Last active
October 19, 2016 12:51
-
-
Save mrdrozdov/ec63f356f077cbd790e5c464c2e77c84 to your computer and use it in GitHub Desktop.
SPINN Chainer Logs
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| $ python -m spinn.models.fat_classifier \ | |
| --batch_size 32 \ | |
| --ckpt_path rnn_scratch_6 \ | |
| --connect_tracking_comp \ | |
| --data_type snli \ | |
| --embedding_data_path ../glove/glove.840B.300d.txt \ | |
| --embedding_keep_rate 0.9 \ | |
| --eval_data_path ../snli_1.0/snli_1.0_dev.jsonl \ | |
| --eval_seq_length 25 \ | |
| --experiment_name model0_scratch_6 \ | |
| --log_path ../logs_chainer \ | |
| --model_dim 300 \ | |
| --model_type SPINN \ | |
| --seq_length 25 \ | |
| --training_data_path ../snli_1.0/snli_1.0_train.jsonl \ | |
| --word_embedding_dim 300 \ | |
| --eval_interval_steps 1500 | |
| [1] Flag values: | |
| { '?': None, | |
| 'allow_gt_transitions_in_eval': False, | |
| 'batch_size': 32, | |
| 'ckpt_interval_steps': 5000, | |
| 'ckpt_on_best_dev_error': True, | |
| 'ckpt_path': 'rnn_scratch_6', | |
| 'classifier_type': 'MLP', | |
| 'clipping_max_value': 5.0, | |
| 'connect_tracking_comp': True, | |
| 'context_sensitive_shift': False, | |
| 'context_sensitive_use_relu': False, | |
| 'data_type': 'snli', | |
| 'embedding_data_path': '../glove/glove.840B.300d.txt', | |
| 'embedding_keep_rate': 0.9, | |
| 'eval_data_limit': -1, | |
| 'eval_data_path': '../snli_1.0/snli_1.0_dev.jsonl', | |
| 'eval_interval_steps': 1500, | |
| 'eval_output_paths': None, | |
| 'eval_seq_length': 25, | |
| 'expanded_eval_only_mode': False, | |
| 'experiment_name': 'model0_scratch_6', | |
| 'gpu': -1, | |
| 'help': None, | |
| 'helpshort': None, | |
| 'helpxml': None, | |
| 'init_range': 0.005, | |
| 'initialize_hyp_tracking_state': False, | |
| 'l2_lambda': 1e-05, | |
| 'learning_rate': 0.001, | |
| 'learning_rate_decay_per_10k_steps': 0.75, | |
| 'log_path': '../logs_chainer', | |
| 'lstm_composition': True, | |
| 'model_dim': 300, | |
| 'model_type': 'SPINN', | |
| 'num_sentence_pair_combination_layers': 2, | |
| 'predict_use_cell': False, | |
| 'resnet_unit_depth': 2, | |
| 'scheduled_sampling_exponent_base': 0.99, | |
| 'semantic_classifier_keep_rate': 0.5, | |
| 'sentence_pair_combination_layer_dim': 1024, | |
| 'seq_length': 25, | |
| 'skip_saved_unsavables': False, | |
| 'statistics_interval_steps': 100, | |
| 'tracking_lstm_hidden_dim': 4, | |
| 'training_data_path': '../snli_1.0/snli_1.0_train.jsonl', | |
| 'training_steps': 500000, | |
| 'transition_cost_scale': 1.0, | |
| 'use_difference_feature': True, | |
| 'use_gru': False, | |
| 'use_product_feature': True, | |
| 'use_tracking_lstm': True, | |
| 'word_embedding_dim': 300, | |
| 'write_predicted_label': False} | |
| Loading ../snli_1.0/snli_1.0_train.jsonl | |
| Loading ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] In open vocabulary mode. Using loaded embeddings without fine-tuning. | |
| [1] Constructing vocabulary... | |
| [1] Found 36675 word types. | |
| [1] Loading vocabulary with 32589 words from ../glove/glove.840B.300d.txt | |
| [1] Preprocessing training data. | |
| [1] Preprocessing eval data: ../snli_1.0/snli_1.0_dev.jsonl | |
| WARNING: May be discarding eval examples. | |
| Skipping 18 examples. | |
| [1] Building model. | |
| Init: /x2h_hypothesis/reduce/treelstm/W_l/W:(750, 150) | |
| Init: /x2h_hypothesis/reduce/treelstm/W_l/b:(750,) | |
| Init: /x2h_hypothesis/reduce/treelstm/W_r/W:(750, 150) | |
| Init: /x2h_hypothesis/reduce/treelstm/W_r/b:(750,) | |
| Init: /h2y/0/W:(1024, 600) | |
| Init: /h2y/0/b:(1024,) | |
| Init: /h2y/1/W:(512, 1024) | |
| Init: /h2y/1/b:(512,) | |
| Init: /h2y/2/W:(3, 512) | |
| Init: /h2y/2/b:(3,) | |
| Init: /projection/W:(300, 300) | |
| Init: /projection/b:(300,) | |
| Init: /x2h_premise/reduce/treelstm/W_l/W:(750, 150) | |
| Init: /x2h_premise/reduce/treelstm/W_l/b:(750,) | |
| Init: /x2h_premise/reduce/treelstm/W_r/W:(750, 150) | |
| Init: /x2h_premise/reduce/treelstm/W_r/b:(750,) | |
| [1] Training. | |
| Training [ ] 1% 2s / 290s | |
| [1] Step: 0 Acc: 0.250000 0.000000 Cost: 1.877214 1.877214 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 100 Acc: 0.500000 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 200 Acc: 0.281250 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 300 Acc: 0.343750 0.000000 Cost: 1.092432 1.092432 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 400 Acc: 0.281250 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 500 Acc: 0.281250 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 600 Acc: 0.343750 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 700 Acc: 0.531250 0.000000 Cost: 1.094383 1.094383 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 800 Acc: 0.250000 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 900 Acc: 0.375000 0.000000 Cost: 1.116331 1.116331 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 1000 Acc: 0.375000 0.000000 Cost: 1.107622 1.107622 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 1100 Acc: 0.531250 0.000000 Cost: 1.097006 1.097006 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 1200 Acc: 0.281250 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 1300 Acc: 0.406250 0.000000 Cost: 1.086858 1.086858 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 1400 Acc: 0.375000 0.000000 Cost: 1.077801 1.077801 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 1500 Acc: 0.437500 0.000000 Cost: 1.106823 1.106823 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 214s / 214s | |
| [1] Step: 1500 Eval acc: 0.355252 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.355252 | |
| Training [================================================================================] 100% 305s / 305s | |
| [1] Step: 1600 Acc: 0.343750 0.000000 Cost: 1.085510 1.085510 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 1700 Acc: 0.375000 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 1800 Acc: 0.375000 0.000000 Cost: 1.081344 1.081344 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 1900 Acc: 0.187500 0.000000 Cost: 1.126509 1.126509 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 2000 Acc: 0.375000 0.000000 Cost: 1.099937 1.099937 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 2100 Acc: 0.218750 0.000000 Cost: 1.087157 1.087157 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 2200 Acc: 0.218750 0.000000 Cost: 1.114965 1.114965 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 2300 Acc: 0.406250 0.000000 Cost: 1.075331 1.075331 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 2400 Acc: 0.156250 0.000000 Cost: 1.111402 1.111402 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 2500 Acc: 0.343750 0.000000 Cost: 1.111017 1.111017 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 2600 Acc: 0.437500 0.000000 Cost: 1.078955 1.078955 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 2700 Acc: 0.375000 0.000000 Cost: 1.085429 1.085429 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 2800 Acc: 0.531250 0.000000 Cost: 1.122796 1.122796 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 2900 Acc: 0.375000 0.000000 Cost: 1.146852 1.146852 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 3000 Acc: 0.312500 0.000000 Cost: 1.078576 1.078576 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 214s / 214s | |
| [1] Step: 3000 Eval acc: 0.379988 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.379988 | |
| Training [================================================================================] 100% 305s / 305s | |
| [1] Step: 3100 Acc: 0.468750 0.000000 Cost: 1.073898 1.073898 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 3200 Acc: 0.218750 0.000000 Cost: 1.082206 1.082206 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 3300 Acc: 0.468750 0.000000 Cost: 1.063477 1.063477 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 3400 Acc: 0.343750 0.000000 Cost: 1.099557 1.099557 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 3500 Acc: 0.437500 0.000000 Cost: 1.099835 1.099835 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 3600 Acc: 0.593750 0.000000 Cost: 1.078814 1.078814 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 3700 Acc: 0.437500 0.000000 Cost: 1.119319 1.119319 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 3800 Acc: 0.312500 0.000000 Cost: 1.119129 1.119129 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 3900 Acc: 0.406250 0.000000 Cost: 1.069315 1.069315 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 4000 Acc: 0.312500 0.000000 Cost: 1.098226 1.098226 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 4100 Acc: 0.250000 0.000000 Cost: 1.096237 1.096237 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 4200 Acc: 0.312500 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 4300 Acc: 0.406250 0.000000 Cost: 1.106863 1.106863 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 4400 Acc: 0.218750 0.000000 Cost: 1.085950 1.085950 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 4500 Acc: 0.406250 0.000000 Cost: 1.098612 1.098612 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 216s / 216s | |
| [1] Step: 4500 Eval acc: 0.372252 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| Training [================================================================================] 100% 307s / 307s | |
| [1] Step: 4600 Acc: 0.500000 0.000000 Cost: 1.084996 1.084996 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 4700 Acc: 0.531250 0.000000 Cost: 1.102973 1.102973 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 4800 Acc: 0.531250 0.000000 Cost: 1.077647 1.077647 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 4900 Acc: 0.375000 0.000000 Cost: 1.076459 1.076459 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 5000 Acc: 0.218750 0.000000 Cost: 1.111341 1.111341 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 5100 Acc: 0.375000 0.000000 Cost: 1.095690 1.095690 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 5200 Acc: 0.281250 0.000000 Cost: 1.081936 1.081936 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 5300 Acc: 0.281250 0.000000 Cost: 1.111412 1.111412 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 5400 Acc: 0.437500 0.000000 Cost: 1.083025 1.083025 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 5500 Acc: 0.375000 0.000000 Cost: 1.100099 1.100099 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 5600 Acc: 0.218750 0.000000 Cost: 1.141485 1.141485 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 5700 Acc: 0.187500 0.000000 Cost: 1.095224 1.095224 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 5800 Acc: 0.312500 0.000000 Cost: 1.134330 1.134330 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 5900 Acc: 0.375000 0.000000 Cost: 1.098454 1.098454 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 6000 Acc: 0.250000 0.000000 Cost: 1.100114 1.100114 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 217s / 217s | |
| [1] Step: 6000 Eval acc: 0.378054 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| Training [================================================================================] 100% 308s / 308s | |
| [1] Step: 6100 Acc: 0.312500 0.000000 Cost: 1.093293 1.093293 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 6200 Acc: 0.187500 0.000000 Cost: 1.100203 1.100203 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 6300 Acc: 0.500000 0.000000 Cost: 1.123378 1.123378 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 6400 Acc: 0.343750 0.000000 Cost: 1.093215 1.093215 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 6500 Acc: 0.406250 0.000000 Cost: 1.065132 1.065132 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 6600 Acc: 0.250000 0.000000 Cost: 1.082592 1.082592 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 6700 Acc: 0.406250 0.000000 Cost: 1.082443 1.082443 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 6800 Acc: 0.406250 0.000000 Cost: 1.126872 1.126872 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 6900 Acc: 0.406250 0.000000 Cost: 1.076969 1.076969 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 7000 Acc: 0.437500 0.000000 Cost: 1.078647 1.078647 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 7100 Acc: 0.468750 0.000000 Cost: 1.097078 1.097078 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 7200 Acc: 0.312500 0.000000 Cost: 1.095784 1.095784 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 7300 Acc: 0.406250 0.000000 Cost: 1.082036 1.082036 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 7400 Acc: 0.437500 0.000000 Cost: 1.095886 1.095886 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 7500 Acc: 0.437500 0.000000 Cost: 1.089329 1.089329 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 216s / 216s | |
| [1] Step: 7500 Eval acc: 0.389963 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.389963 | |
| Training [================================================================================] 100% 306s / 306s | |
| [1] Step: 7600 Acc: 0.312500 0.000000 Cost: 1.137161 1.137161 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 7700 Acc: 0.312500 0.000000 Cost: 1.111806 1.111806 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 7800 Acc: 0.406250 0.000000 Cost: 1.106893 1.106893 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 7900 Acc: 0.468750 0.000000 Cost: 1.146132 1.146132 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 8000 Acc: 0.375000 0.000000 Cost: 1.098773 1.098773 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 8100 Acc: 0.531250 0.000000 Cost: 1.071780 1.071780 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 8200 Acc: 0.468750 0.000000 Cost: 1.088614 1.088614 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 8300 Acc: 0.343750 0.000000 Cost: 1.091694 1.091694 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 8400 Acc: 0.312500 0.000000 Cost: 1.089461 1.089461 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 8500 Acc: 0.406250 0.000000 Cost: 1.158216 1.158216 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 8600 Acc: 0.281250 0.000000 Cost: 1.092488 1.092488 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 8700 Acc: 0.281250 0.000000 Cost: 1.091191 1.091191 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 8800 Acc: 0.406250 0.000000 Cost: 1.115413 1.115413 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 8900 Acc: 0.281250 0.000000 Cost: 1.089024 1.089024 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 9000 Acc: 0.406250 0.000000 Cost: 1.067368 1.067368 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 216s / 216s | |
| [1] Step: 9000 Eval acc: 0.398921 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.398921 | |
| Training [================================================================================] 100% 306s / 306s | |
| [1] Step: 9100 Acc: 0.375000 0.000000 Cost: 1.100011 1.100011 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 9200 Acc: 0.468750 0.000000 Cost: 1.090692 1.090692 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 9300 Acc: 0.250000 0.000000 Cost: 1.133367 1.133367 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 9400 Acc: 0.343750 0.000000 Cost: 1.103071 1.103071 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 9500 Acc: 0.406250 0.000000 Cost: 1.098316 1.098316 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 9600 Acc: 0.375000 0.000000 Cost: 1.082073 1.082073 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 9700 Acc: 0.531250 0.000000 Cost: 1.090313 1.090313 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 9800 Acc: 0.343750 0.000000 Cost: 1.108983 1.108983 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 9900 Acc: 0.500000 0.000000 Cost: 1.087912 1.087912 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 10000 Acc: 0.375000 0.000000 Cost: 1.114786 1.114786 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 10100 Acc: 0.375000 0.000000 Cost: 1.116527 1.116527 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 10200 Acc: 0.500000 0.000000 Cost: 1.058278 1.058278 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 10300 Acc: 0.437500 0.000000 Cost: 1.059135 1.059135 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 10400 Acc: 0.281250 0.000000 Cost: 1.138080 1.138080 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 10500 Acc: 0.468750 0.000000 Cost: 1.091681 1.091681 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 214s / 214s | |
| [1] Step: 10500 Eval acc: 0.408795 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.408795 | |
| Training [================================================================================] 100% 306s / 306s | |
| [1] Step: 10600 Acc: 0.281250 0.000000 Cost: 1.115492 1.115492 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 10700 Acc: 0.468750 0.000000 Cost: 1.061459 1.061459 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 10800 Acc: 0.312500 0.000000 Cost: 1.120672 1.120672 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 10900 Acc: 0.406250 0.000000 Cost: 1.087863 1.087863 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 11000 Acc: 0.406250 0.000000 Cost: 1.071164 1.071164 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 11100 Acc: 0.500000 0.000000 Cost: 1.055906 1.055906 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 11200 Acc: 0.406250 0.000000 Cost: 1.054652 1.054652 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 11300 Acc: 0.375000 0.000000 Cost: 1.050360 1.050360 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 11400 Acc: 0.468750 0.000000 Cost: 1.125689 1.125689 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 11500 Acc: 0.281250 0.000000 Cost: 1.134853 1.134853 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 11600 Acc: 0.406250 0.000000 Cost: 1.078301 1.078301 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 11700 Acc: 0.468750 0.000000 Cost: 1.062546 1.062546 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 11800 Acc: 0.500000 0.000000 Cost: 1.082999 1.082999 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 11900 Acc: 0.468750 0.000000 Cost: 1.060507 1.060507 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 12000 Acc: 0.375000 0.000000 Cost: 1.092496 1.092496 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 215s / 215s | |
| [1] Step: 12000 Eval acc: 0.401059 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| Training [================================================================================] 100% 306s / 306s | |
| [1] Step: 12100 Acc: 0.312500 0.000000 Cost: 1.094483 1.094483 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 12200 Acc: 0.343750 0.000000 Cost: 1.041826 1.041826 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 12300 Acc: 0.406250 0.000000 Cost: 1.068268 1.068268 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 12400 Acc: 0.500000 0.000000 Cost: 1.064596 1.064596 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 12500 Acc: 0.281250 0.000000 Cost: 1.098655 1.098655 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 12600 Acc: 0.468750 0.000000 Cost: 1.038236 1.038236 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 12700 Acc: 0.468750 0.000000 Cost: 1.054318 1.054318 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 12800 Acc: 0.375000 0.000000 Cost: 1.083427 1.083427 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 12900 Acc: 0.437500 0.000000 Cost: 1.060724 1.060724 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 13000 Acc: 0.312500 0.000000 Cost: 1.084953 1.084953 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 13100 Acc: 0.406250 0.000000 Cost: 1.072052 1.072052 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 13200 Acc: 0.218750 0.000000 Cost: 1.081473 1.081473 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 13300 Acc: 0.312500 0.000000 Cost: 1.101915 1.101915 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 13400 Acc: 0.375000 0.000000 Cost: 1.060277 1.060277 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 13500 Acc: 0.375000 0.000000 Cost: 1.108615 1.108615 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 216s / 216s | |
| [1] Step: 13500 Eval acc: 0.411543 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| Training [================================================================================] 100% 307s / 307s | |
| [1] Step: 13600 Acc: 0.531250 0.000000 Cost: 1.050899 1.050899 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 13700 Acc: 0.437500 0.000000 Cost: 1.019719 1.019719 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 13800 Acc: 0.343750 0.000000 Cost: 1.073750 1.073750 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 13900 Acc: 0.406250 0.000000 Cost: 1.090007 1.090007 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 14000 Acc: 0.437500 0.000000 Cost: 1.108856 1.108856 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 14100 Acc: 0.406250 0.000000 Cost: 1.110420 1.110420 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 14200 Acc: 0.375000 0.000000 Cost: 1.070153 1.070153 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 14300 Acc: 0.437500 0.000000 Cost: 1.076140 1.076140 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 14400 Acc: 0.343750 0.000000 Cost: 1.143887 1.143887 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 14500 Acc: 0.437500 0.000000 Cost: 1.045461 1.045461 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 14600 Acc: 0.312500 0.000000 Cost: 1.074042 1.074042 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 14700 Acc: 0.406250 0.000000 Cost: 1.069158 1.069158 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 14800 Acc: 0.468750 0.000000 Cost: 1.077461 1.077461 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 14900 Acc: 0.437500 0.000000 Cost: 1.067259 1.067259 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 15000 Acc: 0.375000 0.000000 Cost: 1.094726 1.094726 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 216s / 216s | |
| [1] Step: 15000 Eval acc: 0.415106 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.415106 | |
| Training [================================================================================] 100% 305s / 305s | |
| [1] Step: 15100 Acc: 0.500000 0.000000 Cost: 1.030034 1.030034 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 15200 Acc: 0.406250 0.000000 Cost: 1.082673 1.082673 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 15300 Acc: 0.218750 0.000000 Cost: 1.163438 1.163438 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 15400 Acc: 0.468750 0.000000 Cost: 1.115485 1.115485 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 15500 Acc: 0.406250 0.000000 Cost: 1.103464 1.103464 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 15600 Acc: 0.281250 0.000000 Cost: 1.064159 1.064159 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 15700 Acc: 0.437500 0.000000 Cost: 1.115651 1.115651 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 15800 Acc: 0.500000 0.000000 Cost: 1.065050 1.065050 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 15900 Acc: 0.375000 0.000000 Cost: 1.072550 1.072550 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 16000 Acc: 0.375000 0.000000 Cost: 1.090301 1.090301 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 16100 Acc: 0.343750 0.000000 Cost: 1.105175 1.105175 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 16200 Acc: 0.406250 0.000000 Cost: 1.021194 1.021194 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 16300 Acc: 0.312500 0.000000 Cost: 1.087172 1.087172 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 16400 Acc: 0.437500 0.000000 Cost: 1.117678 1.117678 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 16500 Acc: 0.468750 0.000000 Cost: 1.065603 1.065603 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 215s / 215s | |
| [1] Step: 16500 Eval acc: 0.429662 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.429662 | |
| Training [================================================================================] 100% 306s / 306s | |
| [1] Step: 16600 Acc: 0.343750 0.000000 Cost: 1.140348 1.140348 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 16700 Acc: 0.281250 0.000000 Cost: 1.072178 1.072178 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 16800 Acc: 0.468750 0.000000 Cost: 1.117922 1.117922 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 16900 Acc: 0.406250 0.000000 Cost: 1.084913 1.084913 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 17000 Acc: 0.406250 0.000000 Cost: 1.052146 1.052146 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 17100 Acc: 0.468750 0.000000 Cost: 1.055959 1.055959 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 17200 Acc: 0.468750 0.000000 Cost: 1.065861 1.065861 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 17300 Acc: 0.468750 0.000000 Cost: 1.028856 1.028856 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 17400 Acc: 0.468750 0.000000 Cost: 1.130354 1.130354 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 17500 Acc: 0.312500 0.000000 Cost: 1.116898 1.116898 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 17600 Acc: 0.468750 0.000000 Cost: 1.073465 1.073465 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 17700 Acc: 0.500000 0.000000 Cost: 1.066458 1.066458 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 17800 Acc: 0.500000 0.000000 Cost: 1.065701 1.065701 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 17900 Acc: 0.406250 0.000000 Cost: 0.994078 0.994078 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 18000 Acc: 0.437500 0.000000 Cost: 1.077106 1.077106 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 214s / 214s | |
| [1] Step: 18000 Eval acc: 0.435566 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.435566 | |
| Training [================================================================================] 100% 306s / 306s | |
| [1] Step: 18100 Acc: 0.531250 0.000000 Cost: 1.078030 1.078030 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 18200 Acc: 0.343750 0.000000 Cost: 1.056519 1.056519 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 18300 Acc: 0.500000 0.000000 Cost: 1.041512 1.041512 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 18400 Acc: 0.531250 0.000000 Cost: 1.099308 1.099308 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 18500 Acc: 0.468750 0.000000 Cost: 1.080408 1.080408 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 18600 Acc: 0.406250 0.000000 Cost: 1.034689 1.034689 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 18700 Acc: 0.500000 0.000000 Cost: 1.078049 1.078049 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 18800 Acc: 0.343750 0.000000 Cost: 1.034947 1.034947 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 18900 Acc: 0.531250 0.000000 Cost: 1.074095 1.074095 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 19000 Acc: 0.437500 0.000000 Cost: 1.094167 1.094167 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 19100 Acc: 0.437500 0.000000 Cost: 1.100367 1.100367 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 19200 Acc: 0.562500 0.000000 Cost: 1.015020 1.015020 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 19300 Acc: 0.625000 0.000000 Cost: 0.998726 0.998726 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 19400 Acc: 0.500000 0.000000 Cost: 1.068777 1.068777 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 19500 Acc: 0.437500 0.000000 Cost: 1.110436 1.110436 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 214s / 214s | |
| [1] Step: 19500 Eval acc: 0.457655 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.457655 | |
| Training [================================================================================] 100% 305s / 305s | |
| [1] Step: 19600 Acc: 0.437500 0.000000 Cost: 1.050200 1.050200 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 19700 Acc: 0.437500 0.000000 Cost: 1.083301 1.083301 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 19800 Acc: 0.375000 0.000000 Cost: 1.112352 1.112352 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 19900 Acc: 0.468750 0.000000 Cost: 1.037724 1.037724 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 20000 Acc: 0.468750 0.000000 Cost: 1.090298 1.090298 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 20100 Acc: 0.375000 0.000000 Cost: 1.140127 1.140127 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 20200 Acc: 0.468750 0.000000 Cost: 1.066080 1.066080 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 20300 Acc: 0.593750 0.000000 Cost: 0.987435 0.987435 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 20400 Acc: 0.343750 0.000000 Cost: 1.074129 1.074129 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 20500 Acc: 0.343750 0.000000 Cost: 1.044995 1.044995 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 20600 Acc: 0.343750 0.000000 Cost: 1.165304 1.165304 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 20700 Acc: 0.406250 0.000000 Cost: 1.019419 1.019419 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 20800 Acc: 0.343750 0.000000 Cost: 1.084186 1.084186 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 90s / 90s | |
| [1] Step: 20900 Acc: 0.406250 0.000000 Cost: 1.101259 1.101259 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 21000 Acc: 0.437500 0.000000 Cost: 1.038850 1.038850 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 216s / 216s | |
| [1] Step: 21000 Eval acc: 0.483713 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.483713 | |
| Training [================================================================================] 100% 307s / 307s | |
| [1] Step: 21100 Acc: 0.593750 0.000000 Cost: 0.869412 0.869412 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 89s / 89s | |
| [1] Step: 21200 Acc: 0.593750 0.000000 Cost: 0.965400 0.965400 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 93s / 93s | |
| [1] Step: 21300 Acc: 0.500000 0.000000 Cost: 1.011637 1.011637 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 94s / 94s | |
| [1] Step: 21400 Acc: 0.656250 0.000000 Cost: 1.006077 1.006077 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 91s / 91s | |
| [1] Step: 21500 Acc: 0.531250 0.000000 Cost: 0.986071 0.986071 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 93s / 93s | |
| [1] Step: 21600 Acc: 0.562500 0.000000 Cost: 0.811805 0.811805 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 96s / 96s | |
| [1] Step: 21700 Acc: 0.437500 0.000000 Cost: 1.051043 1.051043 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 103s / 103s | |
| [1] Step: 21800 Acc: 0.406250 0.000000 Cost: 1.096644 1.096644 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 93s / 93s | |
| [1] Step: 21900 Acc: 0.562500 0.000000 Cost: 1.032705 1.032705 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 92s / 92s | |
| [1] Step: 22000 Acc: 0.468750 0.000000 Cost: 1.000828 1.000828 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 95s / 95s | |
| [1] Step: 22100 Acc: 0.437500 0.000000 Cost: 1.107982 1.107982 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 101s / 101s | |
| [1] Step: 22200 Acc: 0.437500 0.000000 Cost: 1.017677 1.017677 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 101s / 101s | |
| [1] Step: 22300 Acc: 0.468750 0.000000 Cost: 1.033083 1.033083 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 99s / 99s | |
| [1] Step: 22400 Acc: 0.343750 0.000000 Cost: 1.157516 1.157516 0.000000 l2-not-exposed | |
| Training [================================================================================] 100% 96s / 96s | |
| [1] Step: 22500 Acc: 0.437500 0.000000 Cost: 1.085885 1.085885 0.000000 l2-not-exposed | |
| Run Eval [================================================================================] 100% 243s / 243s | |
| [1] Step: 22500 Eval acc: 0.510586 0.000000 ../snli_1.0/snli_1.0_dev.jsonl | |
| [1] [TODO: NOT IMPLEMENTED] Checkpointing with new best dev accuracy of 0.510586 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment