模型的张量不一致怎么办

[HCl0vo]

(env) ysl@PrimeAI:~/doc/Digital_Life_Server-master$ python SocketServer.py --chatVer 3 --stream False --character paimon --model gpt-3.5-turbo 2023-09-21 11:43:10,596 INFO Initializing Server... 2023-09-21 11:43:10,596 INFO Initializing ASR Service... 2023-09-21 11:43:11,842 INFO Initializing ChatGPT Service... 2023-09-21 11:43:11,842 INFO chatGPT prompt: paimon35.txt 2023-09-21 11:43:11,844 INFO machine identifier: 08:bf:b8:9d:73:06 2023-09-21 11:43:11,882 INFO using custom API proxy, with rate limit. 2023-09-21 11:43:11,996 INFO API Chatbot initialized. 2023-09-21 11:43:11,996 INFO Initializing TTS Service for character_paimon... 2023-09-21 11:43:12,363 ERROR Error(s) in loading state_dict for SynthesizerTrn: size mismatch for enc_p.emb.weight: copying a param with shape torch.Size([52, 192]) from checkpoint, the shape in current model is torch.Size([178, 192]). 2023-09-21 11:43:12,363 ERROR Traceback (most recent call last): File "/home/ysl/doc/Digital_Life_Server-master/SocketServer.py", line 198, in s = Server(args) File "/home/ysl/doc/Digital_Life_Server-master/SocketServer.py", line 88, in init self.tts = TTService.TTService(*self.char_name[args.character]) File "/home/ysl/doc/Digital_LifeServer-master/TTS/TTService.py", line 42, in init = utils.load_checkpoint(model, self.net_g, None) File "/home/ysl/doc/Digital_Life_Server-master/TTS/vits/utils.py", line 40, in load_checkpoint model.load_state_dict(new_state_dict) File "/home/ysl/miniconda3/envs/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 2041, in load_state_dict raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format( RuntimeError: Error(s) in loading state_dict for SynthesizerTrn: size mismatch for enc_p.emb.weight: copying a param with shape torch.Size([52, 192]) from checkpoint, the shape in current model is torch.Size([178, 192]).

Traceback (most recent call last): File "/home/ysl/doc/Digital_Life_Server-master/SocketServer.py", line 203, in raise e File "/home/ysl/doc/Digital_Life_Server-master/SocketServer.py", line 198, in s = Server(args) File "/home/ysl/doc/Digital_Life_Server-master/SocketServer.py", line 88, in init self.tts = TTService.TTService(*self.char_name[args.character]) File "/home/ysl/doc/Digital_LifeServer-master/TTS/TTService.py", line 42, in init = utils.load_checkpoint(model, self.net_g, None) File "/home/ysl/doc/Digital_Life_Server-master/TTS/vits/utils.py", line 40, in load_checkpoint model.load_state_dict(new_state_dict) File "/home/ysl/miniconda3/envs/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 2041, in load_state_dict raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format( RuntimeError: Error(s) in loading state_dict for SynthesizerTrn: size mismatch for enc_p.emb.weight: copying a param with shape torch.Size([52, 192]) from checkpoint, the shape in current model is torch.Size([178, 192]).

[sanqiuu]

我也是这样

[sanqiuu]

解决了 下载项目时加上 --recursive 参数 才能下载完整项目

[qingxuanlin]

能讲一下咋解决这个问题吗？我也遇到了这个问题

[sanqiuu]

能讲一下咋解决这个问题吗？我也遇到了这个问题

重新下载项目 使用这个命令 git clone https://github.com/zixiiu/Digital_Life_Server.git --recursive