LAC实体识别进行增量训练

需要的数据一定得是tsv的格式吗？
我通过下面的代码将空格分开的txt转成tsv，训练时报警WARNING:root:Data type error:
代码：
def write_data(data_, path_):
with open(path_, 'w+', encoding='utf-8') as f:
for each in data_:
f.write('\t'.join(each.split())+'\n')
输入的“词/type“用空格分割，上述代码里转成了\t
然后训练的时候，每条语料进去，都有个WARNING:root:Data type error:这样报错

Start with the provided write_data function and reproduce the warning during incremental LAC named-entity recognition training. Compare the generated TSV records with the training input requirements; done means identifying the accepted format and resolving the reported Data type error for each training record.

Written by the indexing model from the issue text.