[BUG]

[ines-gpt]

when excute the command python tools/llama/generate.py \ --text "要转换的文本" \ --prompt-text "你的参考文本" \ --prompt-tokens "fake.npy" \ --checkpoint-path "checkpoints/text2semantic-sft-medium-v1.1-4k.pth" \ --num-samples 2 \ --compile the following error happened

**checkpoints/text2semantic-sft-medium-v1.1-4k.pth Traceback (most recent call last): File "/home/ysl/00_work/01_ines/01_rsd/02_fish_speech/tools/llama/generate.py", line 713, in main() File "/home/ysl/miniconda3/envs/fish-speech/lib/python3.10/site-packages/click/core.py", line 1157, in call return self.main(args, kwargs) File "/home/ysl/miniconda3/envs/fish-speech/lib/python3.10/site-packages/click/core.py", line 1078, in main rv = self.invoke(ctx) File "/home/ysl/miniconda3/envs/fish-speech/lib/python3.10/site-packages/click/core.py", line 1434, in invoke return ctx.invoke(self.callback, ctx.params) File "/home/ysl/miniconda3/envs/fish-speech/lib/python3.10/site-packages/click/core.py", line 783, in invoke return __callback(args, **kwargs) File "/home/ysl/00_work/01_ines/01_rsd/02_fish_speech/tools/llama/generate.py", line 660, in main model, decode_one_token = load_model( File "/home/ysl/00_work/01_ines/01_rsd/02_fish_speech/tools/llama/generate.py", line 344, in load_model model: Union[NaiveTransformer, DualARTransformer] = BaseTransformer.from_pretrained( File "/home/ysl/00_work/01_ines/01_rsd/02_fish_speech/fish_speech/models/text2semantic/llama.py", line 325, in from_pretrained config = BaseModelArgs.from_pretrained(path) File "/home/ysl/00_work/01_ines/01_rsd/02_fish_speech/fish_speech/models/text2semantic/llama.py", line 77, in from_pretrained data = json.load(f) File "/home/ysl/miniconda3/envs/fish-speech/lib/python3.10/json/init.py", line 293, in load return loads(fp.read(), File "/home/ysl/miniconda3/envs/fish-speech/lib/python3.10/codecs.py", line 322, in decode (result, consumed) = self._buffer_decode(data, self.errors, final) UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 128: invalid start byte

[PoTaTo-Mika]

非常抱歉，文档此处有些小问题，--checkpoint-path应该为 "checkpoints/fish-speech-1.2/model.pth" Sorry for the bug in docs, the args --checkpoint-path should be "checkpoints/fish-speech-1.2/model.pth"

[ines-gpt]

Thank you very much for your quickly reply, just now

1. update the source code,
2. download the model again with [huggingface-cli download fishaudio/fish-speech-1.2 model.pth --local-dir checkpoints]
3. excute that command again, but the same error come out error happened place is [fish_speech/models/text2semantic/llama.py", line 77, in from_pretrained data = json.load(f)] By the way the parameter [ --config-name ] can not be used in current source [ tools/llama/generate.py]

[PoTaTo-Mika]

We are trying our best to fix the bugs, if you want to use the inferrence part, try use the command below: activate your virtual environment cd /your_path_to_the_object/fish-speech python tools/webui.py

[ines-gpt]

thank you very much! I found that the parameter[heckpoint-path "checkpoints/model.pth" should be folder [checkpoints], then source will load the config.json and model.pth,

1. I download the config file by [huggingface-cli download fishaudio/fish-speech-1.2 config.json --local-dir checkpoints]

2. excute the command again, 　and when loading mode.pth, the flow error ocurred.

   File "/home/ysl/00_work/01_ines/01_rsd/02_fish_speech/tools/llama/generate.py", line 656, in main model, decode_one_token = load_model( File "/home/ysl/00_work/01_ines/01_rsd/02_fish_speech/tools/llama/generate.py", line 340, in load_model model: Union[NaiveTransformer, DualARTransformer] = BaseTransformer.from_pretrained( File "/home/ysl/00_work/01_ines/01_rsd/02_fish_speech/fish_speech/models/text2semantic/llama.py", line 341, in from_pretrained tokenizer = AutoTokenizer.from_pretrained(str(path)) File "/home/ysl/miniconda3/envs/fish-speech/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py", line 846, in from_pretrained config = AutoConfig.from_pretrained( File "/home/ysl/miniconda3/envs/fish-speech/lib/python3.10/site-packages/transformers/models/auto/configuration_auto.py", line 984, in from_pretrained raise ValueError( ValueError: The checkpoint you are trying to load has model typedual_arbut Transformers does not recognize this architecture. This could be because of an issue with the checkpoint, or because your version of Transformers is out of date.

[PoTaTo-Mika]

The bugs about path has been fixed in the pr, you can wait for merge and then pull the main branch again :) Also remember to check the version of your Transformers. Try: pip install --upgrade transformers If the problem still exists, we suggest you to wait for the pull requests to be merged.

[ines-gpt]

from [https://huggingface.co/fishaudio/fish-speech-1.2/tree/main], download all the file ,and it become good, execute successfully. thank you

[ines-gpt]

resolve the problem, and close it.