Skip to content

Actions: saitejasrivilli/flash-attn-from-scratch

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
8 workflow runs
8 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
Add CUTLASS fp16 GEMM benchmark vs cuBLAS on A30 SM80 CI #8: Commit 1e15d5f pushed by saitejasrivilli
1d 0h 0m 13s main
main
1d 0h 0m 13s
Update benchmark results with verified A30 measurements (PyTorch 2.10… CI #7: Commit 64303d9 pushed by saitejasrivilli
1h 44m 33s main
main
1h 44m 33s
Add attention ablation: MHA vs MQA vs GQA vs FlashAttention on perple… CI #6: Commit f7d81fb pushed by saitejasrivilli
1d 0h 0m 15s main
main
1d 0h 0m 15s
Fix flash_attn_bwd.py orphaned constexpr lines, black clean CI #5: Commit 08a51b5 pushed by saitejasrivilli
1d 0h 0m 12s main
main
1d 0h 0m 12s
Add .gitignore, remove compiled artifacts from tracking CI #4: Commit 1bb4d5e pushed by saitejasrivilli
1m 8s main
main
1m 8s
Auto-format with black and isort CI #3: Commit fa1425e pushed by saitejasrivilli
28s main
main
28s
Fix CI: push all source files, soft-fail lint on missing dirs CI #2: Commit 77a250f pushed by saitejasrivilli
1m 38s main
main
1m 38s
Add Flash Attention, int8 GEMM, fused RMSNorm+Linear kernels with A30… CI #1: Commit 27ab842 pushed by saitejasrivilli
10s main
main
10s