Skip to content

Actions: HKUSTDial/flash-sparse-attention

Actions

Auto assign reviewers and assignees

Actions

Loading...
Loading

Show workflow options

Create status badge

Loading
auto-assign.yml
86 workflow runs
86 workflow runs

Filter by Event

Filter by Status

Filter by Branch

Filter by Actor

Refactor log_sigmoid function for improved performance and accuracy Auto assign reviewers and assignees #64: Pull request #237 opened by LoserCheems
10s
10s
[FEATURE SUPPORT] Add Configurable Sparse Gate Modes and Adaptive Thresholding in Triton Forward Kernel Auto assign reviewers and assignees #63: Pull request #236 opened by LoserCheems
9s
9s
[FEATURE SUPPORT] Add Triton backward support Auto assign reviewers and assignees #62: Pull request #235 opened by LoserCheems
9s
9s
Refactor qheads_per_kvhead calculations for clarity Auto assign reviewers and assignees #61: Pull request #234 opened by LoserCheems
10s
10s
[BUG FIX] Improve forward kernel configuration and validation Auto assign reviewers and assignees #60: Pull request #233 opened by LoserCheems
7s
7s
Refactor rescaling logic in online_softmax and rescale_o functions Auto assign reviewers and assignees #59: Pull request #232 opened by LoserCheems
9s
9s
Improve performance and configuration for SM90 forward path Auto assign reviewers and assignees #58: Pull request #231 opened by LoserCheems
10s
10s
Renames variables and refactors functions for clarity Auto assign reviewers and assignees #57: Pull request #229 opened by LoserCheems
10s
10s
Improves softmax stability with log2 scaling Auto assign reviewers and assignees #56: Pull request #228 opened by LoserCheems
10s
10s
[FEATURE] Enhance forward combine kernel and split attention Auto assign reviewers and assignees #55: Pull request #227 opened by LoserCheems
9s
9s
[PERFORMANCE OPTIMIZATION] Triton Sparse Base Forward Kernel with Gate-Based Sparsity Auto assign reviewers and assignees #54: Pull request #226 opened by LoserCheems
7s
7s
Add utility functions for device management and input validation Auto assign reviewers and assignees #53: Pull request #225 opened by LoserCheems
7s
7s
[BUG FIX] Improve packed GQA handling Auto assign reviewers and assignees #52: Pull request #224 opened by LoserCheems
10s
10s
[BUG FIX] Refactor block min/max calculations Auto assign reviewers and assignees #51: Pull request #223 opened by LoserCheems
10s
10s
[PERFORMANCE OPTIMIZATION] Flash Sparse Attention Auto assign reviewers and assignees #50: Pull request #221 opened by LoserCheems
9s
9s
Improve numerical stability in sparse attention with sink auxiliary logits Auto assign reviewers and assignees #49: Pull request #220 opened by LoserCheems
10s
10s
Refactor create_mask function parameters Auto assign reviewers and assignees #48: Pull request #218 opened by LoserCheems
9s
9s
Simplify attention mechanisms Auto assign reviewers and assignees #47: Pull request #217 opened by LoserCheems
11s
11s
Improve code readability and linting workflow Auto assign reviewers and assignees #46: Pull request #216 opened by LoserCheems
11s
11s
Add rotary-aware attention modules for improved inference Auto assign reviewers and assignees #45: Pull request #215 opened by LoserCheems
11s
11s
Add gradient computation for bias and token-level KV sparsity support Auto assign reviewers and assignees #44: Pull request #214 opened by LoserCheems
11s
11s
[BUG FIX] Correct causal mask handling for longer KV pairs Auto assign reviewers and assignees #43: Pull request #213 opened by LoserCheems
10s
10s
Corrects issue links in README guides Auto assign reviewers and assignees #42: Pull request #212 opened by LoserCheems
12s
12s
[BUG FIX] Unify masking utilities and improve performance Auto assign reviewers and assignees #41: Pull request #209 opened by LoserCheems
10s
10s
Chore/sync after move Auto assign reviewers and assignees #40: Pull request #208 opened by LoserCheems
12s
12s