This is the repository card of kernels-community/flash-attn2 that has been pushed on the Hub. It was built to be used with the kernels library. This card was automatically generated.

How to use

# make sure `kernels` is installed: `pip install -U kernels`
from kernels import get_kernel

kernel_module = get_kernel("kernels-community/flash-attn2", version=3)
flash_attn_func = kernel_module.flash_attn_func

flash_attn_func(...)

Available functions

  • flash_attn_func
  • flash_attn_kvpacked_func
  • flash_attn_qkvpacked_func
  • flash_attn_varlen_func
  • flash_attn_varlen_kvpacked_func
  • flash_attn_varlen_qkvpacked_func
  • flash_attn_with_kvcache
  • fwd
  • varlen_fwd
  • bwd
  • varlen_bwd
  • fwd_kvcache

Benchmarks

Benchmarking script is available for this kernel. Run kernels benchmark kernels-community/flash-attn2 --version 3.

Downloads last month
6,281
bsd-3-clause
Supported hardwares new
XPU
Intel Arc
Arc A380
6GB
Intel Arc
Arc A750
8GB
Intel Arc
Arc A770
8GB
Intel Arc
Arc B570
10GB
Intel Arc
Arc B580
12GB
Intel Arc
Arc B50
16GB
Intel Arc
Arc B60
24GB
Intel Arc
Arc Pro B70
32GB
OS
linux
Arch
x86_64
Kernel Builder
81580bb