packaging
wheel
psutil
numpy
torch==2.14.0
apache-tvm-ffi==0.1.13.post3
tokenspeed-proton>=3.8.10.post20260920
tokenspeed-triton>=3.8.10.post20260920
nvidia-cutlass-dsl[cu13]==4.8.0
nvidia-cutlass-dsl-libs-cu13==4.8.0
flashinfer-python==0.7.1rc2
nvidia-cudnn-frontend==1.30.0
nvidia-ml-py==13.580.82
nvtx
tokenspeed-deepep==2.1.0.post20260810
tokenspeed-deepgemm==2.8.0.post20260924
tokenspeed-flashmla==1.0.0.post20260924
tokenspeed-deepselect==1.0.0.post20260910
tokenspeed-mla==0.2.16
tokenspeed-fast-hadamard-transform==1.1.0.post20260920
tokenspeed-fa3==3.0.0.post20260810
tokenspeed-fa4==4.0.0.post20260923
tokenspeed-quack==0.6.4.post20260923
tokenspeed-trtllm-kernel==1.3.0.post20260924
tokenspeed-flashkda==0.0.1.post20260901
tokenspeed-cutedsl-kda==0.1.0.post20260923
