# Install a CUDA-enabled PyTorch build matching your host first.
# The tested worker used torch 2.10.0+cu128 on Linux.
transformers==5.17.0
peft==0.21.0
accelerate==1.15.0
Pillow==12.3.0
fastapi==0.141.1
starlette==1.6.0
pydantic==2.13.5
uvicorn>=0.30,<1
huggingface_hub>=1.32,<2
triton>=3.7.1
flash-linear-attention[cuda]