这里以BERT恢复模型为例,说明恢复ckpt的两种常见方法
# 方法一:
init_checkpoint = "chinese_L-12_H-768_A-12/bert_model.ckpt" use_tpu = False # 获取模型中所有的训练参数。 tvars = tf.trainable_variables() # 加载BERT模型 (assignment_map, initialized_variable_names) = modeling.get_assignment_map_from_checkpoint(tvars, init_checkpoint) tf.train.init_from_checkpoint(init_checkpoint, assignment_map) tf.logging.info("**** Trainable Variables ****") # 打印加载模型的参数 for var in tvars: init_string = "" if var.name in initialized_variable_names: init_string = ", *INIT_FROM_CKPT*" tf.logging.info(" name = %s, shape = %s%s", var.name, var.shape, init_string) with tf.Session() as sess: sess.run(tf.global_variables_initializer())
# 方法二:
pathname = "chinese_L-12_H-768_A-12/bert_model.ckpt" # 模型地址
bert_config = modeling.BertConfig.from_json_file("chinese_L-12_H-768_A-12/bert_config.json")# 配置文件地址。
configsession = tf.ConfigProto()
configsession.gpu_options.allow_growth = True
sess = tf.Session(config=configsession)
input_ids = tf.placeholder(shape=[64, 128], dtype=tf.int32, name="input_ids")
input_mask = tf.placeholder(shape=[64, 128], dtype=tf.int32, name="input_mask")
segment_ids = tf.placeholder(shape=[64, 128], dtype=tf.int32, name="segment_ids")
with sess.as_default():
model = modeling.BertModel(
config=bert_config,
is_training=True,
input_ids=input_ids,
input_mask=input_mask,
token_type_ids=segment_ids,
use_one_hot_embeddings=False)
saver = tf.train.Saver()
sess.run(tf.global_variables_initializer())# 这里尤其注意,先初始化,在加载参数,否者会把bert的参数重新初始化。这里和demo1是有区别的
saver.restore(sess, pathname)