 inv=0.4096875 sub=0.0000000
17-03-26 12:41:22 [1] Step: 95700 Acc: 0.69656 0.85872 Cost: 0.83163 0.56147 0.15296 0.11721 Time: 0.00077
17-03-26 12:41:22 [1] Train Extra: lr=0.0000191 inv=0.4265625 sub=0.0000000
17-03-26 12:42:54 [1] Step: 95800 Acc: 0.69688 0.85766 Cost: 0.97562 0.59703 0.26135 0.11723 Time: 0.00080
17-03-26 12:42:54 [1] Train Extra: lr=0.0000191 inv=0.4220313 sub=0.0000000
17-03-26 12:44:21 [1] Step: 95900 Acc: 0.70688 0.84774 Cost: 1.10305 0.72083 0.26492 0.11730 Time: 0.00075
17-03-26 12:44:21 [1] Train Extra: lr=0.0000190 inv=0.4285937 sub=0.0000000
17-03-26 12:45:40 [1] Step: 96000 Acc: 0.72281 0.84658 Cost: 1.15840 0.77771 0.26331 0.11738 Time: 0.00074
17-03-26 12:45:40 [1] Train Extra: lr=0.0000190 inv=0.4165625 sub=0.0000000
17-03-26 12:46:36 [1] Step: 96000 Eval acc: 0.68905 0.85613 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 12:46:36 [1] Eval Extra: inv=0.3995693
17-03-26 12:48:00 [1] Step: 96100 Acc: 0.71625 0.84736 Cost: 0.81497 0.56846 0.12912 0.11739 Time: 0.00075
17-03-26 12:48:00 [1] Train Extra: lr=0.0000189 inv=0.4295312 sub=0.0000000
17-03-26 12:49:15 [1] Step: 96200 Acc: 0.71125 0.85208 Cost: 0.78915 0.57706 0.09462 0.11748 Time: 0.00073
17-03-26 12:49:15 [1] Train Extra: lr=0.0000188 inv=0.4165625 sub=0.0000000
17-03-26 12:50:40 [1] Step: 96300 Acc: 0.72500 0.85024 Cost: 0.89598 0.57158 0.20685 0.11754 Time: 0.00078
17-03-26 12:50:40 [1] Train Extra: lr=0.0000188 inv=0.4035937 sub=0.0000000
17-03-26 12:51:55 [1] Step: 96400 Acc: 0.72875 0.84659 Cost: 0.88487 0.55310 0.21412 0.11766 Time: 0.00070
17-03-26 12:51:55 [1] Train Extra: lr=0.0000187 inv=0.4206250 sub=0.0000000
17-03-26 12:53:15 [1] Step: 96500 Acc: 0.72094 0.84576 Cost: 1.30216 0.89582 0.28860 0.11774 Time: 0.00076
17-03-26 12:53:15 [1] Train Extra: lr=0.0000187 inv=0.4295312 sub=0.0000000
17-03-26 12:54:13 [1] Step: 96500 Eval acc: 0.68364 0.85459 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 12:54:13 [1] Eval Extra: inv=0.4384386
17-03-26 12:55:40 [1] Step: 96600 Acc: 0.70656 0.84647 Cost: 0.71879 0.52372 0.07733 0.11774 Time: 0.00074
17-03-26 12:55:40 [1] Train Extra: lr=0.0000186 inv=0.4551562 sub=0.0000000
17-03-26 12:56:58 [1] Step: 96700 Acc: 0.71250 0.85508 Cost: 0.79023 0.60899 0.06344 0.11780 Time: 0.00075
17-03-26 12:56:58 [1] Train Extra: lr=0.0000186 inv=0.4092188 sub=0.0000000
17-03-26 12:58:15 [1] Step: 96800 Acc: 0.71875 0.84915 Cost: 0.78830 0.55107 0.11947 0.11776 Time: 0.00074
17-03-26 12:58:15 [1] Train Extra: lr=0.0000185 inv=0.4120313 sub=0.0000000
17-03-26 12:59:40 [1] Step: 96900 Acc: 0.71906 0.85061 Cost: 0.88066 0.59664 0.16615 0.11787 Time: 0.00076
17-03-26 12:59:40 [1] Train Extra: lr=0.0000185 inv=0.4148438 sub=0.0000000
17-03-26 13:01:02 [1] Step: 97000 Acc: 0.70875 0.84513 Cost: 1.08884 0.73858 0.23237 0.11789 Time: 0.00073
17-03-26 13:01:02 [1] Train Extra: lr=0.0000184 inv=0.4148438 sub=0.0000000
17-03-26 13:02:00 [1] Step: 97000 Eval acc: 0.68617 0.85289 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 13:02:00 [1] Eval Extra: inv=0.4142005
17-03-26 13:03:28 [1] Step: 97100 Acc: 0.71031 0.85416 Cost: 0.96061 0.72979 0.11286 0.11796 Time: 0.00078
17-03-26 13:03:28 [1] Train Extra: lr=0.0000184 inv=0.4117188 sub=0.0000000
17-03-26 13:04:46 [1] Step: 97200 Acc: 0.71844 0.84953 Cost: 1.34607 1.00172 0.22635 0.11800 Time: 0.00074
17-03-26 13:04:46 [1] Train Extra: lr=0.0000183 inv=0.3985938 sub=0.0000000
17-03-26 13:06:17 [1] Step: 97300 Acc: 0.70500 0.85032 Cost: 0.82727 0.60789 0.10129 0.11809 Time: 0.00078
17-03-26 13:06:17 [1] Train Extra: lr=0.0000183 inv=0.4300000 sub=0.0000000
17-03-26 13:07:36 [1] Step: 97400 Acc: 0.70312 0.85069 Cost: 1.22045 0.83679 0.26558 0.11808 Time: 0.00073
17-03-26 13:07:36 [1] Train Extra: lr=0.0000182 inv=0.4212500 sub=0.0000000
17-03-26 13:09:11 [1] Step: 97500 Acc: 0.70844 0.85677 Cost: 1.04129 0.67458 0.24855 0.11815 Time: 0.00082
17-03-26 13:09:11 [1] Train Extra: lr=0.0000182 inv=0.4068750 sub=0.0000000
17-03-26 13:10:06 [1] Step: 97500 Eval acc: 0.68275 0.85262 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00016
17-03-26 13:10:06 [1] Eval Extra: inv=0.4255190
17-03-26 13:11:29 [1] Step: 97600 Acc: 0.69750 0.84974 Cost: 1.03621 0.65518 0.26286 0.11817 Time: 0.00076
17-03-26 13:11:29 [1] Train Extra: lr=0.0000181 inv=0.4279688 sub=0.0000000
17-03-26 13:12:47 [1] Step: 97700 Acc: 0.70875 0.85126 Cost: 1.19963 0.82406 0.25733 0.11824 Time: 0.00074
17-03-26 13:12:47 [1] Train Extra: lr=0.0000180 inv=0.4246875 sub=0.0000000
17-03-26 13:14:07 [1] Step: 97800 Acc: 0.71156 0.85220 Cost: 1.34708 0.86978 0.35901 0.11829 Time: 0.00076
17-03-26 13:14:07 [1] Train Extra: lr=0.0000180 inv=0.4200000 sub=0.0000000
17-03-26 13:15:34 [1] Step: 97900 Acc: 0.70813 0.84592 Cost: 1.18399 0.79530 0.27037 0.11832 Time: 0.00074
17-03-26 13:15:34 [1] Train Extra: lr=0.0000179 inv=0.4459375 sub=0.0000000
17-03-26 13:16:53 [1] Step: 98000 Acc: 0.71656 0.84989 Cost: 0.87784 0.49676 0.26266 0.11841 Time: 0.00074
17-03-26 13:16:53 [1] Train Extra: lr=0.0000179 inv=0.4309375 sub=0.0000000
17-03-26 13:17:50 [1] Step: 98000 Eval acc: 0.68617 0.85582 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 13:17:50 [1] Eval Extra: inv=0.4336904
17-03-26 13:19:11 [1] Step: 98100 Acc: 0.71437 0.85026 Cost: 0.83518 0.52378 0.19298 0.11842 Time: 0.00075
17-03-26 13:19:11 [1] Train Extra: lr=0.0000178 inv=0.4210937 sub=0.0000000
17-03-26 13:20:43 [1] Step: 98200 Acc: 0.69969 0.84563 Cost: 1.40089 0.97715 0.30525 0.11849 Time: 0.00077
17-03-26 13:20:43 [1] Train Extra: lr=0.0000178 inv=0.4379688 sub=0.0000000
17-03-26 13:22:15 [1] Step: 98300 Acc: 0.69437 0.85356 Cost: 0.87453 0.61600 0.13997 0.11857 Time: 0.00081
17-03-26 13:22:15 [1] Train Extra: lr=0.0000177 inv=0.4062500 sub=0.0000000
17-03-26 13:23:37 [1] Step: 98400 Acc: 0.71656 0.85077 Cost: 0.79037 0.52054 0.15131 0.11852 Time: 0.00073
17-03-26 13:23:37 [1] Train Extra: lr=0.0000177 inv=0.4442187 sub=0.0000000
17-03-26 13:24:56 [1] Step: 98500 Acc: 0.70656 0.85426 Cost: 0.92499 0.70310 0.10333 0.11856 Time: 0.00075
17-03-26 13:24:56 [1] Train Extra: lr=0.0000176 inv=0.4104687 sub=0.0000000
17-03-26 13:25:55 [1] Step: 98500 Eval acc: 0.68949 0.84893 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 13:25:55 [1] Eval Extra: inv=0.4145870
17-03-26 13:27:22 [1] Step: 98600 Acc: 0.70500 0.84469 Cost: 1.39102 1.00061 0.27178 0.11863 Time: 0.00075
17-03-26 13:27:22 [1] Train Extra: lr=0.0000176 inv=0.4285937 sub=0.0000000
17-03-26 13:28:45 [1] Step: 98700 Acc: 0.71406 0.85010 Cost: 0.79330 0.52680 0.14788 0.11862 Time: 0.00076
17-03-26 13:28:45 [1] Train Extra: lr=0.0000175 inv=0.3978125 sub=0.0000000
17-03-26 13:30:18 [1] Step: 98800 Acc: 0.70312 0.84379 Cost: 0.89354 0.66733 0.10761 0.11860 Time: 0.00075
17-03-26 13:30:18 [1] Train Extra: lr=0.0000175 inv=0.4518750 sub=0.0000000
17-03-26 13:31:45 [1] Step: 98900 Acc: 0.71469 0.85164 Cost: 1.12957 0.81107 0.19987 0.11864 Time: 0.00074
17-03-26 13:31:45 [1] Train Extra: lr=0.0000174 inv=0.4584375 sub=0.0000000
17-03-26 13:33:04 [1] Step: 99000 Acc: 0.70781 0.84970 Cost: 0.79467 0.56829 0.10771 0.11867 Time: 0.00075
17-03-26 13:33:04 [1] Train Extra: lr=0.0000174 inv=0.4006250 sub=0.0000000
17-03-26 13:34:02 [1] Step: 99000 Eval acc: 0.68231 0.84676 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 13:34:02 [1] Eval Extra: inv=0.4112743
17-03-26 13:35:16 [1] Step: 99100 Acc: 0.69594 0.85430 Cost: 0.92651 0.66442 0.14339 0.11870 Time: 0.00073
17-03-26 13:35:16 [1] Train Extra: lr=0.0000173 inv=0.3962500 sub=0.0000000
17-03-26 13:36:28 [1] Step: 99200 Acc: 0.69281 0.84907 Cost: 1.20328 0.78565 0.29890 0.11873 Time: 0.00073
17-03-26 13:36:28 [1] Train Extra: lr=0.0000173 inv=0.4110937 sub=0.0000000
17-03-26 13:37:54 [1] Step: 99300 Acc: 0.69719 0.84325 Cost: 0.95566 0.66471 0.17220 0.11875 Time: 0.00073
17-03-26 13:37:54 [1] Train Extra: lr=0.0000172 inv=0.4525000 sub=0.0000000
17-03-26 13:39:21 [1] Step: 99400 Acc: 0.71188 0.84363 Cost: 1.19452 0.86897 0.20679 0.11875 Time: 0.00075
17-03-26 13:39:21 [1] Train Extra: lr=0.0000172 inv=0.4350000 sub=0.0000000
17-03-26 13:40:44 [1] Step: 99500 Acc: 0.72094 0.85757 Cost: 1.05065 0.74023 0.19168 0.11873 Time: 0.00078
17-03-26 13:40:44 [1] Train Extra: lr=0.0000171 inv=0.4245313 sub=0.0000000
17-03-26 13:41:41 [1] Step: 99500 Eval acc: 0.68242 0.85191 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 13:41:41 [1] Eval Extra: inv=0.4355124
17-03-26 13:43:12 [1] Step: 99600 Acc: 0.71125 0.84876 Cost: 1.50719 1.06284 0.32560 0.11875 Time: 0.00077
17-03-26 13:43:12 [1] Train Extra: lr=0.0000171 inv=0.4432813 sub=0.0000000
17-03-26 13:44:45 [1] Step: 99700 Acc: 0.69969 0.85808 Cost: 0.54435 0.35236 0.07313 0.11886 Time: 0.00079
17-03-26 13:44:45 [1] Train Extra: lr=0.0000170 inv=0.4084375 sub=0.0000000
17-03-26 13:46:06 [1] Step: 99800 Acc: 0.70344 0.85345 Cost: 0.98802 0.66099 0.20818 0.11885 Time: 0.00074
17-03-26 13:46:06 [1] Train Extra: lr=0.0000170 inv=0.4051562 sub=0.0000000
17-03-26 13:47:26 [1] Step: 99900 Acc: 0.71562 0.85743 Cost: 1.02052 0.66699 0.23463 0.11890 Time: 0.00076
17-03-26 13:47:26 [1] Train Extra: lr=0.0000169 inv=0.4121875 sub=0.0000000
17-03-26 13:48:38 [1] Step: 100000 Acc: 0.72281 0.85445 Cost: 1.24126 0.86152 0.26085 0.11889 Time: 0.00075
17-03-26 13:48:38 [1] Train Extra: lr=0.0000169 inv=0.3893750 sub=0.0000000
17-03-26 13:49:35 [1] Step: 100000 Eval acc: 0.68209 0.85454 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 13:49:35 [1] Eval Extra: inv=0.4079064
17-03-26 13:49:35 [1] Checkpointing.
17-03-26 13:51:00 [1] Step: 100100 Acc: 0.69781 0.84767 Cost: 1.17913 0.71065 0.34954 0.11895 Time: 0.00075
17-03-26 13:51:00 [1] Train Extra: lr=0.0000168 inv=0.4501562 sub=0.0000000
17-03-26 13:52:37 [1] Step: 100200 Acc: 0.69500 0.84927 Cost: 0.87828 0.54796 0.21125 0.11908 Time: 0.00079
17-03-26 13:52:37 [1] Train Extra: lr=0.0000168 inv=0.4476562 sub=0.0000000
17-03-26 13:53:52 [1] Step: 100300 Acc: 0.70375 0.85137 Cost: 0.81044 0.48257 0.20877 0.11910 Time: 0.00072
17-03-26 13:53:52 [1] Train Extra: lr=0.0000167 inv=0.4142188 sub=0.0000000
17-03-26 13:55:17 [1] Step: 100400 Acc: 0.69906 0.85521 Cost: 1.44001 1.02463 0.29624 0.11914 Time: 0.00077
17-03-26 13:55:17 [1] Train Extra: lr=0.0000167 inv=0.4160937 sub=0.0000000
17-03-26 13:56:42 [1] Step: 100500 Acc: 0.70156 0.84750 Cost: 1.13423 0.79157 0.22356 0.11910 Time: 0.00075
17-03-26 13:56:42 [1] Train Extra: lr=0.0000167 inv=0.4498437 sub=0.0000000
17-03-26 13:57:41 [1] Step: 100500 Eval acc: 0.68485 0.85090 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 13:57:41 [1] Eval Extra: inv=0.4295495
17-03-26 13:59:09 [1] Step: 100600 Acc: 0.70844 0.85197 Cost: 1.53558 1.11135 0.30509 0.11914 Time: 0.00079
17-03-26 13:59:09 [1] Train Extra: lr=0.0000166 inv=0.4203125 sub=0.0000000
17-03-26 14:00:39 [1] Step: 100700 Acc: 0.70156 0.85691 Cost: 1.04694 0.68333 0.24447 0.11914 Time: 0.00081
17-03-26 14:00:39 [1] Train Extra: lr=0.0000166 inv=0.4356250 sub=0.0000000
17-03-26 14:02:11 [1] Step: 100800 Acc: 0.72250 0.85341 Cost: 0.96689 0.58363 0.26398 0.11928 Time: 0.00078
17-03-26 14:02:11 [1] Train Extra: lr=0.0000165 inv=0.4489063 sub=0.0000000
17-03-26 14:03:25 [1] Step: 100900 Acc: 0.74031 0.84973 Cost: 0.87622 0.57558 0.18120 0.11944 Time: 0.00071
17-03-26 14:03:25 [1] Train Extra: lr=0.0000165 inv=0.3834375 sub=0.0000000
17-03-26 14:04:43 [1] Step: 101000 Acc: 0.72313 0.84845 Cost: 1.01136 0.64326 0.24853 0.11957 Time: 0.00073
17-03-26 14:04:43 [1] Train Extra: lr=0.0000164 inv=0.4093750 sub=0.0000000
17-03-26 14:05:40 [1] Step: 101000 Eval acc: 0.67977 0.85126 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 14:05:40 [1] Eval Extra: inv=0.4268441
17-03-26 14:06:56 [1] Step: 101100 Acc: 0.73594 0.84828 Cost: 0.97834 0.59921 0.25952 0.11961 Time: 0.00073
17-03-26 14:06:56 [1] Train Extra: lr=0.0000164 inv=0.3946875 sub=0.0000000
17-03-26 14:08:12 [1] Step: 101200 Acc: 0.73156 0.85236 Cost: 1.08921 0.80287 0.16660 0.11974 Time: 0.00071
17-03-26 14:08:12 [1] Train Extra: lr=0.0000163 inv=0.4237500 sub=0.0000000
17-03-26 14:09:46 [1] Step: 101300 Acc: 0.72000 0.84845 Cost: 0.78216 0.53093 0.13137 0.11987 Time: 0.00080
17-03-26 14:09:46 [1] Train Extra: lr=0.0000163 inv=0.4481250 sub=0.0000000
17-03-26 14:11:17 [1] Step: 101400 Acc: 0.72969 0.84264 Cost: 0.81478 0.55087 0.14392 0.12000 Time: 0.00076
17-03-26 14:11:17 [1] Train Extra: lr=0.0000162 inv=0.4535938 sub=0.0000000
17-03-26 14:12:38 [1] Step: 101500 Acc: 0.74531 0.84356 Cost: 0.87407 0.59390 0.16007 0.12010 Time: 0.00071
17-03-26 14:12:38 [1] Train Extra: lr=0.0000162 inv=0.4337500 sub=0.0000000
17-03-26 14:13:35 [1] Step: 101500 Eval acc: 0.68154 0.85383 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 14:13:35 [1] Eval Extra: inv=0.4236418
17-03-26 14:14:58 [1] Step: 101600 Acc: 0.73281 0.84237 Cost: 1.13734 0.72889 0.28820 0.12025 Time: 0.00075
17-03-26 14:14:58 [1] Train Extra: lr=0.0000161 inv=0.4415625 sub=0.0000000
17-03-26 14:16:23 [1] Step: 101700 Acc: 0.71688 0.85677 Cost: 1.19865 0.86135 0.21692 0.12038 Time: 0.00076
17-03-26 14:16:23 [1] Train Extra: lr=0.0000161 inv=0.4218750 sub=0.0000000
17-03-26 14:17:41 [1] Step: 101800 Acc: 0.72781 0.84622 Cost: 1.17239 0.78962 0.26237 0.12040 Time: 0.00072
17-03-26 14:17:41 [1] Train Extra: lr=0.0000160 inv=0.4248438 sub=0.0000000
17-03-26 14:19:09 [1] Step: 101900 Acc: 0.71906 0.84839 Cost: 1.06793 0.72782 0.21962 0.12049 Time: 0.00075
17-03-26 14:19:09 [1] Train Extra: lr=0.0000160 inv=0.4360938 sub=0.0000000
17-03-26 14:20:26 [1] Step: 102000 Acc: 0.73687 0.84763 Cost: 0.72622 0.47770 0.12792 0.12060 Time: 0.00073
17-03-26 14:20:26 [1] Train Extra: lr=0.0000159 inv=0.4135937 sub=0.0000000
17-03-26 14:21:24 [1] Step: 102000 Eval acc: 0.68165 0.85301 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 14:21:24 [1] Eval Extra: inv=0.4011705
17-03-26 14:22:41 [1] Step: 102100 Acc: 0.73406 0.84892 Cost: 0.84969 0.51918 0.20972 0.12079 Time: 0.00074
17-03-26 14:22:41 [1] Train Extra: lr=0.0000159 inv=0.3925000 sub=0.0000000
17-03-26 14:24:06 [1] Step: 102200 Acc: 0.71875 0.85397 Cost: 0.97085 0.57595 0.27399 0.12091 Time: 0.00076
17-03-26 14:24:06 [1] Train Extra: lr=0.0000159 inv=0.4112500 sub=0.0000000
17-03-26 14:25:26 [1] Step: 102300 Acc: 0.72344 0.84083 Cost: 0.88522 0.64614 0.11812 0.12096 Time: 0.00072
17-03-26 14:25:26 [1] Train Extra: lr=0.0000158 inv=0.4153125 sub=0.0000000
17-03-26 14:26:47 [1] Step: 102400 Acc: 0.72875 0.84996 Cost: 1.16144 0.71981 0.32060 0.12103 Time: 0.00075
17-03-26 14:26:47 [1] Train Extra: lr=0.0000158 inv=0.3935938 sub=0.0000000
17-03-26 14:28:06 [1] Step: 102500 Acc: 0.72719 0.85343 Cost: 0.78750 0.59364 0.07265 0.12120 Time: 0.00076
17-03-26 14:28:06 [1] Train Extra: lr=0.0000157 inv=0.4020313 sub=0.0000000
17-03-26 14:29:03 [1] Step: 102500 Eval acc: 0.68065 0.84982 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 14:29:03 [1] Eval Extra: inv=0.3843308
17-03-26 14:30:31 [1] Step: 102600 Acc: 0.72406 0.85327 Cost: 0.93532 0.57407 0.23993 0.12133 Time: 0.00077
17-03-26 14:30:31 [1] Train Extra: lr=0.0000157 inv=0.4376563 sub=0.0000000
17-03-26 14:32:02 [1] Step: 102700 Acc: 0.71469 0.85278 Cost: 0.86136 0.56937 0.17056 0.12142 Time: 0.00079
17-03-26 14:32:02 [1] Train Extra: lr=0.0000156 inv=0.4334375 sub=0.0000000
17-03-26 14:33:16 [1] Step: 102800 Acc: 0.71688 0.85552 Cost: 1.05448 0.69816 0.23479 0.12153 Time: 0.00073
17-03-26 14:33:16 [1] Train Extra: lr=0.0000156 inv=0.4123438 sub=0.0000000
17-03-26 14:34:36 [1] Step: 102900 Acc: 0.72875 0.84456 Cost: 0.97327 0.58896 0.26276 0.12154 Time: 0.00075
17-03-26 14:34:36 [1] Train Extra: lr=0.0000155 inv=0.4304688 sub=0.0000000
17-03-26 14:35:57 [1] Step: 103000 Acc: 0.72594 0.85616 Cost: 1.03788 0.68567 0.23055 0.12166 Time: 0.00075
17-03-26 14:35:57 [1] Train Extra: lr=0.0000155 inv=0.4285937 sub=0.0000000
17-03-26 14:36:54 [1] Step: 103000 Eval acc: 0.68805 0.85878 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 14:36:54 [1] Eval Extra: inv=0.3923918
17-03-26 14:38:12 [1] Step: 103100 Acc: 0.73156 0.85509 Cost: 0.71920 0.42585 0.17162 0.12173 Time: 0.00075
17-03-26 14:38:12 [1] Train Extra: lr=0.0000155 inv=0.3892188 sub=0.0000000
17-03-26 14:39:31 [1] Step: 103200 Acc: 0.72281 0.84849 Cost: 1.15814 0.76688 0.26947 0.12180 Time: 0.00073
17-03-26 14:39:31 [1] Train Extra: lr=0.0000154 inv=0.4210937 sub=0.0000000
17-03-26 14:41:03 [1] Step: 103300 Acc: 0.72469 0.85160 Cost: 1.19356 0.79340 0.27823 0.12193 Time: 0.00078
17-03-26 14:41:03 [1] Train Extra: lr=0.0000154 inv=0.4421875 sub=0.0000000
17-03-26 14:42:28 [1] Step: 103400 Acc: 0.71562 0.84853 Cost: 0.83777 0.61267 0.10312 0.12199 Time: 0.00077
17-03-26 14:42:28 [1] Train Extra: lr=0.0000153 inv=0.4354688 sub=0.0000000
17-03-26 14:43:43 [1] Step: 103500 Acc: 0.71750 0.84586 Cost: 1.02493 0.63067 0.27220 0.12207 Time: 0.00072
17-03-26 14:43:43 [1] Train Extra: lr=0.0000153 inv=0.4179688 sub=0.0000000
17-03-26 14:44:42 [1] Step: 103500 Eval acc: 0.67833 0.85566 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 14:44:42 [1] Eval Extra: inv=0.4358989
17-03-26 14:46:00 [1] Step: 103600 Acc: 0.72437 0.85087 Cost: 1.02778 0.66547 0.24013 0.12218 Time: 0.00075
17-03-26 14:46:00 [1] Train Extra: lr=0.0000152 inv=0.4093750 sub=0.0000000
17-03-26 14:47:21 [1] Step: 103700 Acc: 0.72875 0.85498 Cost: 1.12699 0.77974 0.22506 0.12219 Time: 0.00075
17-03-26 14:47:21 [1] Train Extra: lr=0.0000152 inv=0.4110937 sub=0.0000000
17-03-26 14:48:46 [1] Step: 103800 Acc: 0.72094 0.84675 Cost: 1.10252 0.68604 0.29426 0.12222 Time: 0.00075
17-03-26 14:48:46 [1] Train Extra: lr=0.0000151 inv=0.4425000 sub=0.0000000
17-03-26 14:50:16 [1] Step: 103900 Acc: 0.72375 0.84798 Cost: 0.94052 0.61038 0.20782 0.12232 Time: 0.00077
17-03-26 14:50:16 [1] Train Extra: lr=0.0000151 inv=0.4456250 sub=0.0000000
17-03-26 14:51:58 [1] Step: 104000 Acc: 0.69437 0.85186 Cost: 1.07132 0.74641 0.20253 0.12239 Time: 0.00078
17-03-26 14:51:58 [1] Train Extra: lr=0.0000151 inv=0.4700000 sub=0.0000000
17-03-26 14:52:55 [1] Step: 104000 Eval acc: 0.67193 0.85361 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 14:52:55 [1] Eval Extra: inv=0.4190592
17-03-26 14:54:26 [1] Step: 104100 Acc: 0.71437 0.85142 Cost: 0.85921 0.51676 0.22000 0.12245 Time: 0.00078
17-03-26 14:54:26 [1] Train Extra: lr=0.0000150 inv=0.4479688 sub=0.0000000
17-03-26 14:55:38 [1] Step: 104200 Acc: 0.72188 0.85067 Cost: 1.44495 1.07018 0.25225 0.12253 Time: 0.00073
17-03-26 14:55:38 [1] Train Extra: lr=0.0000150 inv=0.3931250 sub=0.0000000
17-03-26 14:57:04 [1] Step: 104300 Acc: 0.72781 0.84831 Cost: 0.90318 0.63061 0.14998 0.12259 Time: 0.00076
17-03-26 14:57:04 [1] Train Extra: lr=0.0000149 inv=0.4129687 sub=0.0000000
17-03-26 14:58:25 [1] Step: 104400 Acc: 0.70094 0.85273 Cost: 0.85297 0.49397 0.23639 0.12261 Time: 0.00073
17-03-26 14:58:25 [1] Train Extra: lr=0.0000149 inv=0.4385938 sub=0.0000000
17-03-26 14:59:44 [1] Step: 104500 Acc: 0.72344 0.84350 Cost: 0.90269 0.54300 0.23695 0.12274 Time: 0.00072
17-03-26 14:59:44 [1] Train Extra: lr=0.0000148 inv=0.4293750 sub=0.0000000
17-03-26 15:00:42 [1] Step: 104500 Eval acc: 0.68330 0.85298 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 15:00:42 [1] Eval Extra: inv=0.4132619
17-03-26 15:02:00 [1] Step: 104600 Acc: 0.71313 0.85602 Cost: 0.95590 0.61126 0.22189 0.12275 Time: 0.00074
17-03-26 15:02:00 [1] Train Extra: lr=0.0000148 inv=0.4068750 sub=0.0000000
17-03-26 15:03:20 [1] Step: 104700 Acc: 0.72281 0.85176 Cost: 0.98094 0.64883 0.20930 0.12281 Time: 0.00074
17-03-26 15:03:20 [1] Train Extra: lr=0.0000148 inv=0.4342187 sub=0.0000000
17-03-26 15:04:52 [1] Step: 104800 Acc: 0.71781 0.85020 Cost: 0.95032 0.63403 0.19344 0.12285 Time: 0.00078
17-03-26 15:04:52 [1] Train Extra: lr=0.0000147 inv=0.4503125 sub=0.0000000
17-03-26 15:06:14 [1] Step: 104900 Acc: 0.72500 0.84940 Cost: 1.21699 0.80366 0.29040 0.12293 Time: 0.00076
17-03-26 15:06:14 [1] Train Extra: lr=0.0000147 inv=0.4056250 sub=0.0000000
17-03-26 15:07:38 [1] Step: 105000 Acc: 0.72062 0.85560 Cost: 0.78330 0.46826 0.19204 0.12300 Time: 0.00079
17-03-26 15:07:38 [1] Train Extra: lr=0.0000146 inv=0.4001562 sub=0.0000000
17-03-26 15:08:36 [1] Step: 105000 Eval acc: 0.67966 0.85369 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 15:08:36 [1] Eval Extra: inv=0.4615172
17-03-26 15:08:36 [1] Checkpointing.
17-03-26 15:09:53 [1] Step: 105100 Acc: 0.71844 0.85931 Cost: 0.99343 0.69239 0.17795 0.12310 Time: 0.00076
17-03-26 15:09:53 [1] Train Extra: lr=0.0000146 inv=0.3810938 sub=0.0000000
17-03-26 15:11:19 [1] Step: 105200 Acc: 0.72281 0.85527 Cost: 0.76698 0.59015 0.05367 0.12316 Time: 0.00080
17-03-26 15:11:19 [1] Train Extra: lr=0.0000145 inv=0.4339062 sub=0.0000000
17-03-26 15:12:52 [1] Step: 105300 Acc: 0.70625 0.84740 Cost: 1.00366 0.63474 0.24576 0.12316 Time: 0.00077
17-03-26 15:12:52 [1] Train Extra: lr=0.0000145 inv=0.4732812 sub=0.0000000
17-03-26 15:14:09 [1] Step: 105400 Acc: 0.72000 0.84821 Cost: 0.86764 0.57961 0.16485 0.12319 Time: 0.00073
17-03-26 15:14:09 [1] Train Extra: lr=0.0000145 inv=0.4142188 sub=0.0000000
17-03-26 15:15:28 [1] Step: 105500 Acc: 0.72875 0.86110 Cost: 0.76096 0.42376 0.21394 0.12325 Time: 0.00077
17-03-26 15:15:28 [1] Train Extra: lr=0.0000144 inv=0.3926562 sub=0.0000000
17-03-26 15:16:26 [1] Step: 105500 Eval acc: 0.68595 0.85233 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 15:16:26 [1] Eval Extra: inv=0.4392668
17-03-26 15:17:50 [1] Step: 105600 Acc: 0.71375 0.85656 Cost: 0.96206 0.68442 0.15435 0.12329 Time: 0.00077
17-03-26 15:17:50 [1] Train Extra: lr=0.0000144 inv=0.4293750 sub=0.0000000
17-03-26 15:19:04 [1] Step: 105700 Acc: 0.71656 0.84949 Cost: 0.99337 0.71725 0.15277 0.12336 Time: 0.00071
17-03-26 15:19:04 [1] Train Extra: lr=0.0000143 inv=0.4150000 sub=0.0000000
17-03-26 15:20:35 [1] Step: 105800 Acc: 0.71625 0.85258 Cost: 0.87341 0.58310 0.16687 0.12345 Time: 0.00078
17-03-26 15:20:35 [1] Train Extra: lr=0.0000143 inv=0.4403125 sub=0.0000000
17-03-26 15:21:47 [1] Step: 105900 Acc: 0.72406 0.85105 Cost: 0.70589 0.51685 0.06553 0.12351 Time: 0.00070
17-03-26 15:21:47 [1] Train Extra: lr=0.0000143 inv=0.4300000 sub=0.0000000
17-03-26 15:23:05 [1] Step: 106000 Acc: 0.71844 0.84774 Cost: 0.98946 0.69307 0.17279 0.12360 Time: 0.00073
17-03-26 15:23:05 [1] Train Extra: lr=0.0000142 inv=0.4451562 sub=0.0000000
17-03-26 15:24:02 [1] Step: 106000 Eval acc: 0.68098 0.84960 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 15:24:02 [1] Eval Extra: inv=0.4574867
17-03-26 15:25:26 [1] Step: 106100 Acc: 0.72000 0.84646 Cost: 0.80700 0.50585 0.17744 0.12371 Time: 0.00074
17-03-26 15:25:26 [1] Train Extra: lr=0.0000142 inv=0.4615625 sub=0.0000000
17-03-26 15:26:41 [1] Step: 106200 Acc: 0.73125 0.84555 Cost: 0.93357 0.62740 0.18241 0.12376 Time: 0.00071
17-03-26 15:26:41 [1] Train Extra: lr=0.0000141 inv=0.4329688 sub=0.0000000
17-03-26 15:28:09 [1] Step: 106300 Acc: 0.70188 0.84664 Cost: 0.93873 0.65051 0.16445 0.12376 Time: 0.00076
17-03-26 15:28:09 [1] Train Extra: lr=0.0000141 inv=0.4348437 sub=0.0000000
17-03-26 15:29:28 [1] Step: 106400 Acc: 0.71562 0.85535 Cost: 1.11536 0.90722 0.08438 0.12376 Time: 0.00077
17-03-26 15:29:28 [1] Train Extra: lr=0.0000141 inv=0.4118750 sub=0.0000000
17-03-26 15:30:54 [1] Step: 106500 Acc: 0.70594 0.85129 Cost: 1.03167 0.68727 0.22061 0.12379 Time: 0.00076
17-03-26 15:30:54 [1] Train Extra: lr=0.0000140 inv=0.4484375 sub=0.0000000
17-03-26 15:31:52 [1] Step: 106500 Eval acc: 0.68408 0.85461 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 15:31:52 [1] Eval Extra: inv=0.4449536
17-03-26 15:33:11 [1] Step: 106600 Acc: 0.71813 0.85346 Cost: 1.19405 0.81103 0.25918 0.12384 Time: 0.00074
17-03-26 15:33:11 [1] Train Extra: lr=0.0000140 inv=0.4048438 sub=0.0000000
17-03-26 15:34:41 [1] Step: 106700 Acc: 0.71062 0.84983 Cost: 0.89370 0.59739 0.17241 0.12391 Time: 0.00078
17-03-26 15:34:41 [1] Train Extra: lr=0.0000139 inv=0.4284375 sub=0.0000000
17-03-26 15:36:14 [1] Step: 106800 Acc: 0.71188 0.84747 Cost: 0.95349 0.61960 0.21003 0.12386 Time: 0.00075
17-03-26 15:36:14 [1] Train Extra: lr=0.0000139 inv=0.4625000 sub=0.0000000
17-03-26 15:37:30 [1] Step: 106900 Acc: 0.72656 0.84939 Cost: 0.83988 0.58614 0.12984 0.12390 Time: 0.00074
17-03-26 15:37:30 [1] Train Extra: lr=0.0000139 inv=0.3942188 sub=0.0000000
17-03-26 15:38:50 [1] Step: 107000 Acc: 0.72219 0.84257 Cost: 1.18046 0.82347 0.23310 0.12389 Time: 0.00073
17-03-26 15:38:50 [1] Train Extra: lr=0.0000138 inv=0.4415625 sub=0.0000000
17-03-26 15:39:48 [1] Step: 107000 Eval acc: 0.68463 0.85448 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 15:39:48 [1] Eval Extra: inv=0.3815702
17-03-26 15:41:08 [1] Step: 107100 Acc: 0.70125 0.84734 Cost: 1.06248 0.69156 0.24693 0.12399 Time: 0.00075
17-03-26 15:41:08 [1] Train Extra: lr=0.0000138 inv=0.4271875 sub=0.0000000
17-03-26 15:42:27 [1] Step: 107200 Acc: 0.71062 0.85636 Cost: 0.93558 0.69230 0.11928 0.12399 Time: 0.00076
17-03-26 15:42:27 [1] Train Extra: lr=0.0000137 inv=0.4250000 sub=0.0000000
17-03-26 15:43:55 [1] Step: 107300 Acc: 0.71313 0.85013 Cost: 1.12180 0.68107 0.31674 0.12399 Time: 0.00076
17-03-26 15:43:55 [1] Train Extra: lr=0.0000137 inv=0.4471875 sub=0.0000000
17-03-26 15:45:20 [1] Step: 107400 Acc: 0.70813 0.84869 Cost: 1.09992 0.69228 0.28356 0.12408 Time: 0.00078
17-03-26 15:45:20 [1] Train Extra: lr=0.0000137 inv=0.4054687 sub=0.0000000
17-03-26 15:46:43 [1] Step: 107500 Acc: 0.72688 0.85381 Cost: 1.09113 0.80495 0.16212 0.12405 Time: 0.00076
17-03-26 15:46:43 [1] Train Extra: lr=0.0000136 inv=0.4160937 sub=0.0000000
17-03-26 15:47:42 [1] Step: 107500 Eval acc: 0.68352 0.85626 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 15:47:42 [1] Eval Extra: inv=0.4132067
17-03-26 15:49:04 [1] Step: 107600 Acc: 0.71031 0.84781 Cost: 0.92293 0.65760 0.14129 0.12404 Time: 0.00074
17-03-26 15:49:04 [1] Train Extra: lr=0.0000136 inv=0.4535938 sub=0.0000000
17-03-26 15:50:20 [1] Step: 107700 Acc: 0.71500 0.84688 Cost: 0.87240 0.51314 0.23520 0.12406 Time: 0.00074
17-03-26 15:50:20 [1] Train Extra: lr=0.0000135 inv=0.4135937 sub=0.0000000
17-03-26 15:51:46 [1] Step: 107800 Acc: 0.72313 0.85348 Cost: 1.14554 0.80518 0.21623 0.12412 Time: 0.00075
17-03-26 15:51:46 [1] Train Extra: lr=0.0000135 inv=0.4293750 sub=0.0000000
17-03-26 15:53:00 [1] Step: 107900 Acc: 0.71531 0.85414 Cost: 1.43633 0.96546 0.34677 0.12410 Time: 0.00072
17-03-26 15:53:00 [1] Train Extra: lr=0.0000135 inv=0.3984375 sub=0.0000000
17-03-26 15:54:25 [1] Step: 108000 Acc: 0.71656 0.84146 Cost: 1.04252 0.64123 0.27713 0.12416 Time: 0.00074
17-03-26 15:54:25 [1] Train Extra: lr=0.0000134 inv=0.4526562 sub=0.0000000
17-03-26 15:55:22 [1] Step: 108000 Eval acc: 0.68341 0.85791 /home/dexter/data/multinli_0.1/multinli_0.1_dev_matched.jsonl Time: 0.00017
17-03-26 15:55:22 [1] Eval Extra: inv=0.3849934
17-03-26 15:56:41 [1] Step: 108100 Acc: 0.71219 0.85214 Cost: 0.80909 0.48194 0.20293 0.12422 Time: 0.00074
17-03-26 15:56:41 [1] Train Extra: lr=0.0000134 inv=0.3892188 sub=0.0000000
17-03-26 15:58:02 [1] Step: 108200 Acc: 0.71500 0.85868 Cost: 1.18406 0.84517 0.21463 0.12426 Time: 0.00075
17-03-26 15:58:02 [1] Train Extra: lr=0.0000133 inv=0.4014063 sub=0.0000000