[FINETUNE]

[kunci115]

I want to finetune it with around 30mins of english singaporean accent data, but when the models was merge with lora after trained, i test the model with "hello my name is laura", but the model generate longer than that with french language, can anyone tell what happen ?

this is to train following https://speech.fish.audio/en/finetune/ `` python fish_speech/train.py --config-name text2semantic_finetune project=$project [email protected]_config=r_8_alpha_16

[PoTaTo-Mika]

30mins' data is so little for finetune, we recommend you to finetune with data more than 10 hours.

[kunci115]

ok so the bare minimum to fitune is 11 hours I assume ? also my dataset format is in .lab like this inside 0.00 0.50 hello \n 0.51 1.00 world is that correct or no?

[leng-yue]

The text in dataset should be hello world directly.