Skip to content

Issue when fine-tune dataset #31

Open
@colafly

Description

@colafly

Thanks for taking time.

I tried fine-tune with my own dataset, following default instruction. However, when I ran inference with the new model, I saw this error:

2024-09-23 22:29:19.347 | INFO | __main__:main:662 - Loading model ... Traceback (most recent call last): model: Union[NaiveTransformer, DualARTransformer] = BaseTransformer.from_pretrained( File "/home/colafly/Repo/fish-speech/fish_speech/models/text2semantic/llama.py", line 325, in from_pretrained config = BaseModelArgs.from_pretrained(str(path)) File "/home/colafly/Repo/fish-speech/fish_speech/models/text2semantic/llama.py", line 77, in from_pretrained data = json.load(f) File "/home/colafly/.pyenv/versions/3.10.14/lib/python3.10/json/__init__.py", line 293, in load return loads(fp.read(), File "/home/colafly/.pyenv/versions/3.10.14/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 64: invalid start byte

any idea what it might be the issue?

I did not modify any of the steps, except reduce batch size to 4 to fit into my GPU. Any help is much appreciated.
python tools/llama/generate.py --text "深呼吸" --checkpoint-path "checkpoints/fish-speech-1.4-yth-lora/model.pth" --num-samples 2 --compile

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions