torch>=2.7
triton
numpy
safetensors
jinja2
pyelftools
nvidia-ml-py
cuda-bindings
tqdm
tabulate

[cu12]
nvidia-cuda-runtime-cu12
nvidia-cuda-cccl-cu12
nvidia-cuda-nvcc-cu12
nvidia-cuda-nvrtc-cu12

[cu13]
nvidia-cuda-runtime
nvidia-cuda-cccl
nvidia-cuda-nvcc
nvidia-cuda-nvrtc
