STDNN_MRPC.log

01/05/2021 02:32:44 1
01/05/2021 02:32:44 Launching the MT-DNN training
01/05/2021 02:32:44 Loading data/canonical_data/bert_base_uncased_lower/mrpc_train.json as task 0
01/05/2021 02:32:45 ####################
01/05/2021 02:32:45 {'log_file': 'checkpoints/2021-01-05T1432_bert-base-uncased_mrpc/log.log', 'tensorboard': False, 'tensorboard_logdir': 'tensorboard_logdir', 'init_checkpoint': 'bert-base-uncased', 'data_dir': 'data/canonical_data/bert_base_uncased_lower', 'data_sort_on': False, 'name': 'farmer', 'task_def': 'experiments/glue/glue_task_def.yml', 'train_datasets': ['mrpc'], 'test_datasets': ['mrpc'], 'glue_format_on': False, 'mkd_opt': 0, 'do_padding': False, 'update_bert_opt': 0, 'multi_gpu_on': True, 'mem_cum_type': 'simple', 'answer_num_turn': 5, 'answer_mem_drop_p': 0.1, 'answer_att_hidden_size': 128, 'answer_att_type': 'bilinear', 'answer_rnn_type': 'gru', 'answer_sum_att_type': 'bilinear', 'answer_merge_opt': 1, 'answer_mem_type': 1, 'max_answer_len': 10, 'answer_dropout_p': 0.1, 'answer_weight_norm_on': False, 'dump_state_on': False, 'answer_opt': 1, 'mtl_opt': 0, 'ratio': 0, 'mix_opt': 0, 'max_seq_len': 512, 'init_ratio': 1, 'encoder_type': <EncoderModelType.BERT: 1>, 'num_hidden_layers': -1, 'bert_model_type': 'bert-base-uncased', 'do_lower_case': False, 'masked_lm_prob': 0.15, 'short_seq_prob': 0.2, 'max_predictions_per_seq': 128, 'bin_on': False, 'bin_size': 64, 'bin_grow_ratio': 0.5, 'cuda': True, 'log_per_updates': 500, 'save_per_updates': 10000, 'save_per_updates_on': False, 'epochs': 20, 'batch_size': 16, 'batch_size_eval': 8, 'optimizer': 'adamax', 'grad_clipping': 0.0, 'global_grad_clipping': 1.0, 'weight_decay': 0, 'learning_rate': 5e-05, 'momentum': 0, 'warmup': 0.1, 'warmup_schedule': 'warmup_linear', 'adam_eps': 1e-06, 'vb_dropout': True, 'dropout_p': 0.1, 'dropout_w': 0.0, 'bert_dropout_p': 0.1, 'model_ckpt': 'checkpoints/model_0.pt', 'resume': False, 'have_lr_scheduler': True, 'multi_step_lr': '10,20,30', 'lr_gamma': 0.5, 'scheduler_type': 'ms', 'output_dir': 'checkpoints/2021-01-05T1432_bert-base-uncased_mrpc', 'seed': 2018, 'grad_accumulation_step': 1, 'fp16': False, 'fp16_opt_level': 'O1', 'adv_train': False, 'adv_opt': 0, 'adv_norm_level': 0, 'adv_p_norm': 'inf', 'adv_alpha': 1, 'adv_k': 1, 'adv_step_size': 0.001, 'adv_noise_var': 1e-05, 'adv_epsilon': 1e-06, 'loss_pred': True, 'collect_uncertainty': None, 'collect_topk': 0.1, 'load_ranked_data': None, 'mc_dropout': 0, 'finetune': False, 'encode_mode': False, 'task_def_list': [{'self': '{}', 'label_vocab': 'None', 'n_class': '2', 'data_type': '<DataFormat.PremiseAndOneHypothesis: 2>', 'task_type': '<TaskType.Classification: 1>', 'metric_meta': '(<Metric.ACC: 0>, <Metric.F1: 1>)', 'split_names': "['train', 'dev', 'test']", 'enable_san': 'False', 'dropout_p': 'None', 'loss': '<LossCriterion.CeCriterion: 0>', 'kd_loss': '<LossCriterion.MseCriterion: 1>', 'adv_loss': '<LossCriterion.SymKlCriterion: 7>', '__class__': "<class 'experiments.exp_def.TaskDef'>"}]}
01/05/2021 02:32:45 ####################
01/05/2021 02:32:45 ############# Gradient Accumulation Info #############
01/05/2021 02:32:45 number of step: 4600
01/05/2021 02:32:45 number of grad grad_accumulation step: 1
01/05/2021 02:32:45 adjusted number of step: 4600
01/05/2021 02:32:45 ############# Gradient Accumulation Info #############
01/05/2021 02:33:01 
############# Model Arch of MT-DNN #############
SANBertNetwork(
  (dropout_list): ModuleList(
    (0): DropoutWrapper()
  )
  (bert): BertModel(
    (embeddings): BertEmbeddings(
      (word_embeddings): Embedding(30522, 768, padding_idx=0)
      (position_embeddings): Embedding(512, 768)
      (token_type_embeddings): Embedding(2, 768)
      (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
      (dropout): Dropout(p=0.1, inplace=False)
    )
    (encoder): BertEncoder(
      (layer): ModuleList(
        (0): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (1): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (2): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (3): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (4): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (5): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (6): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (7): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (8): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (9): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (10): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
        (11): BertLayer(
          (attention): BertAttention(
            (self): BertSelfAttention(
              (query): Linear(in_features=768, out_features=768, bias=True)
              (key): Linear(in_features=768, out_features=768, bias=True)
              (value): Linear(in_features=768, out_features=768, bias=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
            (output): BertSelfOutput(
              (dense): Linear(in_features=768, out_features=768, bias=True)
              (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
              (dropout): Dropout(p=0.1, inplace=False)
            )
          )
          (intermediate): BertIntermediate(
            (dense): Linear(in_features=768, out_features=3072, bias=True)
          )
          (output): BertOutput(
            (dense): Linear(in_features=3072, out_features=768, bias=True)
            (LayerNorm): LayerNorm((768,), eps=1e-12, elementwise_affine=True)
            (dropout): Dropout(p=0.1, inplace=False)
          )
        )
      )
    )
    (pooler): BertPooler(
      (dense): Linear(in_features=768, out_features=768, bias=True)
      (activation): Tanh()
    )
  )
  (loss_pred_fc): Linear(in_features=768, out_features=1, bias=True)
  (scoring_list): ModuleList(
    (0): Linear(in_features=768, out_features=2, bias=True)
  )
)

01/05/2021 02:33:01 Total number of params: 109484547
01/05/2021 02:33:01 At epoch 0
01/05/2021 02:33:03 Task [ 0] updates[     1] train loss[1.56808] remaining[0:05:48]
01/05/2021 02:33:45 Task mrpc -- epoch 0 -- Dev ACC: 68.382
01/05/2021 02:33:45 Task mrpc -- epoch 0 -- Dev F1: 81.223
01/05/2021 02:33:49 [new test scores saved.]
01/05/2021 02:33:54 At epoch 1
01/05/2021 02:34:34 Task mrpc -- epoch 1 -- Dev ACC: 74.020
01/05/2021 02:34:34 Task mrpc -- epoch 1 -- Dev F1: 79.924
01/05/2021 02:34:38 [new test scores saved.]
01/05/2021 02:34:42 At epoch 2
01/05/2021 02:34:49 Task [ 0] updates[   500] train loss[1.39559] remaining[0:00:29]
01/05/2021 02:35:19 Task mrpc -- epoch 2 -- Dev ACC: 81.618
01/05/2021 02:35:19 Task mrpc -- epoch 2 -- Dev F1: 86.339
01/05/2021 02:35:24 [new test scores saved.]
01/05/2021 02:35:35 At epoch 3
01/05/2021 02:36:17 Task mrpc -- epoch 3 -- Dev ACC: 84.804
01/05/2021 02:36:17 Task mrpc -- epoch 3 -- Dev F1: 88.809
01/05/2021 02:36:22 [new test scores saved.]
01/05/2021 02:36:26 At epoch 4
01/05/2021 02:36:41 Task [ 0] updates[  1000] train loss[1.08183] remaining[0:00:26]
01/05/2021 02:37:09 Task mrpc -- epoch 4 -- Dev ACC: 84.314
01/05/2021 02:37:09 Task mrpc -- epoch 4 -- Dev F1: 88.811
01/05/2021 02:37:13 [new test scores saved.]
01/05/2021 02:37:17 At epoch 5
01/05/2021 02:37:59 Task mrpc -- epoch 5 -- Dev ACC: 84.069
01/05/2021 02:37:59 Task mrpc -- epoch 5 -- Dev F1: 88.246
01/05/2021 02:38:03 [new test scores saved.]
01/05/2021 02:38:07 At epoch 6
01/05/2021 02:38:29 Task [ 0] updates[  1500] train loss[0.88479] remaining[0:00:19]
01/05/2021 02:38:49 Task mrpc -- epoch 6 -- Dev ACC: 80.882
01/05/2021 02:38:49 Task mrpc -- epoch 6 -- Dev F1: 86.780
01/05/2021 02:38:54 [new test scores saved.]
01/05/2021 02:38:58 At epoch 7
01/05/2021 02:39:41 Task mrpc -- epoch 7 -- Dev ACC: 83.333
01/05/2021 02:39:41 Task mrpc -- epoch 7 -- Dev F1: 88.396
01/05/2021 02:39:45 [new test scores saved.]
01/05/2021 02:39:49 At epoch 8
01/05/2021 02:40:18 Task [ 0] updates[  2000] train loss[0.74605] remaining[0:00:12]
01/05/2021 02:40:31 Task mrpc -- epoch 8 -- Dev ACC: 83.578
01/05/2021 02:40:31 Task mrpc -- epoch 8 -- Dev F1: 88.428
01/05/2021 02:40:36 [new test scores saved.]
01/05/2021 02:40:40 At epoch 9
01/05/2021 02:41:22 Task mrpc -- epoch 9 -- Dev ACC: 85.049
01/05/2021 02:41:22 Task mrpc -- epoch 9 -- Dev F1: 89.317
01/05/2021 02:41:27 [new test scores saved.]
01/05/2021 02:41:31 At epoch 10
01/05/2021 02:42:02 Task [ 0] updates[  2500] train loss[0.66090] remaining[0:00:04]
01/05/2021 02:42:07 Task mrpc -- epoch 10 -- Dev ACC: 83.333
01/05/2021 02:42:07 Task mrpc -- epoch 10 -- Dev F1: 88.112
01/05/2021 02:42:11 [new test scores saved.]
01/05/2021 02:42:15 At epoch 11
01/05/2021 02:42:56 Task mrpc -- epoch 11 -- Dev ACC: 83.824
01/05/2021 02:42:56 Task mrpc -- epoch 11 -- Dev F1: 88.339
01/05/2021 02:43:01 [new test scores saved.]
01/05/2021 02:43:05 At epoch 12
01/05/2021 02:43:47 Task mrpc -- epoch 12 -- Dev ACC: 83.088
01/05/2021 02:43:47 Task mrpc -- epoch 12 -- Dev F1: 88.042
01/05/2021 02:43:50 [new test scores saved.]
01/05/2021 02:43:54 At epoch 13
01/05/2021 02:43:56 Task [ 0] updates[  3000] train loss[0.59217] remaining[0:00:25]
01/05/2021 02:44:22 Task mrpc -- epoch 13 -- Dev ACC: 85.294
01/05/2021 02:44:22 Task mrpc -- epoch 13 -- Dev F1: 89.761
01/05/2021 02:44:25 [new test scores saved.]
01/05/2021 02:44:33 At epoch 14
01/05/2021 02:45:04 Task mrpc -- epoch 14 -- Dev ACC: 83.824
01/05/2021 02:45:04 Task mrpc -- epoch 14 -- Dev F1: 88.660
01/05/2021 02:45:06 [new test scores saved.]
01/05/2021 02:45:11 At epoch 15
01/05/2021 02:45:16 Task [ 0] updates[  3500] train loss[0.53450] remaining[0:00:20]
01/05/2021 02:45:38 Task mrpc -- epoch 15 -- Dev ACC: 83.824
01/05/2021 02:45:38 Task mrpc -- epoch 15 -- Dev F1: 89.037
01/05/2021 02:45:41 [new test scores saved.]
01/05/2021 02:45:45 At epoch 16
01/05/2021 02:46:12 Task mrpc -- epoch 16 -- Dev ACC: 84.804
01/05/2021 02:46:12 Task mrpc -- epoch 16 -- Dev F1: 89.161
01/05/2021 02:46:15 [new test scores saved.]
01/05/2021 02:46:19 At epoch 17
01/05/2021 02:46:29 Task [ 0] updates[  4000] train loss[0.49609] remaining[0:00:16]
01/05/2021 02:46:46 Task mrpc -- epoch 17 -- Dev ACC: 83.824
01/05/2021 02:46:46 Task mrpc -- epoch 17 -- Dev F1: 88.737
01/05/2021 02:46:49 [new test scores saved.]
01/05/2021 02:46:53 At epoch 18
01/05/2021 02:47:20 Task mrpc -- epoch 18 -- Dev ACC: 84.559
01/05/2021 02:47:20 Task mrpc -- epoch 18 -- Dev F1: 89.157
01/05/2021 02:47:23 [new test scores saved.]
01/05/2021 02:47:27 At epoch 19
01/05/2021 02:47:42 Task [ 0] updates[  4500] train loss[0.46235] remaining[0:00:11]
01/05/2021 02:47:55 Task mrpc -- epoch 19 -- Dev ACC: 84.559
01/05/2021 02:47:55 Task mrpc -- epoch 19 -- Dev F1: 89.157
01/05/2021 02:47:58 [new test scores saved.]