Search results

Compiling on Python 3.14 fails due to redefinition errors

Status: Open.

#114 In thu-ml/SpargeAttn;

· blepping opened on Jan 18, 2026

does spas_sage2_attn_meansim_topk_cuda support Deepseek V3.1?

Status: Open.

#113 In thu-ml/SpargeAttn;

· staugust opened on Jan 15, 2026

Random black output when using <code>spas_sage2_attn_meansim_topk_cuda</code>

Status: Open.

#110 In thu-ml/SpargeAttn;

· xuanmingShang opened on Jan 14, 2026

5090 black is not support？

Status: Open.

#109 In thu-ml/SpargeAttn;

· secret-deus opened on Jan 12, 2026

First, I created a sparse attention module for using Flux 1.D which I tested. Some thoughts, questions...

Status: Open.

#107 In thu-ml/SpargeAttn;

· ukaprch opened on Dec 29, 2025

Using the same selection algorithm, but employing non-quantized Flashinfer for computation, why are the generated results all garbled?

Status: Open.

#106 In thu-ml/SpargeAttn;

· hjdjs12 opened on Dec 27, 2025

Possible logical issues in some code within the <code>setup.py</code> file.

Status: Open.

#104 In thu-ml/SpargeAttn;

· junyii37 opened on Dec 25, 2025

windows 下 源码安装的时候 执行link 报错 如何解决？

Status: Open.

#103 In thu-ml/SpargeAttn;

· tailuchen opened on Dec 24, 2025

Does SparseAttn support NVIDIA GB10 architecture?

Status: Open.

#102 In thu-ml/SpargeAttn;

· GarfieldHuang opened on Dec 22, 2025

Please use <code>spas_sage2_attn_meansim_topk_cuda</code> and <code>block_sparse_sage2_attn_cuda</code> APIs

enhancementNew feature or requestNew feature or request

Status: Open.

Feature

#101 In thu-ml/SpargeAttn;

· jt-zhang opened on Dec 16, 2025

Question about TopK

Status: Open.

#99 In thu-ml/SpargeAttn;

· jhl13 opened on Nov 24, 2025

Calling SpargeAttn Twice Corrupts CUDA Context (unspecified launch failure)

Status: Open.

#97 In thu-ml/SpargeAttn;

· sajalmaheshwari624 opened on Nov 19, 2025