datasets[audio]>=3.6.0
peft>=0.16.0
pydantic>=2.11.7
pydantic-settings[yaml]>=2.10.1
transformers[torch,torch-speech]>=4.56.2
typing-extensions>=4.14.1

[modelscope]
modelscope>=1.29.1

[quantization]
bitsandbytes>=0.47.0
