Closed skyeKing closed 1 year ago
基于PTTS-basemodel微调训练的时候报错，代码跟ipy上一样
File ~/anaconda3/envs/modelscope_py38/lib/python3.8/site-packages/modelscope/trainers/audio/tts_trainer.py:242, in KanttsTrainer.train(self, *args, **kwargs) 232 dir_dict = { 233 'work_dir': self.work_dir, 234 'am_tmp_dir': self.am_tmp_dir, 235 'voc_tmp_dir': self.voc_tmp_dir, 236 'data_dir': self.data_dir 237 } 238 config_dict = { 239 'am_config': self.am_config_path, 240 'voc_config': self.voc_config_path 241 } --> 242 self.model.train(self.speaker, dir_dict, self.train_type, config_dict,
size mismatch for mel_decoder.mel_dec.pnca.7.pnca_attn.w_h_kv.weight: copying a param with shape torch.Size([256, 640]) from checkpoint, the shape in current model is torch.Size([256, 320])
基于PTTS-basemodel微调训练的时候报错，代码跟ipy上一样
File ~/anaconda3/envs/modelscope_py38/lib/python3.8/site-packages/modelscope/trainers/audio/tts_trainer.py:242, in KanttsTrainer.train(self, *args, **kwargs) 232 dir_dict = { 233 'work_dir': self.work_dir, 234 'am_tmp_dir': self.am_tmp_dir, 235 'voc_tmp_dir': self.voc_tmp_dir, 236 'data_dir': self.data_dir 237 } 238 config_dict = { 239 'am_config': self.am_config_path, 240 'voc_config': self.voc_config_path 241 } --> 242 self.model.train(self.speaker, dir_dict, self.train_type, config_dict,
size mismatch for mel_decoder.mel_dec.pnca.7.pnca_attn.w_h_kv.weight: copying a param with shape torch.Size([256, 640]) from checkpoint, the shape in current model is torch.Size([256, 320])