Closed wjddd closed 4 weeks ago
It seems that the codebook has out of bounds. Is it because the VQ features were not regenerated from the previous data?
It seems that the codebook has out of bounds. Is it because the VQ features were not regenerated from the previous data?
Hi, maybe it's because the dataset is too small. Now I use a new dataset with more than 100 audio files, and it works.
Self Checks
Cloud or Self Hosted
Self Hosted (Docker)
Steps to reproduce
Docker version: lengyue233/fish-speech:v1.2.1
I followed instructions on: https://speech.fish.audio/zh/finetune/#4-lora
and error occured:
✔️ Expected Behavior
No response
❌ Actual Behavior
No response