Skip to content

Actions: HKUSTDial/flash-sparse-attention

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
324 workflow runs
324 workflow runs

Filter by Workflow

⌥ + click/return to exclude

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
Merge pull request #296 from HKUSTDial/flex-window Publish docs #38: Commit ddadf6d pushed by LoserCheems
57s main
main
57s
Refactor flash_dec_combine to flash_fwd_combine in forward attention … Auto assign reviewers and assignees #123: Pull request #296 opened by LoserCheems
11s
11s
Merge pull request #295 from HKUSTDial/flex-window Publish docs #37: Commit 320f555 pushed by LoserCheems
1m 1s main
main
1m 1s
Refactor MASK_CAUSAL parameter to use IS_CAUSAL Auto assign reviewers and assignees #122: Pull request #295 opened by LoserCheems
8s
8s
Refactor MASK_CAUSAL parameter to use IS_CAUSAL Auto assign reviewers and assignees #121: Pull request #294 opened by LoserCheems
9s
9s
Merge pull request #293 from HKUSTDial/flex-window Publish docs #36: Commit 3c7c606 pushed by LoserCheems
1m 7s main
main
1m 7s
Update autotune cache system and local attention Auto assign reviewers and assignees #120: Pull request #293 opened by LoserCheems
7s
7s
Revert "local attention and cache system" Auto assign reviewers and assignees #119: Pull request #292 opened by LoserCheems
6s
6s
local attention and cache system Auto assign reviewers and assignees #118: Pull request #291 opened by LoserCheems
7s
7s
Graph Update: pip in /flash_sparse_attn/ops/cute #1381151678 Dependency Graph #1: by dependabot Bot
50s main
main
50s
Rewrite vendored CuTe namespace to flash_sparse_attn.ops.cute Publish docs #35: Commit f12fa37 pushed by LoserCheems
1m 10s main
main
1m 10s
Merge pull request #290 from HKUSTDial/quant Publish docs #34: Commit 7f561a6 pushed by LoserCheems
1m 31s main
main
1m 31s
Quant backward Auto assign reviewers and assignees #117: Pull request #290 opened by LoserCheems
11s
11s
Merge pull request #289 from HKUSTDial/quant Publish docs #33: Commit 8e72e9f pushed by LoserCheems
1m 10s main
main
1m 10s
Add optional output and logsumexp tensors to attention functions Auto assign reviewers and assignees #116: Pull request #289 opened by LoserCheems
14s
14s
Merge pull request #288 from HKUSTDial/quant Publish docs #32: Commit 52a6f4e pushed by LoserCheems
1m 22s main
main
1m 22s
Refactor attention functions and add autotuning support in Triton Auto assign reviewers and assignees #115: Pull request #288 opened by LoserCheems
8s
8s
Merge pull request #287 from HKUSTDial/optime4triton3.7 Publish docs #31: Commit 852114c pushed by LoserCheems
1m 14s main
main
1m 14s
Enhance attention kernel performance and quantization support Auto assign reviewers and assignees #114: Pull request #287 opened by LoserCheems
7s
7s
Merge pull request #286 from HKUSTDial/optime4triton3.7 Publish docs #30: Commit fd0b21b pushed by LoserCheems
59s main
main
59s
FEATURE] Optimize kernel representations for Triton Auto assign reviewers and assignees #113: Pull request #286 opened by LoserCheems
14s
14s
Merge pull request #285 from HKUSTDial/optime4triton3.7 Publish docs #29: Commit 8055c95 pushed by LoserCheems
1m 8s main
main
1m 8s
Refactor activation functions for optimization Auto assign reviewers and assignees #112: Pull request #285 opened by LoserCheems
13s
13s
Revert "Refactor activation functions for performance and consistency" Auto assign reviewers and assignees #111: Pull request #284 opened by LoserCheems
9s
9s
Refactor activation functions for performance and consistency Auto assign reviewers and assignees #110: Pull request #283 opened by LoserCheems
11s
11s