python – Tensorflow tf.constant_initializer非常慢
发布时间:2020-12-20 13:17:54 所属栏目:Python 来源:网络整理
导读:尝试使用100 dim的训练有素的word2vec嵌入来训练LSTM @staticmethoddef load_embeddings(pre_trained_embeddings_path,word_embed_size): embd = [] import time start_time = time.time() cnt = 4 with codecs.open(pre_trained_embeddings_path,mode="r",e
尝试使用100 dim的训练有素的word2vec嵌入来训练LSTM
@staticmethod def load_embeddings(pre_trained_embeddings_path,word_embed_size): embd = [] import time start_time = time.time() cnt = 4 with codecs.open(pre_trained_embeddings_path,mode="r",encoding='utf-8') as f: for line in f.readlines(): values = line.strip().split(' ') embd.append(values[1:]) cnt += 1 if cnt % 100000 == 0: print("word-vectors loaded: %d" % cnt) embedding,vocab_size,embed_dim = embd,len(embd),len(embd[0]) load_end_time = time.time() print("word vectors loaded from and start initialising,cnt: %d,time taken: %d secs " % (vocab_size,load_end_time - start_time)) embedding_init = tf.constant_initializer(embedding,dtype=tf.float16) src_word_embedding = tf.get_variable(shape=[vocab_size,embed_dim],initializer=embedding_init,trainable=False,name='word_embedding',dtype=tf.float16) print("word-vectors loaded and initialised,time taken: %d secs" % (vocab_size,time.time() - load_end_time)) return src_word_embedding 运行此方法时输出的结果如下: word vectors loaded from and start initialising,cnt: 2419080,time taken: 74 secs word-vectors loaded and initialised,time taken: 1647 secs 系统信息:tensorflow 1.1.0,tcmalloc,python 3.6,ubuntu 14.04 HALF一小时初始化似乎很慢或是正常行为?知道可能是什么问题还是有问题? 更新:使用@sirfz方法提供嵌入使得加载嵌入的速度非常快,初始化完成时间为85秒 解决方法
将大常量加载到图形中不仅速度较慢,而且还会泄漏大量内存.我有一个类似的问题
I reported not long ago和我最好的解决方法是:
# placeholder for loading your saved embeddings embedding_init = tf.placeholder(tf.float16,shape=[vocab_size,embed_dim]) src_word_embedding = tf.get_variable(initializer=embedding_init,dtype=tf.float16) # run initialization with the value of embeddings placeholder session.run(tf.global_variables_initializer(),feed_dict={embedding_init: embedding}) (编辑:李大同) 【声明】本站内容均来自网络,其相关言论仅代表作者个人观点,不代表本站立场。若无意侵犯到您的权利,请及时与联系站长删除相关内容! |