V1.5 Finetune bug

[Charles679]

Self Checks

• This template is only for bug reports. For questions, please visit Discussions.
• I have thoroughly reviewed the project documentation (installation, training, inference) but couldn't find information to solve my problem. English 中文 日本語 Portuguese (Brazil)
• I have searched for existing issues, including closed ones. Search issues
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
• Please do not modify this template and fill in all required fields.

Cloud or Self Hosted

Cloud

Environment Details

Ubuntu 22.04.4 LTS , python 3.10.12 , torch 2.4.1

Steps to Reproduce

I tried to finetune fish-speech 1.5 according to the doc https://speech.fish.audio/finetune/ . Everything went well before the last step.

4. Finally, fine-tuning with LoRA

I downloaded LLAMA weights using the command
huggingface-cli download fishaudio/fish-speech-1.5 --local-dir checkpoints/fish-speech-1.5

And then I started the fine-tuning by running the following command:
python fish_speech/train.py --config-name text2semantic_finetune \ project=$project \ [email protected]_config=r_8_alpha_16

✔️ Expected Behavior

No bugs

❌ Actual Behavior

It is said that
requests.exceptions.HTTPError: 404 Client Error: Not Found for url: https://huggingface.co/checkpoints/fish-speech-1.4/resolve/main/tokenizer_config.json

But I was training using fish-speech-1.5 at that time. I don't know how to fix the bug.Could you give me a solution to complete the finetune?

[Charles679]

I found a similar issue and tried to replace all the 1.4 in the project to 1.5.The problem seems solved

[gokul-eloop]

@Charles679 how was the output after finetune? please share your experience.

[Charles679]

I gave up finetune on version 1.5. After I solved the problem I mentioned above ,I came across some other problems. I also found an issue similar to the new problem. It is said that finetune on 1.5 is not supported yet. So i'm currently trying finetune on 1.4.

[gokul-eloop]

Cool, thanks. Thats my understanding as well. Thanks for confirmation.