Search results

[FEATURE REQUEST] Support for Sparse Attention via attn_indices

Status: Open.

#210 In flash-algo/flash-sparse-attention;

· LoserCheems opened on Nov 14, 2025

[PERFORMANCE] Top‑k windowed mask generation dominates end‑to‑end latency for large sequences

performanceperformance problems or optimizationperformance problems or optimization

Status: Open.

#196 In flash-algo/flash-sparse-attention;

· LoserCheems opened on Oct 22, 2025

[FEATURE REQUEST] Support {num_heads, num_kv_heads, 1} shaped bias in attention functions

featureNew feature requestNew feature request

Status: Open.

#189 In flash-algo/flash-sparse-attention;

· LoserCheems opened on Oct 11, 2025

[FEATURE REQUEST] Support total_k-head broadcastable mask/bias in varlen batch inference

featureNew feature requestNew feature request

Status: Open.

#183 In flash-algo/flash-sparse-attention;

· LoserCheems opened on Oct 3, 2025

[BUG REPORT] INF occurs in backward phrase of the first training step

bugSomething isn't workingSomething isn't working

Status: Open.

#180 In flash-algo/flash-sparse-attention;

· ftgreat opened on Sep 26, 2025

CUDA Illegal Memory Access Error in Flash-DMA during Training

bugSomething isn't workingSomething isn't working

Status: Open.

#169 In flash-algo/flash-sparse-attention;

· helloworld01001 opened on Sep 17, 2025

[FEATURE REQUEST] Unified Sparse Mask Strategy with Block-Level Skipping

featureNew feature requestNew feature request

Status: Open.

#163 In flash-algo/flash-sparse-attention;

· LoserCheems opened on Sep 11, 2025