Skip to content

Pull requests: deepseek-ai/DeepGEMM

New pull request New
73 Open 181 Closed
73 Open 181 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

[Bugfix] Reset MegaMoE combine readiness for CUDA graph replay
#451 opened Sep 17, 2026 by foreverrookie Loading…
3
feat(gemm): opt-in labels-contract checker for m-grouped contiguous GEMM
#450 opened Sep 16, 2026 by lucifer1004 Collaborator Loading…
18
[BugFix][sm90] tf32 gemm race condition
#448 opened Sep 15, 2026 by leevan Loading…
6
Integrate main APIs and SM120 kernels while preserving nv_dev features
#447 opened Sep 15, 2026 by lucifer1004 Collaborator Loading…
29
Fix test helper discovery in pytest and sanitizer runner
#446 opened Sep 14, 2026 by lucifer1004 Collaborator Loading…
8
fix(sm90): support parameterized SwiGLU in FP8 MegaMoE
#445 opened Sep 14, 2026 by shiyang814-cpu Loading…
11
Add torch dispatcher op for tf32_hc_prenorm_gemm
#442 opened Sep 14, 2026 by leevan Loading…
10
Add a fused SM90 FP8 MegaMoE: one kernel, a lag-ordered schedule and a ring activation pool
#440 opened Sep 13, 2026 by zkyue Loading…
15
[Bugfix][Layout] Add cuda_fp8.h include to fix __nv_fp8_e4m3 undefined
#439 opened Sep 11, 2026 by XFDG Loading…
1
3
Fix unknown CMake command "cuda_add_library" in CMakeLists.txt
#437 opened Sep 10, 2026 by PerryLink Loading…
4
Improve SM100 MX MQA logits clean-output performance
#436 opened Sep 10, 2026 by OutWhite Loading…
13
fix: derive the SM120 fp8 mqa_logits swizzle mode from head_dim
#434 opened Sep 10, 2026 by aganhui Loading…
5
fix: port the #343 tensormap drain to the SM120 k_grouped NT contiguous kernel
#431 opened Sep 9, 2026 by aganhui Loading…
1
[BUG] Fix sm90 grouped multicast deadlock
#429 opened Sep 8, 2026 by baptiste-bt Loading…
7
fix: reject non-positive block size multiples
#428 opened Sep 7, 2026 by nisbenz Loading…
2
Improve error when cuobjdump is missing or fails
#427 opened Sep 6, 2026 by monperrus Loading…
4
[Bug] Fix bench_kineto returning zero on the first call in a process
#426 opened Sep 2, 2026 by DiegoCao Loading…
4
docs: note Mega MoE requires PyTorch >= 2.9 in Requirements
#424 opened Aug 28, 2026 by pengfaming Loading…
2
docs: fix stale test file link in README
#423 opened Aug 28, 2026 by pengfaming Loading…
2
[Feat] add support sm90 fp8 mega moe
#422 opened Aug 27, 2026 by lengrongfu Loading…
7
fix: make pack_ue8m0_to_int CUDA-graph capture safe (#414)
#421 opened Aug 26, 2026 by XFDG Loading…
8
[SM12x] Restore pure-fp8 1d1d path (removed in 8b1392b); fix kernel TU preamble
#419 opened Aug 26, 2026 by maci0 Loading…
18
[SM100][MegaMoE] Fuse and publish Kimi K3's BF16 shared expert
#416 opened Aug 24, 2026 by gcanlin Loading…
15
Fix SM90 k-grouped GEMM hang and silent corruption when TMA multicast clusters straddle group boundaries
#415 opened Aug 24, 2026 by shoemoney Loading…
3
feat: add SM90 FP8 x MXFP4 MegaMoE support
#411 opened Aug 20, 2026 by huangzhilin-hzl Loading…
13
ProTip! Type g i on any issue or pull request to go back to the issue listing page.