search

FunAudioLLM / CosyVoice

Multi-lingual large voice generation model, providing inference, training and deployment full-stack ability.
https://funaudiollm.github.io/
Apache License 2.0
6.54k stars 703 forks source link

About "base continue train". #659

Open sunnnnnnnny opened 1 week ago

sunnnnnnnny commented 1 week ago

there are sft training examples;if i want to continue train extra area data from base model, how to do it?

aluminumbox commented 1 week ago

train from base follow same recipe with sft. only data amount is different

sunnnnnnnny commented 1 week ago

Well, thank you for your answer; I understand that the sft model will use spk_embed or utt_embed. If i want to continue training the base model with new data, what key parameters in the configuration file need to be changed?