parser.add_argument('--ignoreEntities', action="store_true", default=False, help='tag entities', required=False) args = vars(parser.parse_args()) print(args) mode=args["mode"] tagset_flat=sequence_layered_reader.read_tagset(args["tagFile_flat"]) tagset=sequence_layered_reader.read_tagset(args["tagFile_layered"]) model_file=args["modelFile"] cache_dir = os.path.join(str(PYTORCH_PRETRAINED_BERT_CACHE), 'distributed_{}'.format(0)) model = Tagger.from_pretrained('bert-base-cased', cache_dir=cache_dir, freeze_bert=True, tagset_flat=tagset_flat, tagset=tagset, device=device) model.to(device) if mode == "train": # train_folder_flat=args["trainFolder_flat"] # dev_folder_flat=args["devFolder_flat"] train_folder_layered=args["trainFolder_layered"] dev_folder_layered=args["devFolder_layered"] # flat_metric=None # if args["flat_metric"].lower() == "fscore": # flat_metric=sequence_eval.check_f1_two_lists # elif args["flat_metric"].lower() == "accuracy":