def bilstm_crf_train_eval(config,
import_model,
train_examples,
dev_examples=None,
test_examples=None):
processor = DataProcessor(config.data_dir, config.do_lower_case)
word2id = processor.get_vocab()
config.vocab_size = len(word2id)
train_features = convert_examples_to_features_crf(
examples=train_examples,
word2id=word2id,
label_list=config.label_list,
)
train_dataset = BuildDataSet(train_features)
train_loader = DataLoader(train_dataset,
batch_size=config.batch_size,
shuffle=True)
if dev_examples:
dev_features = convert_examples_to_features_crf(
examples=dev_examples,
word2id=word2id,
label_list=config.label_list,
)
dev_dataset = BuildDataSet(dev_features)
dev_loader = DataLoader(dev_dataset,
batch_size=config.batch_size,
shuffle=True)
else:
dev_loader = None
if test_examples:
test_features = convert_examples_to_features_crf(
examples=test_examples,
word2id=word2id,
label_list=config.label_list,
)
test_dataset = BuildDataSet(test_features)
test_loader = DataLoader(test_dataset,
batch_size=config.batch_size,
shuffle=False)
else:
test_loader = None
logger.info("self config:\n {}".format(config_to_json_string(config)))
model = import_model.Model(config).to(config.device)
best_model = model_train(config,
model,
train_loader,
dev_loader,
to_crf=True)
model_test(config, best_model, test_loader, to_crf=True)
def train_dev_test(
config,
model,
tokenizer,
train_data=None,
dev_data=None,
test_examples=None,
):
dev_acc = 0.
dev_f1 = 0.
predict_label = []
# 加载模型
model_example = copy.deepcopy(model).to(config.device)
best_model = None
convert_to_features, build_data_set, data_loader = MODEL_CLASSES[config.use_model]
if train_data:
config.train_num_examples = len(train_data)
# 特征转化
train_features = convert_to_features(
examples=train_data,
tokenizer=tokenizer,
label_list=config.class_list,
max_length=config.pad_size,
to_pair=config.to_pair,
speaker_dict=config.speaker_dict
)
train_dataset = build_data_set(train_features, config=config)
train_loader = data_loader(train_dataset, device=config.device, batch_size=config.batch_size, shuffle=True)
# dev 数据加载与转换
if dev_data is not None:
config.dev_num_examples = len(dev_data)
dev_features = convert_to_features(
examples=dev_data,
tokenizer=tokenizer,
label_list=config.class_list,
max_length=config.pad_size,
to_pair=config.to_pair,
speaker_dict=config.speaker_dict
)
dev_dataset = build_data_set(dev_features, config=config)
dev_loader = data_loader(dev_dataset, device=config.device, batch_size=config.batch_size, shuffle=True)
else:
dev_loader = None
best_model = model_train(config, model_example, train_loader, dev_loader)
if dev_data is not None:
dev_acc, dev_f1 = model_metrics(config, best_model, dev_loader)
if test_examples is not None or dev_data is not None:
if test_examples is None:
test_examples = dev_data
test_features = convert_to_features(
examples=test_examples,
tokenizer=tokenizer,
label_list=config.class_list,
max_length=config.pad_size,
to_pair=config.to_pair,
speaker_dict=config.speaker_dict
)
test_dataset = build_data_set(test_features, config=config)
test_loader = data_loader(test_dataset, device=config.device, batch_size=config.batch_size, shuffle=False)
predict_label = model_evaluate(config, model_example, test_loader, test=True)
return best_model, (dev_acc, dev_f1), predict_label