NER教程中发现bug

[cocoa0409]

hi,
https://aistudio.baidu.com/aistudio/projectdetail/1317771
在这个NER任务中，主模型代码
class BiGRUWithCRF2(nn.Layer): def __init__(self, emb_size, hidden_size, word_num, label_num): super(BiGRUWithCRF2, self).__init__() self.word_emb = TokenEmbedding(extended_vocab_path='./conf/word.dic', unknown_token='OOV') #EMB
TokenEmbedding的利用有误

我查看了源码，extended_vocab_path的参数会作为读取字典，经过_read_vocab_list_from_file取出词表
def _read_vocab_list_from_file(self, extended_vocab_path): # load new vocab table from file vocab_list = [] with open(extended_vocab_path, "r", encoding="utf-8") as f: for line in f.readlines(): vocab = line.rstrip("\n").split("\t")[0] vocab_list.append(vocab) return vocab_list
该任务对应的字典word.dic ，第一列是索引id，不是vocab

所以TokenEmbedding无法正确加载pretrain的权重

[chenxiaozeng]

您好，经研发同学排查，这里确实写错了。预计会在本周修复bug，感谢您的反馈！

[ZeyuChen]

This bug has already fixed.