PyPI page
Home page
Author:
AlcAI-AnimeHaven
License:
MIT
Summary:
FastFlashAttention: drop-in exact bf16 flash-attention for CUDA with a deterministic backward, tuned for Blackwell (sm_120 / RTX 5090).
Latest version:
0.2.0
Optional dependencies:
flash-attn
|
matplotlib
|
pytest
|
torch
|
triton
Downloads last day:
0
Downloads last week:
2
Downloads last month:
21
7-day smoothing