import code
import json
import torch
import re
import numpy as np
from model.modeling_albert import AlbertConfig, AlbertForSequenceClassification
from model import tokenization_albert
from model.file_utils import WEIGHTS_NAME
tokenizer = tokenization_albert.FullTokenizer(
vocab_file='./trained_bert_model/albert_small/vocab.txt')
config = AlbertConfig.from_pretrained('./trained_bert_model/albert_small/',
num_labels=2,
finetuning_task='lcqmc')
model = AlbertForSequenceClassification.from_pretrained(
'./trained_bert_model/albert_small/', config=config)
space_pattern = re.compile(r'\s')
alphanum_pattern = re.compile(r'\(?[a-zA-Z0-9]+-?[a-zA-Z0-9]*\)?')
pattern_string = r'^你知道|请问|我想知道|谁知道|我很好奇|有谁知道|大家知道|有人知道|请告诉我|我想了解一下|请说出|告诉我|能告诉我|你了解|你清楚|' \
r'你能说出|谁是|你能告诉我|我想请问|我想问问|我想问一下|我想了解|我很想知道|你们知道|问一下|我好奇|谁能告诉我|请问一下|你觉得|什么是'
def bert_output(question, predicate):
question_sep = tokenizer.tokenize(question)
predicate_sep = tokenizer.tokenize(predicate)
text = ["[CLS]"] + question_sep + ["[SEP]"] + predicate_sep + ["[SEP]"]
indexed_tokens = tokenizer.convert_tokens_to_ids(text)
token_type_ids = [0] * (len(question_sep) +
2) + [1] * (len(predicate_sep) + 1)
tokens_tensor = torch.tensor([indexed_tokens])
def main():
parser = argparse.ArgumentParser()
# parser.add_argument("--arch", default='albert_xlarge', type=str)
parser.add_argument("--arch", default='albert_large', type=str)
parser.add_argument('--bert_dir',
default='pretrain/pytorch/albert_large_zh',
type=str)
parser.add_argument('--albert_config_path',
default='configs/albert_config_large.json',
type=str)
parser.add_argument('--task_name', default='lcqmc', type=str)
parser.add_argument(
"--train_max_seq_len",
default=64,
type=int,
help=
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded.")
parser.add_argument(
"--eval_max_seq_len",
default=64,
type=int,
help=
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded.")
parser.add_argument('--share_type',
default='all',
type=str,
choices=['all', 'attention', 'ffn', 'None'])
parser.add_argument("--do_train",
action='store_true',
help="Whether to run training.")
parser.add_argument("--do_eval",
action='store_true',
help="Whether to run eval on the dev set.")
parser.add_argument("--do_test",
action='store_true',
help="Whether to run eval on the test set.")
parser.add_argument(
"--evaluate_during_training",
action='store_true',
help="Rul evaluation during training at each logging step.")
parser.add_argument(
"--do_lower_case",
action='store_true',
help="Set this flag if you are using an uncased model.")
parser.add_argument("--train_batch_size",
default=32,
type=int,
help="Batch size per GPU/CPU for training.")
parser.add_argument("--eval_batch_size",
default=16,
type=int,
help="Batch size per GPU/CPU for evaluation.")
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=2e-5,
type=float,
help="The initial learning rate for Adam.")
parser.add_argument("--weight_decay",
default=0.1,
type=float,
help="Weight deay 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=5.0,
type=float,
help="Max gradient norm.")
parser.add_argument("--num_train_epochs",
default=3.0,
type=float,
help="Total number of training epochs to perform.")
parser.add_argument(
"--warmup_proportion",
default=0.1,
type=int,
help=
"Proportion of training to perform linear learning rate warmup for,E.g., 0.1 = 10% of training."
)
parser.add_argument(
"--eval_all_checkpoints",
action='store_true',
help=
"Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number"
)
parser.add_argument("--no_cuda",
action='store_true',
help="Avoid using CUDA when available")
parser.add_argument('--overwrite_output_dir',
action='store_true',
help="Overwrite the content of the output directory")
parser.add_argument(
'--overwrite_cache',
action='store_true',
help="Overwrite the cached training and evaluation sets")
parser.add_argument('--seed',
type=int,
default=42,
help="random seed for initialization")
parser.add_argument(
'--fp16',
action='store_true',
help=
"Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit"
)
parser.add_argument(
'--fp16_opt_level',
type=str,
default='O1',
help=
"For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
"See details at https://nvidia.github.io/apex/amp.html")
parser.add_argument("--local_rank",
type=int,
default=-1,
help="For distributed training: local_rank")
parser.add_argument('--server_ip',
type=str,
default='',
help="For distant debugging.")
parser.add_argument('--server_port',
type=str,
default='',
help="For distant debugging.")
args = parser.parse_args()
# Fix bug: Config is wrong from base.py if it is not base
config['bert_dir'] = args.bert_dir
config['albert_config_path'] = args.albert_config_path
args.model_save_path = config['checkpoint_dir'] / f'{args.arch}'
args.model_save_path.mkdir(exist_ok=True)
# Setudistant debugging if needed
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port),
redirect_output=True)
ptvsd.wait_for_attach()
# Setup CUDA, GPU & distributed training
if args.local_rank == -1 or args.no_cuda:
device = torch.device("cuda" if torch.cuda.is_available()
and not args.no_cuda else "cpu")
args.n_gpu = torch.cuda.device_count()
else: # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
torch.distributed.init_process_group(backend='nccl')
args.n_gpu = 1
args.device = device
init_logger(log_file=config['log_dir'] / 'finetuning.log')
logger.warning(
"Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
args.local_rank, device, args.n_gpu, bool(args.local_rank != -1),
args.fp16)
# Set seed
seed_everything(args.seed)
# --------- data
processor = AlbertProcessor(vocab_path=config['albert_vocab_path'],
do_lower_case=args.do_lower_case)
label_list = processor.get_labels()
num_labels = len(label_list)
if args.local_rank not in [-1, 0]:
torch.distributed.barrier(
) # Make sure only the first process in distributed training will download model & vocab
bert_config = AlbertConfig.from_pretrained(str(
config['albert_config_path']),
share_type=args.share_type,
num_labels=num_labels)
logger.info("Training/evaluation parameters %s", args)
metrics = Accuracy(topK=1)
# Training
if args.do_train:
train_data = processor.get_train(config['data_dir'] / "train.txt")
train_examples = processor.create_examples(
lines=train_data,
example_type='train',
cached_examples_file=config['data_dir'] /
f"cached_train_examples_{args.arch}")
train_features = processor.create_features(
examples=train_examples,
max_seq_len=args.train_max_seq_len,
cached_features_file=config['data_dir'] /
"cached_train_features_{}_{}".format(args.train_max_seq_len,
args.arch))
train_dataset = processor.create_dataset(train_features)
train_sampler = RandomSampler(train_dataset)
train_dataloader = DataLoader(train_dataset,
sampler=train_sampler,
batch_size=args.train_batch_size)
valid_data = processor.get_dev(config['data_dir'] / "dev.txt")
valid_examples = processor.create_examples(
lines=valid_data,
example_type='valid',
cached_examples_file=config['data_dir'] /
f"cached_valid_examples_{args.arch}")
valid_features = processor.create_features(
examples=valid_examples,
max_seq_len=args.eval_max_seq_len,
cached_features_file=config['data_dir'] /
"cached_valid_features_{}_{}".format(args.eval_max_seq_len,
args.arch))
valid_dataset = processor.create_dataset(valid_features)
valid_sampler = SequentialSampler(valid_dataset)
valid_dataloader = DataLoader(valid_dataset,
sampler=valid_sampler,
batch_size=args.eval_batch_size)
model = AlbertForSequenceClassification.from_pretrained(
config['bert_dir'], config=bert_config)
if args.local_rank == 0:
torch.distributed.barrier(
) # Make sure only the first process in distributed training will download model & vocab
model.to(args.device)
train(args, train_dataloader, valid_dataloader, metrics, model)
if args.do_test:
test_data = processor.get_train(config['data_dir'] / "test.txt")
test_examples = processor.create_examples(
lines=test_data,
example_type='test',
cached_examples_file=config['data_dir'] /
f"cached_test_examples_{args.arch}")
test_features = processor.create_features(
examples=test_examples,
max_seq_len=args.eval_max_seq_len,
cached_features_file=config['data_dir'] /
"cached_test_features_{}_{}".format(args.eval_max_seq_len,
args.arch))
test_dataset = processor.create_dataset(test_features)
test_sampler = SequentialSampler(test_dataset)
test_dataloader = DataLoader(test_dataset,
sampler=test_sampler,
batch_size=args.eval_batch_size)
model = AlbertForSequenceClassification.from_pretrained(
args.model_save_path, config=bert_config)
model.to(args.device)
test_log = evaluate(args, model, test_dataloader, metrics)
print(test_log)
def for_server(
text: str,
task: str,
):
if args.local_rank == -1 or args.no_cuda:
device = torch.device("cuda" if torch.cuda.is_available()
and not args.no_cuda else "cpu")
args.n_gpu = torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
args.n_gpu = 1
args.device = device
text = [''] + [['0'] + ['0'] + [text]]
processor = processors[task]
output_mode = output_modes[task]
label_list = processor.get_labels()
# num_labels = len(label_list)
examples = processor._create_examples(text, 'predict')
if args.local_rank in [-1, 0]:
tokenizer = tokenization_albert.FullTokenizer(
vocab_file=args.vocab_file,
do_lower_case=args.do_lower_case,
)
checkpoints = [(0, args.output_dir)]
if args.eval_all_checkpoints:
checkpoints = list(
os.path.dirname(c) for c in sorted(
glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME,
recursive=True)))
checkpoints = [(int(checkpoint.split('-')[-1]), checkpoint)
for checkpoint in checkpoints
if checkpoint.find('checkpoint') != -1]
checkpoints = sorted(checkpoints, key=lambda x: x[0])
logger.info("Evaluate the following checkpoints: %s", checkpoints)
if len(checkpoints) == 0:
checkpoints = [(0, args.output_dir)]
else:
checkpoints = [checkpoints[-1]]
for _, checkpoint in checkpoints:
if len(os.listdir(checkpoint)) == 0:
main()
model = AlbertForSequenceClassification.from_pretrained(checkpoint)
model.to(args.device)
features = convert_examples_to_features(
examples,
tokenizer,
label_list=label_list,
max_seq_length=args.max_seq_length,
output_mode=output_mode)
all_input_ids = torch.tensor([f.input_ids for f in features],
dtype=torch.long)
all_attention_mask = torch.tensor(
[f.attention_mask for f in features], dtype=torch.long)
all_token_type_ids = torch.tensor(
[f.token_type_ids for f in features], dtype=torch.long)
all_lens = torch.tensor([f.input_len for f in features],
dtype=torch.long)
all_labels = torch.tensor([f.label for f in features],
dtype=torch.long)
dataset = TensorDataset(all_input_ids, all_attention_mask,
all_token_type_ids, all_lens, all_labels)
for step, batch in enumerate(dataset):
model.eval()
batch = tuple(t.to(args.device) for t in batch)
with torch.no_grad():
inputs = {
'input_ids': batch[0].unsqueeze(0),
'attention_mask': batch[1].unsqueeze(0),
}
inputs['token_type_ids'] = batch[2].unsqueeze(0)
outputs = model(**inputs)
logits = outputs[0]
preds = np.argmax(logits, axis=1)
label = tasks_num_labels[task][preds]
logger.info('label is {}'.format(label))
return label
def main():
if not os.path.exists(args.output_dir):
os.mkdir(args.output_dir)
# type_task = args.model_type + '_' + '{}'.format(args.task_name)
# if not os.path.exists(os.path.join(args.output_dir, type_task)):
# os.mkdir(os.path.join(args.output_dir, type_task))
init_logger(log_file=args.output_dir +
'/{}-{}.log'.format(args.model_type, args.task_name))
if args.local_rank == -1 or args.no_cuda:
device = torch.device("cuda" if torch.cuda.is_available()
and not args.no_cuda else "cpu")
args.n_gpu = torch.cuda.device_count()
else:
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
args.n_gpu = 1
args.device = device
# logger.warning("Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
# args.local_rank, device, args.n_gpu, bool(args.local_rank != -1), args.fp16)
seed_everything(args.seed)
args.task_name = args.task_name.lower()
if args.task_name not in processors:
raise ValueError("Task not found: %s" % (args.task_name))
processor = processors[args.task_name]
args.output_mode = output_modes[args.task_name]
label_list = processor.get_labels()
num_labels = len(label_list)
# if args.local_rank not in [-1, 0]:
# torch.distributed.barrier() # Make sure only the first process in distributed training will download model & vocab
args.model_type = args.model_type.lower()
config = AlbertConfig.from_pretrained(
args.config_name if args.config_name else args.model_name_or_path,
num_labels=num_labels,
finetuning_task=args.task_name)
tokenizer = tokenization_albert.FullTokenizer(
vocab_file=args.vocab_file,
do_lower_case=args.do_lower_case,
)
model = AlbertForSequenceClassification.from_pretrained(
args.model_name_or_path, config=config)
model.to(args.device)
logger.info("Training/evaluation parameters %s", args)
# Training
args.do_train = True
if args.do_train:
train_dataset = load_and_cache_examples(args,
args.task_name,
tokenizer,
data_type='train')
global_step, tr_loss = train(args, train_dataset, model, tokenizer)
logger.info(" global_step = %s, average loss = %s", global_step,
tr_loss)
if args.do_train:
if not os.path.exists(args.output_dir) and args.local_rank in [-1, 0]:
os.makedirs(args.output_dir)
logger.info("Saving model checkpoint to %s", args.output_dir)
model_to_save = model.module if hasattr(
model,
'module') else model # Take care of distributed/parallel training
model_to_save.save_pretrained(args.output_dir)
torch.save(args, os.path.join(args.output_dir, 'training_args.bin'))
# Evaluation
args.do_eval = True
results = []
if args.do_eval and args.local_rank in [-1, 0]:
tokenizer = tokenization_albert.FullTokenizer(
vocab_file=args.vocab_file,
do_lower_case=args.do_lower_case,
)
checkpoints = [(0, args.output_dir)]
if args.eval_all_checkpoints:
checkpoints = list(
os.path.dirname(c) for c in sorted(
glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME,
recursive=True)))
checkpoints = [(int(checkpoint.split('-')[-1]), checkpoint)
for checkpoint in checkpoints
if checkpoint.find('checkpoint') != -1]
checkpoints = sorted(checkpoints, key=lambda x: x[0])
logger.info("Evaluate the following checkpoints: %s", checkpoints)
for _, checkpoint in checkpoints:
global_step = checkpoint.split(
'-')[-1] if len(checkpoints) > 1 else ""
prefix = checkpoint.split(
'/')[-1] if checkpoint.find('checkpoint') != -1 else ""
model = AlbertForSequenceClassification.from_pretrained(checkpoint)
model.to(args.device)
result = evaluate(args, model, tokenizer, prefix=prefix)
results.extend([(k + '_{}'.format(global_step), v)
for k, v in result.items()])
output_eval_file = os.path.join(args.output_dir,
"checkpoint_eval_results.txt")
with open(output_eval_file, "w") as writer:
for key, value in results:
writer.write("%s = %s\n" % (key, str(value)))
args.do_predict = True
predict_results = []
if args.do_predict and args.local_rank in [-1, 0]:
tokenizer = tokenization_albert.FullTokenizer(
vocab_file=args.vocab_file,
do_lower_case=args.do_lower_case,
)
checkpoints = [(0, args.output_dir)]
if args.eval_all_checkpoints:
checkpoints = list(
os.path.dirname(c) for c in sorted(
glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME,
recursive=True)))
checkpoints = [(int(checkpoint.split('-')[-1]), checkpoint)
for checkpoint in checkpoints
if checkpoint.find('checkpoint') != -1]
checkpoints = sorted(checkpoints, key=lambda x: x[0])
logger.info("Evaluate the following checkpoints: %s", checkpoints)
checkpoints = [checkpoints[-1]]
for _, checkpoint in checkpoints:
global_step = checkpoint.split(
'-')[-1] if len(checkpoints) > 1 else ""
prefix = checkpoint.split(
'/')[-1] if checkpoint.find('checkpoint') != -1 else ""
model = AlbertForSequenceClassification.from_pretrained(checkpoint)
model.to(args.device)
result = predict(args, model, tokenizer, prefix=prefix)
predict_results.extend([(k + '_{}'.format(global_step), v)
for k, v in result.items()])
output_eval_file = os.path.join(args.output_dir,
"checkpoint_eval_results.txt")
with open(output_eval_file, "w") as writer:
for key, value in predict_results:
writer.write("%s = %s\n" % (key, str(value)))
def main():
parser = argparse.ArgumentParser()
## Required parameters
parser.add_argument(
"--data_dir",
default=None,
type=str,
required=True,
help=
"The input data dir. Should contain the .tsv files (or other data files) for the task."
)
parser.add_argument("--model_type",
default=None,
type=str,
required=True,
help="Model type selected in the list: ")
parser.add_argument(
"--model_name_or_path",
default=None,
type=str,
required=True,
help="Path to pre-trained model or shortcut name selected in the list")
parser.add_argument(
"--task_name",
default=None,
type=str,
required=True,
help="The name of the task to train selected in the list: " +
", ".join(processors.keys()))
parser.add_argument(
"--output_dir",
default=None,
type=str,
required=True,
help=
"The output directory where the model predictions and checkpoints will be written."
)
parser.add_argument("--vocab_file", default=None, type=str)
parser.add_argument("--spm_model_file",
default=None,
required=True,
type=str)
## Other parameters
parser.add_argument(
"--config_name",
default="",
type=str,
help="Pretrained config name or path if not the same as model_name")
parser.add_argument(
"--tokenizer_name",
default="",
type=str,
help="Pretrained tokenizer name or path if not the same as model_name")
parser.add_argument(
"--cache_dir",
default="",
type=str,
help=
"Where do you want to store the pre-trained models downloaded from s3")
parser.add_argument(
"--max_seq_length",
default=512,
type=int,
help=
"The maximum total input sequence length after tokenization. Sequences longer "
"than this will be truncated, sequences shorter will be padded.")
parser.add_argument("--do_train",
action='store_true',
help="Whether to run training.")
parser.add_argument("--do_eval",
action='store_true',
help="Whether to run eval on the dev set.")
parser.add_argument(
"--do_predict",
action='store_true',
help="Whether to run the model in inference mode on the test set.")
parser.add_argument(
"--do_lower_case",
action='store_true',
help="Set this flag if you are using an uncased model.")
parser.add_argument('--share_type',
default='all',
type=str,
choices=['all', 'attention', 'ffn', 'None'])
parser.add_argument("--per_gpu_train_batch_size",
default=8,
type=int,
help="Batch size per GPU/CPU for training.")
parser.add_argument("--per_gpu_eval_batch_size",
default=8,
type=int,
help="Batch size per GPU/CPU for evaluation.")
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 deay if we apply some.")
parser.add_argument("--adam_epsilon",
default=1e-6,
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("--num_train_epochs",
default=3.0,
type=float,
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_proportion",
default=0.1,
type=float,
help=
"Proportion of training to perform linear learning rate warmup for,E.g., 0.1 = 10% of training."
)
parser.add_argument('--logging_steps',
type=int,
default=10,
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(
"--eval_all_checkpoints",
action='store_true',
help=
"Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number"
)
parser.add_argument("--no_cuda",
action='store_true',
help="Avoid using CUDA when available")
parser.add_argument('--overwrite_output_dir',
action='store_true',
help="Overwrite the content of the output directory")
parser.add_argument(
'--overwrite_cache',
action='store_true',
help="Overwrite the cached training and evaluation sets")
parser.add_argument('--seed',
type=int,
default=42,
help="random seed for initialization")
parser.add_argument(
'--fp16',
action='store_true',
help=
"Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit"
)
parser.add_argument(
'--fp16_opt_level',
type=str,
default='O1',
help=
"For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
"See details at https://nvidia.github.io/apex/amp.html")
parser.add_argument("--local_rank",
type=int,
default=-1,
help="For distributed training: local_rank")
parser.add_argument('--server_ip',
type=str,
default='',
help="For distant debugging.")
parser.add_argument('--server_port',
type=str,
default='',
help="For distant debugging.")
args = parser.parse_args()
if not os.path.exists(args.output_dir):
os.mkdir(args.output_dir)
init_logger(log_file=args.output_dir +
'/{}-{}.log'.format(args.model_type, args.task_name))
if os.path.exists(args.output_dir) and os.listdir(
args.output_dir
) and args.do_train and not args.overwrite_output_dir:
raise ValueError(
"Output directory ({}) already exists and is not empty. Use --overwrite_output_dir to overcome."
.format(args.output_dir))
# Setup distant debugging if needed
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port),
redirect_output=True)
ptvsd.wait_for_attach()
# Setup CUDA, GPU & distributed training
if args.local_rank == -1 or args.no_cuda:
device = torch.device("cuda" if torch.cuda.is_available()
and not args.no_cuda else "cpu")
args.n_gpu = torch.cuda.device_count()
else: # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
torch.distributed.init_process_group(backend='nccl')
args.n_gpu = 1
args.device = device
# Setup logging
logger.warning(
"Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
args.local_rank, device, args.n_gpu, bool(args.local_rank != -1),
args.fp16)
# Set seed
seed_everything(args.seed)
# Prepare GLUE task
args.task_name = args.task_name.lower()
if args.task_name not in processors:
raise ValueError("Task not found: %s" % (args.task_name))
processor = processors[args.task_name]()
args.output_mode = output_modes[args.task_name]
label_list = processor.get_labels()
num_labels = len(label_list)
# Load pretrained model and tokenizer
if args.local_rank not in [-1, 0]:
torch.distributed.barrier(
) # Make sure only the first process in distributed training will download model & vocab
args.model_type = args.model_type.lower()
config = BertConfig.from_pretrained(
args.config_name if args.config_name else args.model_name_or_path,
num_labels=num_labels,
finetuning_task=args.task_name,
share_type=args.share_type)
tokenizer = tokenization.FullTokenizer(vocab_file=args.vocab_file,
do_lower_case=args.do_lower_case,
spm_model_file=args.spm_model_file)
model = AlbertForSequenceClassification.from_pretrained(
args.model_name_or_path,
from_tf=bool('.ckpt' in args.model_name_or_path),
config=config)
if args.local_rank == 0:
torch.distributed.barrier(
) # Make sure only the first process in distributed training will download model & vocab
model.to(args.device)
logger.info("Training/evaluation parameters %s", args)
# Training
if args.do_train:
train_dataset = load_and_cache_examples(args,
args.task_name,
tokenizer,
data_type='train')
global_step, tr_loss = train(args, train_dataset, model, tokenizer)
logger.info(" global_step = %s, average loss = %s", global_step,
tr_loss)
# Saving best-practices: if you use defaults names for the model, you can reload it using from_pretrained()
if args.do_train and (args.local_rank == -1
or torch.distributed.get_rank() == 0):
# Create output directory if needed
if not os.path.exists(args.output_dir) and args.local_rank in [-1, 0]:
os.makedirs(args.output_dir)
logger.info("Saving model checkpoint to %s", args.output_dir)
# Save a trained model, configuration and tokenizer using `save_pretrained()`.
# They can then be reloaded using `from_pretrained()`
model_to_save = model.module if hasattr(
model,
'module') else model # Take care of distributed/parallel training
model_to_save.save_pretrained(args.output_dir)
# Good practice: save your training arguments together with the trained model
torch.save(args, os.path.join(args.output_dir, 'training_args.bin'))
# Evaluation
results = []
if args.do_eval and args.local_rank in [-1, 0]:
tokenizer = tokenization.FullTokenizer(
vocab_file=args.vocab_file,
do_lower_case=args.do_lower_case,
spm_model_file=args.spm_model_file)
checkpoints = [(0, args.output_dir)]
if args.eval_all_checkpoints:
checkpoints = list(
os.path.dirname(c) for c in sorted(
glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME,
recursive=True)))
logging.getLogger("transformers.modeling_utils").setLevel(
logging.WARN) # Reduce logging
checkpoints = [(int(checkpoint.split('-')[-1]), checkpoint)
for checkpoint in checkpoints
if checkpoint.find('checkpoint') != -1]
checkpoints = sorted(checkpoints, key=lambda x: x[0])
logger.info("Evaluate the following checkpoints: %s", checkpoints)
for _, checkpoint in checkpoints:
global_step = checkpoint.split(
'-')[-1] if len(checkpoints) > 1 else ""
prefix = checkpoint.split(
'/')[-1] if checkpoint.find('checkpoint') != -1 else ""
model = AlbertForSequenceClassification.from_pretrained(checkpoint)
model.to(args.device)
result = evaluate(args, model, tokenizer, prefix=prefix)
results.extend([(k + '_{}'.format(global_step), v)
for k, v in result.items()])
output_eval_file = os.path.join(args.output_dir,
"checkpoint_eval_results.txt")
with open(output_eval_file, "w") as writer:
for key, value in results:
writer.write("%s = %s\n" % (key, str(value)))
def main():
parser = argparse.ArgumentParser()
## Required parameters
parser.add_argument("--data_dir", default='dataset/car_data', type=str, required=False,
help="输入数据文件地址")
parser.add_argument("--model_type", default='albert', type=str, required=False,
help="模型种类")
parser.add_argument("--model_name_or_path", default='prev_trained_model/albert_chinese_small', type=str,
required=False,
help="模型参数文件地址")
parser.add_argument("--task_name", default='car', type=str, required=False,
help="那个种类数据" + ", ".join(processors.keys()))
parser.add_argument("--output_dir", default='outputs', type=str, required=False,
help="输出文件地址")
parser.add_argument("--vocab_file", default='prev_trained_model/albert_chinese_small/vocab.txt', type=str)
## Other parameters
parser.add_argument("--config_name", default="", type=str,
help="配置文件地址")
parser.add_argument("--tokenizer_name", default="", type=str,
help="Pretrained tokenizer name or path if not the same as model_name")
parser.add_argument("--cache_dir", default="", type=str,
help="Where do you want to store the pre-trained models downloaded from s3")
parser.add_argument("--max_seq_length", default=512, type=int,
help="句子最大长度")
parser.add_argument("--do_train", action='store_true',
help="训练")
parser.add_argument("--do_eval", action='store_true',
help="验证")
parser.add_argument("--do_predict", action='store_true',
help="预测")
parser.add_argument("--do_lower_case", action='store_true',
help="Set this flag if you are using an uncased model.")
parser.add_argument("--per_gpu_train_batch_size", default=8, type=int,
help="批量大小")
parser.add_argument("--per_gpu_eval_batch_size", default=8, type=int,
help="验证批量大小")
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="Adam学习率")
parser.add_argument("--weight_decay", default=0.0, type=float,
help="Weight deay if we apply some.")
parser.add_argument("--adam_epsilon", default=1e-6, 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("--num_train_epochs", default=3.0, type=float,
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_proportion", default=0.1, type=float,
help="Proportion of training to perform linear learning rate warmup for,E.g., 0.1 = 10% of training.")
parser.add_argument('--logging_steps', type=int, default=10,
help="Log every X updates steps.")
parser.add_argument('--save_steps', type=int, default=1000,
help="每多少部保存一次")
parser.add_argument("--eval_all_checkpoints",type=str,default='do',# action='store_true',
help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number")
parser.add_argument("--no_cuda", type=int, default=0, # action='store_true',
help="GPU")
parser.add_argument('--overwrite_output_dir', action='store_true',
help="Overwrite the content of the output directory")
parser.add_argument('--overwrite_cache', action='store_true',
help="Overwrite the cached training and evaluation sets")
parser.add_argument('--seed', type=int, default=42,
help="随机种子")
parser.add_argument('--fp16', action='store_true',
help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit")
parser.add_argument('--fp16_opt_level', type=str, default='O1',
help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
"See details at https://nvidia.github.io/apex/amp.html")
parser.add_argument("--local_rank", type=int, default=0,
help="For distributed training: local_rank")
args = parser.parse_args()
if not os.path.exists(args.output_dir):
os.mkdir(args.output_dir)
type_task = args.model_type + '_' + '{}'.format(args.task_name)
if not os.path.exists(os.path.join(args.output_dir, type_task)):
os.mkdir(os.path.join(args.output_dir, type_task))
init_logger(log_file=args.output_dir + '/{}-{}.log'.format(args.model_type, args.task_name))
# Setup CUDA, GPU & distributed training
if args.local_rank == -1 or args.no_cuda:
device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
args.n_gpu = torch.cuda.device_count()
else: # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
# torch.distributed.init_process_group(backend='nccl')
args.n_gpu = 1
args.device = device
# Setup logging
logger.warning("Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
args.local_rank, device, args.n_gpu, bool(args.local_rank != -1), args.fp16)
# Set seed
seed_everything(args.seed)
# Prepare GLUE task
args.task_name = args.task_name.lower()
if args.task_name not in processors:
raise ValueError("Task not found: %s" % (args.task_name))
processor = processors[args.task_name]()
args.output_mode = output_modes[args.task_name]
label_list = processor.get_labels()
num_labels = len(label_list)
args.model_type = args.model_type.lower()
config = AlbertConfig.from_pretrained(args.config_name if args.config_name else args.model_name_or_path,
num_labels=num_labels,
finetuning_task=args.task_name)
tokenizer = tokenization_albert.FullTokenizer(vocab_file=args.vocab_file, do_lower_case=args.do_lower_case,
)
model =AlbertForSequenceClassification.from_pretrained(args.model_name_or_path, config=config)
#if args.local_rank == 0:
# torch.distributed.barrier() # Make sure only the first process in distributed training will download model & vocab
model.to(args.device)
logger.info("Training/evaluation parameters %s", args)
# Training
# args.do_train = True
if args.do_train:
train_dataset = load_and_cache_examples(args, args.task_name, tokenizer, data_type='train')
global_step, tr_loss = train(args, train_dataset, model, tokenizer)
logger.info(" global_step = %s, average loss = %s", global_step, tr_loss)
# Saving best-practices: if you use defaults names for the model, you can reload it using from_pretrained()
if args.do_train:# and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
# Create output directory if needed
if not os.path.exists(args.output_dir) and args.local_rank in [-1, 0]:
os.makedirs(args.output_dir)
logger.info("Saving model checkpoint to %s", args.output_dir)
# Save a trained model, configuration and tokenizer using `save_pretrained()`.
# They can then be reloaded using `from_pretrained()`
model_to_save = model.module if hasattr(model,
'module') else model # Take care of distributed/parallel training
model_to_save.save_pretrained(args.output_dir)
# Good practice: save your training arguments together with the trained model
torch.save(args, os.path.join(args.output_dir, 'training_args.bin'))
# Evaluation
# args.do_eval = True
results = []
if args.do_eval and args.local_rank in [-1, 0]:
tokenizer = tokenization_albert.FullTokenizer(vocab_file=args.vocab_file,
do_lower_case=args.do_lower_case,
)
checkpoints = [(0,args.output_dir)]
if args.eval_all_checkpoints:
checkpoints = list(
os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME, recursive=True)))
checkpoints = [(int(checkpoint.split('-')[-1]),checkpoint) for checkpoint in checkpoints if checkpoint.find('checkpoint') != -1]
checkpoints = sorted(checkpoints,key =lambda x:x[0])
logger.info("Evaluate the following checkpoints: %s", checkpoints)
for _,checkpoint in checkpoints:
global_step = checkpoint.split('-')[-1] if len(checkpoints) > 1 else ""
prefix = checkpoint.split('/')[-1] if checkpoint.find('checkpoint') != -1 else ""
model =AlbertForSequenceClassification.from_pretrained(checkpoint)
model.to(args.device)
result = evaluate(args, model, tokenizer, prefix=prefix)
results.extend([(k + '_{}'.format(global_step), v) for k, v in result.items()])
output_eval_file = os.path.join(args.output_dir, "checkpoint_eval_results.txt")
with open(output_eval_file, "w") as writer:
for key,value in results:
writer.write("%s = %s\n" % (key, str(value)))
# args.do_predict = True
predict_results = []
if args.do_predict and args.local_rank in [-1, 0]:
tokenizer = tokenization_albert.FullTokenizer(vocab_file=args.vocab_file,
do_lower_case=args.do_lower_case,
)
# checkpoints_path = os.path.join(args.output_dir, 'checkpoint-4000')
checkpoints = [(0, args.output_dir)]
if args.eval_all_checkpoints:
checkpoints = list(
os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME, recursive=True)))
checkpoints = [(int(checkpoint.split('-')[-1]), checkpoint) for checkpoint in checkpoints if
checkpoint.find('checkpoint') != -1]
checkpoints = sorted(checkpoints, key=lambda x: x[0])
logger.info("Evaluate the following checkpoints: %s", checkpoints)
checkpoints = [checkpoints[-1]]
for _, checkpoint in checkpoints:
global_step = checkpoint.split('-')[-1] if len(checkpoints) > 1 else ""
prefix = checkpoint.split('/')[-1] if checkpoint.find('checkpoint') != -1 else ""
model = AlbertForSequenceClassification.from_pretrained(checkpoint)
model.to(args.device)
result = predict(args, model, tokenizer, prefix=prefix)
predict_results.extend([(k + '_{}'.format(global_step), v) for k, v in result.items()])
output_eval_file = os.path.join(args.output_dir, "checkpoint_eval_results.txt")
with open(output_eval_file, "w") as writer:
for key, value in predict_results:
writer.write("%s = %s\n" % (key, str(value)))