Search results

[Info Sharing] Success Combination for Blackwell (RTX5090) / NO BUILD TIME (using whl)

Status: Open.

#2016 In Dao-AILab/flash-attention;

· MinChoi0129 opened on Nov 15, 2025

Can some computations be masked out by setting cache seqlens?

Status: Open.

#2015 In Dao-AILab/flash-attention;

· nannaer opened on Nov 15, 2025

Is fa v4 cute dsl, not directly compatible with torch.compile?

Status: Open.

#2013 In Dao-AILab/flash-attention;

· michael604work opened on Nov 15, 2025

Block sparse attention support for backward pass

Status: Open.

#2011 In Dao-AILab/flash-attention;

· raunaks13 opened on Nov 13, 2025

FlashAttention2 fails to build on CUDA 12.6 + PyTorch 2.6 (symbol not defined linker errors)

Status: Open.

#2010 In Dao-AILab/flash-attention;

· snow-like-kk opened on Nov 13, 2025

FlashAttention-3 Windows + CUDA 13.0 build fails with C2719 “formal parameter with requested alignment of 128 won’t be aligned”

Status: Open.

#2009 In Dao-AILab/flash-attention;

· windreamer opened on Nov 13, 2025

What would be the correct way to install flash-attn for PyTorch 2.9.1+cu130

Status: Open.

#2008 In Dao-AILab/flash-attention;

· n0valis opened on Nov 12, 2025

FA3 support for GB200

Status: Open.

#2003 In Dao-AILab/flash-attention;

· depksingh opened on Nov 11, 2025

Why is torch2.9.0 not supported?

Status: Open.

#2002 In Dao-AILab/flash-attention;

· manglu097 opened on Nov 11, 2025

使用pip install flash_attn==2.7.4.post1 --no-build-isolation，构建失败，请问是什么原因，以下是报错信息

Status: Open.

#2001 In Dao-AILab/flash-attention;

· sanshi9523 opened on Nov 11, 2025

[cute, bwd, sm100] support for head_dim = 64

Status: Open.

#2000 In Dao-AILab/flash-attention;

· XiaomingXu1995 opened on Nov 11, 2025

[Bug] Cannot run cute tests due to version issue

Status: Open.

#1997 In Dao-AILab/flash-attention;

· Edenzzzz opened on Nov 9, 2025