PyPI page
Home page
Author:
Tri Dao
License:
BSD 3-Clause License
Summary:
Flash Attention CUTE (CUDA Template Engine) implementation
Latest version:
4.0.0.post20260923
Required dependencies:
apache-tvm-ffi
|
einops
|
nvidia-cutlass-dsl
|
tokenspeed-quack
|
torch
|
torch-c-dlpack-ext
|
typing-extensions
Optional dependencies:
nvidia-cutlass-dsl
|
pytest
|
pytest-xdist
|
ruff
Downloads last day:
57
Downloads last week:
400
Downloads last month:
1,989
7-day smoothing