apache-tvm-ffi<0.2,>=0.1.11
click
cuda-python>=12.0
cuda-tile>=1.4.0
einops
nccl-extensions>=0.1.0
nccl4py>=0.4.1
ninja
numpy
nvidia-cudnn-frontend>=1.31.0
nvidia-cutlass-dsl>=4.6.2a0
nvidia-ml-py
packaging>=24.2
requests
tabulate
torch
tqdm

[b12x]
torch>=2.12.0
nvidia-cutlass-dsl==4.7.1
nvidia-cutlass-dsl-libs-base==4.7.1
nvidia-cutlass-dsl-libs-core==4.7.1
nvidia-cutlass-dsl-libs-cu12==4.7.1
nvidia-cutlass-dsl-libs-cu13==4.7.1
rich>=13
safetensors>=0.6

[cu12]
nvidia-cutlass-dsl>=4.7.0a0

[cu13]
nvidia-cutlass-dsl[cu13]>=4.7.0a0

[nvep]

[ptx]
nvidia-cuda-nvcc>=13.2
apache-tvm-ffi<0.2,>=0.1.12

[sm107]
nvidia-cutlass-dsl[cu13]>=4.8.0.dev0
nvshmem4py-cu13>=0.3.1
torch>=2.8
