liuyukid
commented
4 years ago You need to specify the values of the following variables： DATA_DIR='' MODEL_TYPE='' MODEL_NAME_OR_PATH='' OUTPUT_DIR='' LABEL=''
For example DATA_DIR='../datasets/conll2003' MODEL_TYPE='bert' MODEL_NAME_OR_PATH='../pretrained_models/bert_base_cased' OUTPUT_DIR='../output/conll2003/bert' LABEL='../datasets/conll2003/labels.txt'
(.venvpy36) transformers-ner/scripts % bash run_softmax_ner.sh usage: run_softmax_ner.py [-h] --data_dir DATA_DIR --model_type MODEL_TYPE --model_name_or_path MODEL_NAME_OR_PATH --output_dir OUTPUT_DIR [--labels LABELS] [--config_name CONFIG_NAME] [--tokenizer_name TOKENIZER_NAME] [--cache_dir CACHE_DIR] [--max_seq_length MAX_SEQ_LENGTH] [--do_train] [--do_eval] [--do_predict] [--evaluate_during_training] [--do_lower_case] [--keep_accents] [--strip_accents] [--use_fast] [--per_gpu_train_batch_size PER_GPU_TRAIN_BATCH_SIZE] [--per_gpu_eval_batch_size PER_GPU_EVAL_BATCH_SIZE] [--gradient_accumulation_steps GRADIENT_ACCUMULATION_STEPS] [--loss_type LOSS_TYPE] [--learning_rate LEARNING_RATE] [--bert_lr BERT_LR] [--classifier_lr CLASSIFIER_LR] [--adv_training {fgm,pgd}] [--weight_decay WEIGHT_DECAY] [--adam_epsilon ADAM_EPSILON] [--max_grad_norm MAX_GRAD_NORM] [--num_train_epochs NUM_TRAIN_EPOCHS] [--max_steps MAX_STEPS] [--warmup_steps WARMUP_STEPS] [--logging_steps LOGGING_STEPS] [--eval_all_checkpoints] [--no_cuda] [--overwrite_output_dir] [--overwrite_cache] [--seed SEED] [--fp16] [--fp16_opt_level FP16_OPT_LEVEL] [--local_rank LOCAL_RANK] [--server_ip SERVER_IP] [--server_port SERVER_PORT] run_softmax_ner.py: error: argument --data_dir: expected one argument