This is the repository card of kernels-community/flash-attn3 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-attn3", version=2)
add_op_namespace_prefix = kernel_module.add_op_namespace_prefix

add_op_namespace_prefix(...)

Available functions

  • add_op_namespace_prefix
  • flash_attn_combine
  • flash_attn_func
  • flash_attn_qkvpacked_func
  • flash_attn_varlen_func
  • flash_attn_with_kvcache
  • get_scheduler_metadata
  • ops

Benchmarks

Benchmarking script is available for this kernel. Run kernels benchmark kernels-community/flash-attn3 --version 2.

Downloads last month
1,623
bsd-3-clause