# Model, evaluation and data
parser.add_argument(
"--model_dir",
required=True,
type=str,
help="Path to save the models, logs, data and other project files")
parser.add_argument(
"--data_dir",
default="../data",
type=str,
help="The input data directory (should live outside the project dir)")
parser.add_argument("--model_type",
required=True,
type=str,
help="Model type selected from the following list: " +
", ".join(MODEL_CLASSES.keys()))
parser.add_argument(
"--num_pair_labels",
default=2,
type=int,
help="How many classes are required for the Pair Classification task?")
parser.add_argument(
"--load_train_model",
default=None,
type=str,
help="Load a specific model for **training**, instead of from: %s" %
str(MODEL_PATH_MAP))
parser.add_argument(
"--load_eval_model",
default=None,
type=str,
if args.do_eval:
trainer.load_model()
trainer.evaluate("test")
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument("--task", default=None, required=True, type=str, help="The name of the task to train")
parser.add_argument("--model_dir", default=None, required=True, type=str, help="Path to save, load model")
parser.add_argument("--data_dir", default="./data", type=str, help="The input data dir")
parser.add_argument("--intent_label_file", default="intent_label.txt", type=str, help="Intent Label file")
parser.add_argument("--slot_label_file", default="slot_label.txt", type=str, help="Slot Label file")
parser.add_argument("--model_type", default="bert", type=str, help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()))
parser.add_argument('--seed', type=int, default=1234, help="random seed for initialization")
parser.add_argument("--train_batch_size", default=32, type=int, help="Batch size for training.")
parser.add_argument("--eval_batch_size", default=64, type=int, help="Batch size for evaluation.")
parser.add_argument("--max_seq_len", default=50, type=int, help="The maximum total input sequence length after tokenization.")
parser.add_argument("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.")
parser.add_argument("--num_train_epochs", default=10.0, type=float, help="Total number of training epochs to perform.")
parser.add_argument("--weight_decay", default=0.0, type=float, help="Weight decay if we apply some.")
parser.add_argument('--gradient_accumulation_steps', type=int, default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.")
parser.add_argument("--adam_epsilon", default=1e-8, type=float, help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
parser.add_argument("--max_steps", default=-1, type=int, help="If > 0: set total number of training steps to perform. Override num_train_epochs.")
parser.add_argument("--warmup_steps", default=0, type=int, help="Linear warmup over warmup_steps.")
parser.add_argument("--dropout_rate", default=0.1, type=float, help="Dropout for fully-connected layers")
write_csvFile(os.path.join(args.data_dir, "result.csv"), results)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument("--task", default="nsmc", type=str, help="The name of the task to train")
parser.add_argument("--model_dir", default="./model", type=str, help="Path to save, load model")
parser.add_argument("--data_dir", default="./data", type=str, help="The input data dir")
parser.add_argument("--train_file", default="news_train.tsv", type=str, help="Train file")
parser.add_argument("--dev_file", default="news_test_temp.tsv", type=str, help="Test file")
parser.add_argument("--test_file", default="news_test.tsv", type=str, help="Test file")
parser.add_argument("--model_type", default="hanbert", type=str,
help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()))
parser.add_argument("--model_name_or_path", default="HanBert-54kN-torch", type=str,
help="Path to pre-trained model or shortcut name")
parser.add_argument('--seed', type=int, default=42, help="random seed for initialization")
parser.add_argument("--train_batch_size", default=32, type=int, help="Batch size for training.")
parser.add_argument("--eval_batch_size", default=64, type=int, help="Batch size for evaluation.")
parser.add_argument("--max_title_len", default=50, type=int,
help="The maximum title input sequence length after tokenization.")
parser.add_argument("--max_sentence_len", default=100, type=int,
help="The maximum sentence of contents input sequence length after tokenization.")
parser.add_argument("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.")
parser.add_argument("--num_train_epochs", default=5.0, type=float,
help="Total number of training epochs to perform.")
parser.add_argument("--weight_decay", default=0.0, type=float, help="Weight decay if we apply some.")
parser.add_argument('--gradient_accumulation_steps', type=int, default=1,
type=str,
help='File path for loading intent_label vocab')
parser.add_argument('--slot_label_file',
default='slot_label.txt',
type=str,
help='File path for loading slot_label vocab ')
parser.add_argument('--word_vocab_file',
default='word_vocab.txt',
type=str,
help='File path for loading word vocab ')
parser.add_argument('--model_type',
default='joint_bert',
type=str,
required=True,
choices=MODEL_CLASSES.keys(),
help='Model type selected in the list:' +
','.join(MODEL_CLASSES.keys()))
parser.add_argument('--random_seed',
type=int,
default=1234,
help='set random seed')
parser.add_argument(
'--max_seqLen',
type=int,
default=50,
help='Set max sequence len After tokenize text.Default is 50')
parser.add_argument(
'--num_train_epochs',
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--per_gpu_eval_batch_size",
default=20,
type=int,
help="Batch size per GPU/CPU for evaluation.")
parser.add_argument("--per_gpu_train_batch_size",
default=20,
type=int,
help="Batch size per GPU/CPU for evaluation.")
parser.add_argument("--no_cuda",
action="store_true",
help="Avoid using CUDA when available")
parser.add_argument("--model_type",
required=True,
type=str,
choices=list(MODEL_CLASSES.keys()),
help="The model architecture to be fine-tuned.")
parser.add_argument(
"--model_name_or_path",
required=True,
type=str,
help="The model checkpoint for weights initialization.",
)
parser.add_argument(
"--overwrite_cache",
action="store_true",
help="Overwrite the cached training and evaluation sets")
parser.add_argument("--sequence_length",
default=128,
type=int,
help="Sequence length for language model.")
parser.add_argument(
"--mlm_probability",
type=float,
default=0.15,
help="Ratio of tokens to mask for masked language modeling loss")
parser.add_argument("--num_train_epochs",
default=5,
type=int,
help="Total number of training epochs to perform.")
parser.add_argument(
"--max_steps",
default=-1,
type=int,
help=
"If > 0: set total number of training steps to perform. Override num_train_epochs.",
)
parser.add_argument("--warmup_steps",
default=0,
type=int,
help="Linear warmup over warmup_steps.")
parser.add_argument("--logging_steps",
type=int,
default=20,
help="Log every X updates steps.")
parser.add_argument("--save_steps",
type=int,
default=1000,
help="Save checkpoint every X updates steps.")
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=1,
help=
"Number of updates steps to accumulate before performing a backward/update pass.",
)
parser.add_argument("--learning_rate",
default=5e-5,
type=float,
help="The initial learning rate for Adam.")
parser.add_argument("--weight_decay",
default=0.0,
type=float,
help="Weight decay if we apply some.")
parser.add_argument("--adam_epsilon",
default=1e-8,
type=float,
help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm",
default=1.0,
type=float,
help="Max gradient norm.")
parser.add_argument(
"--save_total_limit",
type=int,
default=None,
help=
"Limit the total amount of checkpoints, delete the older checkpoints in the output_dir, does not delete by default",
)
parser.add_argument(
"--cache_dir",
default="",
type=str,
help=
"Optional directory to store the pre-trained models downloaded from s3 (instead of the default one)",
)
parser.add_argument(
"--output_dir",
default="",
type=str,
help=
"Optional directory to store the pre-trained models downloaded from s3 (instead of the default one)",
)
parser.add_argument("--training_type",
required=True,
choices=["comment", "post"],
help="Choose between a comment/post fine-tuned model.")
parser.add_argument("--train_files", nargs='+', help='Training file(s)')
parser.add_argument("--seed",
type=int,
default=42,
help="random seed for initialization")
parser.add_argument(
"--legacy",
action="store_true",
help="Legacy code for compatibility with older pytorch versions.")
args = parser.parse_args()
device = torch.device(
"cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
args.n_gpu = torch.cuda.device_count()
args.device = device
# Setup logging
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt='%m/%d/%Y %H:%M:%S',
level=logging.INFO)
logger.warning(
"Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
-1, device, args.n_gpu, False, False)
# Set seed
set_seed(args)
config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
config = config_class.from_pretrained(
args.model_name_or_path,
cache_dir=args.cache_dir if args.cache_dir else None)
config.output_hidden_states = True
tokenizer = tokenizer_class.from_pretrained(
args.model_name_or_path,
cache_dir=args.cache_dir if args.cache_dir else None)
model = model_class.from_pretrained(
args.model_name_or_path,
from_tf=bool('.ckpt' in args.model_name_or_path),
config=config,
cache_dir=args.cache_dir if args.cache_dir else None)
model.to(args.device)
logger.info("Training/evaluation parameters %s", args)
# Training
dataset = load_and_cache_datasets(args, tokenizer)
train(args, dataset, model, tokenizer)