copying a param with shape torch.Size([128, 512]) from checkpoint, the shape in current model is torch.Size([128, 1024])

开始的报错是weight: copying a param with shape torch.Size([70, 512]) from checkpoint, the shape in current model is torch.Size([75, 512])

之后按issue 37改
synthesizer/utils/symbols.py 第11行的内容 改为：
_characters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz12340!\'(),-.:;? '

之后报错变成
copying a param with shape torch.Size([128, 512]) from checkpoint, the shape in current model is torch.Size([128, 1024])

Start with synthesizer/utils/symbols.py at line 11 and compare the symbol change described in the issue with issue 37. Then trace the checkpoint and current-model configuration that produces the reported torch.Size mismatch. Done means the reported checkpoint loads without the parameter shape error, with the intended symbol configuration preserved.

Written by the indexing model from the issue text.