2019-01-29 18:31:51 +08:00
|
|
|
import os
|
2019-01-30 11:39:49 +08:00
|
|
|
import tensorflow as tf
|
|
|
|
|
|
|
|
|
|
tf.logging.set_verbosity(tf.logging.INFO)
|
2019-01-29 18:31:51 +08:00
|
|
|
|
|
|
|
|
file_path = os.path.dirname(__file__)
|
|
|
|
|
|
|
|
|
|
model_dir = os.path.join(file_path, 'chinese_L-12_H-768_A-12/')
|
|
|
|
|
config_name = os.path.join(model_dir, 'bert_config.json')
|
|
|
|
|
ckpt_name = os.path.join(model_dir, 'bert_model.ckpt')
|
|
|
|
|
output_dir = os.path.join(model_dir, '../tmp/result/')
|
|
|
|
|
vocab_file = os.path.join(model_dir, 'vocab.txt')
|
|
|
|
|
data_dir = os.path.join(model_dir, '../data/')
|
|
|
|
|
|
2019-01-30 11:39:49 +08:00
|
|
|
num_train_epochs = 10
|
2019-01-29 18:31:51 +08:00
|
|
|
batch_size = 128
|
2019-01-30 11:39:49 +08:00
|
|
|
learning_rate = 0.00005
|
2019-01-29 18:31:51 +08:00
|
|
|
|
2019-01-30 11:39:49 +08:00
|
|
|
# gpu使用率
|
2019-01-29 18:31:51 +08:00
|
|
|
gpu_memory_fraction = 0.8
|
|
|
|
|
|
2019-01-30 11:39:49 +08:00
|
|
|
# 默认取倒数第二层的输出值作为句向量
|
|
|
|
|
layer_indexes = [-2]
|
2019-01-29 18:31:51 +08:00
|
|
|
|
2019-01-30 11:39:49 +08:00
|
|
|
# 序列的最大程度，单文本建议把该值调小
|
|
|
|
|
max_seq_len = 32
|
