ProHiryu / albert-chinese-ner

使用预训练语言模型ALBERT做中文NER
MIT License
467 stars 124 forks source link

是否需要把输入语料处理成小于max_seq_length #29

Closed agilelab closed 4 years ago

agilelab commented 4 years ago

使用自己的语料训练报错 Dimensions must be equal, but are 45 and 11 for 'loss/mul' (op: 'Mul') with input shapes: [?,128,45], [?,128,11]. 是否需要把输入语料处理成小于max_seq_length的句子。