Khanin Udomchoksakul
Add optimized Mistral-7B-Instruct-v0.3 build (MistralRMSNorm + MistralMLP CUDA kernels)
e57527b verified
Raw
History Blame Contribute Delete
126 Bytes
torch==2.11.0
--extra-index-url https://download.pytorch.org/whl/130
accelerate
datasets
peft
trl
deepspeed
setuptools<82.0.0