FICHA · AUR

python-flash-attention

Fast and memory-efficient exact attention. CUDA version.

  • CUDA attention kernel library
  • LIBRARY
  • AI
  • HARDWARE
  • Dependency only
official+codex · reviewed · Jun 3, 2026 description in en

Description

Transformer attention layers can run faster and use less GPU memory through CUDA kernels. ML engineers use this library when training or serving large neural networks on compatible hardware. GPU kernels, model data, and benchmark results need validation for the target stack.

Permissions

Permissions not analysed for this source yet.