想要用pre-trained模型重新训练时，出现:
RuntimeError: Error(s) in loading state_dict for BIGLM:
size mismatch for tok_embed.weight: copying a param with shape torch.Size([28781, 768]) from checkpoint, the shape in current model is torch.Size([6410, 768]).
的错误，个人推测似乎是vocab.txt上传到songci的版本，想请问怎么处理?
hu0810
commented
1 year ago
想要用pre-trained模型重新训练时，出现: RuntimeError: Error(s) in loading state_dict for BIGLM: size mismatch for tok_embed.weight: copying a param with shape torch.Size([28781, 768]) from checkpoint, the shape in current model is torch.Size([6410, 768]). 的错误，个人推测似乎是vocab.txt上传到songci的版本，想请问怎么处理?