Training/evaluation parameters Namespace(data_dir='intent_example/', model_dir='/data/ryanyip/huggingface-models/', model_name='mc_bert_base', output_dir='./intent_test', do_train=True, do_predict=False, result_output_dir='./intent_test/result/', max_length=128, train_batch_size=8, eval_batch_size=8, learning_rate=5e-05, weight_decay=0.01, adam_epsilon=1e-08, max_grad_norm=1.0, epochs=3, warmup_proportion=0.1, earlystop_patience=2, logging_steps=10, save_steps=1000, seed=2021, device=device(type='cuda'), model_type='bert', task_name='qic')
***** Running training *****
Num samples 1350
Num epochs 3
Num training steps 507
Num warmup steps 50
***** Running evaluation *****
Num samples 170
qic-mc_bert_base acc: 0.1
Saving models checkpoint to ./intent_test/checkpoint-10
***** Running evaluation *****
Num samples 170
qic-mc_bert_base acc: 0.18823529411764706
Saving models checkpoint to ./intent_test/checkpoint-20
***** Running evaluation *****
Num samples 170
qic-mc_bert_base acc: 0.2235294117647059
Saving models checkpoint to ./intent_test/checkpoint-30
***** Running evaluation *****
Num samples 170
qic-mc_bert_base acc: 0.24705882352941178
Saving models checkpoint to ./intent_test/checkpoint-40
***** Running evaluation *****
Num samples 170
qic-mc_bert_base acc: 0.2529411764705882
Saving models checkpoint to ./intent_test/checkpoint-50
***** Running evaluation *****
Num samples 170
qic-mc_bert_base acc: 0.32941176470588235
Saving models checkpoint to ./intent_test/checkpoint-60
***** Running evaluation *****
Num samples 170
qic-mc_bert_base acc: 0.2647058823529412
Earlystopper counter: 1 out of 2
***** Running evaluation *****
Num samples 170
qic-mc_bert_base acc: 0.27647058823529413
Earlystopper counter: 2 out of 2
Training Stop! The best step 60: 0.32941176470588235
Saving models checkpoint to ./intent_test