Skip to content

Pull requests: pytorch/helion

New pull request New
123 Open 2,460 Closed
123 Open 2,460 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

[autotuner] Preserve arg layout in _clone_args CLA Signed This label is managed by the Meta Open Source bot.
#2867 opened Jun 26, 2026 by tarinduj Contributor Loading…
5
[autotuner] pointwise seed heuristic for pure elementwise kernels CLA Signed This label is managed by the Meta Open Source bot.
#2866 opened Jun 26, 2026 by calebmkim Contributor Loading…
Pretuned kernels nightly dashboard CLA Signed This label is managed by the Meta Open Source bot.
#2865 opened Jun 26, 2026 by yushangdi Contributor • Draft
1
[pretuned_kernels] benchmark scaled_mm under CUDA graphs CLA Signed This label is managed by the Meta Open Source bot.
#2863 opened Jun 25, 2026 by yushangdi Contributor Loading…
autotuner: record structured hardware_info in the autotune dataset CLA Signed This label is managed by the Meta Open Source bot.
#2862 opened Jun 24, 2026 by IshanAryendu Contributor Loading…
autotuner: record per-config perf_stats in the autotune dataset CLA Signed This label is managed by the Meta Open Source bot.
#2861 opened Jun 24, 2026 by IshanAryendu Contributor Loading…
[autotuner] allow block sizes to overshoot small tiled dimensions CLA Signed This label is managed by the Meta Open Source bot.
#2856 opened Jun 24, 2026 by yushangdi Contributor Loading…
2
[cute] Drop dead scalar-fallback K-loop branch for flat single padded-M tile CLA Signed This label is managed by the Meta Open Source bot.
#2854 opened Jun 23, 2026 by yushangdi Contributor • Draft
[cute] Coalesced TMA store for padded-M tcgen05 tiles with m_size>=32 CLA Signed This label is managed by the Meta Open Source bot.
#2853 opened Jun 23, 2026 by yushangdi Contributor • Draft
[cute] Vectorize padded-M aux (scale) load in tcgen05 store epilogue CLA Signed This label is managed by the Meta Open Source bot.
#2852 opened Jun 23, 2026 by yushangdi Contributor • Draft
[autotuner] composed-fact seed for fused matmul + reduction-epilogue CLA Signed This label is managed by the Meta Open Source bot.
#2846 opened Jun 23, 2026 by calebmkim Contributor Loading…
3
[cute] Enable N-axis cluster (cluster_n=2, cluster_m=1) A-multicast for tiny-M fp8 CLA Signed This label is managed by the Meta Open Source bot.
#2845 opened Jun 23, 2026 by yushangdi Contributor • Draft
[cute] Leaner masked store for single padded-M tcgen05 tile CLA Signed This label is managed by the Meta Open Source bot.
#2844 opened Jun 23, 2026 by yushangdi Contributor • Draft
[cute] Allow block_m > static_m with padded-M tcgen05 for tiny-M fp8/bf16 GEMM CLA Signed This label is managed by the Meta Open Source bot.
#2843 opened Jun 23, 2026 by yushangdi Contributor • Draft
[cute] Fix codegen crash for fp8 matmul when static M < 64 CLA Signed This label is managed by the Meta Open Source bot.
#2840 opened Jun 22, 2026 by yushangdi Contributor • Draft
[autotuner] specialize the M-reduction seeds via per_feature_accumulator (occupancy + byte-cap) CLA Signed This label is managed by the Meta Open Source bot.
#2830 opened Jun 20, 2026 by calebmkim Contributor Loading…
[autotuner] route the 6 backward M-reduction kernels onto the standard/user-tiled tracks CLA Signed This label is managed by the Meta Open Source bot.
#2829 opened Jun 20, 2026 by calebmkim Contributor Loading…
[autotuner] reduction seed: budgeted r_block + liveness-aware persistent/looped decision CLA Signed This label is managed by the Meta Open Source bot.
#2828 opened Jun 20, 2026 by calebmkim Contributor Loading…
[bench] Add --fp8-gemm-native-shapes to benchmarks/run.py CLA Signed This label is managed by the Meta Open Source bot.
#2814 opened Jun 18, 2026 by yushangdi Contributor • Draft
autotuner: record per-config generated source (Triton/Metal/Cute/Pallas) in the autotune dataset CLA Signed This label is managed by the Meta Open Source bot.
#2809 opened Jun 17, 2026 by IshanAryendu Contributor Loading…
[cute] Enable tcgen05 tensor cores for small-M fp8 GEMM (M<64) CLA Signed This label is managed by the Meta Open Source bot.
#2807 opened Jun 17, 2026 by yushangdi Contributor • Draft
autotuner: capture per-run device IR in the autotune dataset CLA Signed This label is managed by the Meta Open Source bot.
#2796 opened Jun 16, 2026 by IshanAryendu Contributor Loading…
6
autotuner: add device-IR extractor for cost-model dataset CLA Signed This label is managed by the Meta Open Source bot.
#2794 opened Jun 16, 2026 by IshanAryendu Contributor Loading…
30
[autotuner] mem_op_id: cache-robustness (structural fingerprint + AOT length guard) CLA Signed This label is managed by the Meta Open Source bot.
#2790 opened Jun 15, 2026 by calebmkim Contributor • Draft
[autotuner] mem_op_id: id-keyed per-memory-op tunable slots CLA Signed This label is managed by the Meta Open Source bot.
#2789 opened Jun 15, 2026 by calebmkim Contributor • Draft
ProTip! Find all pull requests that aren't related to any open issues with -linked:issue.