Skip to content

Pull requests: RL-Align/RL-Kernel

New pull request New
31 Open 210 Closed
31 Open 210 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

feat(ascend): Qwen-Image qk_rmsnorm & multi_axis_rope kernels (issue #386)
#411 opened Sep 13, 2026 by erfgss Contributor Loading…
4 of 7 tasks
1
Qwen image latent pack unpack multimodal Features, bugs, or optimizations specific to multimodal support.
#410 opened Sep 12, 2026 by nodeeeeee Loading…
1
style: format Python sources and benchmark results
#408 opened Sep 12, 2026 by maxiaosong1124 Collaborator Loading…
1
Add fixed-order MoE shared/residual merge reference deepseek-P6 DSv4 platform: cuda Specific optimizations or bugs in NVIDIA graphics cards (such as FlashInfer, TMA optimizations)
#404 opened Sep 11, 2026 by AsterWang Loading…
1
[DSv4][P5-4] MXFP8×MXFP4 Routed Expert grouped GEMM: strict CUDA kernel deepseek-P5 DSv4 platform: cuda Specific optimizations or bugs in NVIDIA graphics cards (such as FlashInfer, TMA optimizations)
#399 opened Sep 10, 2026 by Bignonia7 Loading…
2
feat(musa): add native deterministic gemm kernel MUSA
#395 opened Sep 9, 2026 by Arlo-mt Collaborator Loading…
@Arlo-mt
4
Musa support native fused logp kernel MUSA
#392 opened Sep 8, 2026 by Arlo-mt Collaborator Loading…
@Arlo-mt
9
[CI] Fix black formatting on main and pin line-length in pyproject.toml
#384 opened Sep 3, 2026 by Dnoob Contributor Loading…
1
4
[DSv4][P1-S0] Start kit for the P1 work package (mHC + RMSNorm) deepseek-P1 DSv4 platform: cuda Specific optimizations or bugs in NVIDIA graphics cards (such as FlashInfer, TMA optimizations)
#383 opened Sep 3, 2026 by zhangj1an Collaborator Loading…
10 tasks
7
feat: enable Triton kernels on MUSA MUSA
#375 opened Sep 1, 2026 by Arlo-mt Collaborator Loading…
@Arlo-mt
6
[DSv4][P5-0] Start kit for the P5 work package (MXFP4 Routed Expert + LoRA + Shared Expert) deepseek-P5 DSv4 platform: cuda Specific optimizations or bugs in NVIDIA graphics cards (such as FlashInfer, TMA optimizations)
#368 opened Sep 1, 2026 by KJLdefeated Collaborator Loading…
@KJLdefeated
12
[WS1] Add batch-invariant h_aggregate kernel deepseek-P1 DSv4 platform: cuda Specific optimizations or bugs in NVIDIA graphics cards (such as FlashInfer, TMA optimizations)
#366 opened Aug 30, 2026 by nodeeeeee Loading…
2
docs: add DCO 1.1 text and contributor sign-off guide type: ci-cd Modify GitHub Actions, automated tests, and packaging/deployment tasks.
#359 opened Aug 29, 2026 by Zhifu-Liu Contributor Loading…
8
[WS2][Logp] Deterministic config option for operator platform: cuda Specific optimizations or bugs in NVIDIA graphics cards (such as FlashInfer, TMA optimizations)
#314 opened Aug 16, 2026 by KJLdefeated Collaborator Loading…
5
[CI][refator]: migrate GPU workflow needs-gpu-ci
#309 opened Aug 14, 2026 by Flink-ddd Collaborator Loading…
12
docs(ws1): batch-invariance contract RFC (#101)
#264 opened Aug 5, 2026 by Billy1900 Loading…
1
17
[WS2][PR1][Logp] Add TP-aware logprob contract and dispatch metadata
#259 opened Aug 2, 2026 by ryankert01 Contributor Loading…
22
[FEAT][kernels] Add fused ratio-clip-aggregate loss primitive needs-gpu-ci
#255 opened Jul 31, 2026 by Chen-BUPT Loading…
1
11
feat(observability): NVTX kernel tracing + Prometheus metrics endpoint
#250 opened Jul 27, 2026 by Billy1900 Loading…
3 of 4 tasks
9
feat(observability): add NVTX tracing and Prometheus metrics
#242 opened Jul 22, 2026 by BruceLoveDecimal • Draft
4
feat(kernels): SM90 TMA + mma.sync FlashAttention, forward + backward (causal, varlen, LSE) needs-gpu-ci stale
#237 opened Jul 20, 2026 by Billy1900 Loading…
5 tasks done
15
[FEAT][kernels]: add backward support for fused logp CUDA kernels needs-gpu-ci
#234 opened Jul 19, 2026 by Dnoob Contributor Loading…
1
11
feat(kernels): Triton FlashAttention LSE export + varlen packing needs-gpu-ci stale
#233 opened Jul 19, 2026 by Billy1900 Loading…
3 tasks done
9
docs(bench): add H100 SXM5 benchmark results alongside A100 baseline stale
#231 opened Jul 19, 2026 by Billy1900 Loading…
4 of 5 tasks
12
feat(executors): add colocated dual-engine with vLLM sleep/wake stale
#218 opened Jul 11, 2026 by icenfly Loading…
16
ProTip! What’s not been updated in a month: updated:<2026-08-13.