fish-speech-v1.5 finetuned model can't speak clearly.

[dawenxi-only]

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

Self Hosted (Docker)

Environment Details

I use the docker image pulled by "docker pull fishaudio/fish-speech:latest-dev"
My gpu is V100S,and driver version is 535.104.12, cuda version is 12.2

Steps to Reproduce

according to finetune.md, I use 1000 audio to finetune model.
I test the new model merged lora weight, including 1 step, 100 step, 1000 step,and 10000 step, the synthesized audio cannot speak clearly.
here is an example file:

✔️ Expected Behavior

I expect finetuned model can speak better than zero-shot.

❌ Actual Behavior

finetuned model can't speak clearly

[dawenxi-only]

My mistake.
When using new model for inference, new prompt token should be re generate using tools/vqgan/inference.py.
Thanks