Benchmarks uploaded using `kernels`.

Build uploaded using `kernels`.

feat: add benchmark file

Build uploaded using `kernels`.

Build uploaded using `kernels`.

Build uploaded using `kernels`.

Fix repo ID in example

Build uploaded using `kernels`.

Fix kernel module name

Update tags

Build uploaded using `kernels`.

Build uploaded using `kernels`.

fix: bump build for torch 2.8

feat: bump api on readme

feat: bump build

fix: add readme example script

fix: bump readme

fix: refactor bindings and update py library for correctness

fix: add warning for accuracy issue

fix: adjust comma typos

fix: simplify and align signatures further

feat: bump builds

fix: align kernel source with latest reference source

feat: add tag for hfjob build

Add license and tag metadata

Build (aarch64)

Build

Sync capabilities with upstream

Update flake

feat improve readme and library code

fix: update library for all ops

Update build/torch26-cxx98-cu124-x86_64-linux/flash_attn/__init__.py

feat: bump build

fix: revise bindings and wrapper typing

feat: build for all caps 8 and up

feat: add remaining torch binding def and impls

fix: adjust types

fix: expand build combinations and include all files

fix: limit build head size and cuda caps for dev

feat: pass vars into fwd and include build

feat: include source and enable build

initial commit