Training Llasa-3B for French
#21
by
Delfshkrimm
- opened
Hi! How would one tackle adding French (or any other) language support to the model? Finetuning the model? What kind of dataset would be needed? What's the best approach to do so? I have a 100GB VRAM cluster of GPUs (mostly RTX 3070s 8GB) available for training.
Delfshkrimm
changed discussion title from
Llasa-3B for French TTS
to Training Llasa-3B for French
You can finetune this one https://huggingface.co/HKUSTAudio/Llasa-1B-Multilingual A detailed fine-tuning instruction will be updated soon. https://github.com/zhenye234/LLaSA_training
Thank you! Already tested the current one and it's really good already :) Will also try the fine tuning when instructions are ready! Congratz for the amazing work