Skip to content

Pull requests: RL-Align/RL-Kernel

New pull request New
30 Open 83 Closed
30 Open 83 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

[WS1][kernels] Batch-invariant attention (standard softmax--cuda version) needs-gpu-ci
#240 opened Jul 21, 2026 by maxiaosong1124 Collaborator Loading…
1
5
feat(attention): add deterministic CP reference
#238 opened Jul 20, 2026 by inaniloquentee Collaborator Loading…
5
feat(kernels): SM90 TMA + mma.sync FlashAttention, forward + backward (causal, varlen, LSE)
#237 opened Jul 20, 2026 by Billy1900 Loading…
5 tasks done
12
[WS2] Add CP-aware Attention contract and dispatch metadata
#236 opened Jul 19, 2026 by Bsgg1 Contributor Loading…
5
[FEAT][kernels]: add backward support for fused logp CUDA kernels
#234 opened Jul 19, 2026 by Dnoob Contributor Loading…
1
5
feat(kernels): Triton FlashAttention LSE export + varlen packing needs-gpu-ci
#233 opened Jul 19, 2026 by Billy1900 Loading…
3 tasks done
7
docs(bench): add H100 SXM5 benchmark results alongside A100 baseline
#231 opened Jul 19, 2026 by Billy1900 Loading…
4 of 5 tasks
8
feat(alignment): add cross-configuration framework
#230 opened Jul 19, 2026 by CyberSecurityErial Collaborator Loading…
16
[WS1][kernels] RoPE (Triton, CUDA on SM90) needs-gpu-ci
#228 opened Jul 18, 2026 by zhangj1an Collaborator Loading…
8
feat(executors): add colocated dual-engine with vLLM sleep/wake
#218 opened Jul 11, 2026 by icenfly Loading…
15
[FEAT][kernels] Fused Gumbel-Softmax sampler — Triton forward, ST hard path, fused backward needs-gpu-ci
#215 opened Jul 10, 2026 by JLiu4Coding Loading…
17
chore: fix repo-wide pre-commit lint (black/isort/trailing-whitespace)
#213 opened Jul 9, 2026 by Liuzin521 Contributor Loading…
1
[Feat][Kernel] Add TP Fused linear logp Triton needs-gpu-ci
#208 opened Jul 5, 2026 by KJLdefeated Collaborator Loading…
6
feat(ci): cross-platform Docker matrix and hardware CI for CUDA/ROCm (closes #173)
#206 opened Jul 4, 2026 by haoruilee Contributor Loading…
1
10
[WS1][kernels] Batch-invariant logprob (CUDA) needs-gpu-ci
#204 opened Jul 2, 2026 by KJLdefeated Collaborator Loading…
13
feat(kernels): add batch-invariant RMSNorm needs-gpu-ci
#201 opened Jun 29, 2026 by EthanZero2Hero Collaborator Loading…
1
13
[WS1][kernels] Batch-invariant logprob (Native, Triton) needs-gpu-ci
#199 opened Jun 28, 2026 by hihaluemen Contributor Loading…
23
test: add selected logprob parity harness
#198 opened Jun 28, 2026 by Txy02 Loading…
8
[FEAT][kernels] route LM-head vocab projection through deterministic GEMM needs-gpu-ci
#194 opened Jun 26, 2026 by inaniloquentee Collaborator Loading…
7
[WS1]feat: add embedding invariance and LM-head linear_logp routing
#190 opened Jun 25, 2026 by inaniloquentee Collaborator Loading…
3
Add OpenAI-compatible local serving
#183 opened Jun 23, 2026 by haoruilee Contributor Loading…
1
46
feat(kernels): add fused masking + variable-length pack-and-pad op needs-gpu-ci
#182 opened Jun 23, 2026 by Chen-BUPT Loading…
16
[WS2][distributed] Add deterministic all-reduce contract and smoke tests
#181 opened Jun 23, 2026 by CyberSecurityErial Collaborator Loading…
10
[WS1][kernels] Batch-invariant deterministic GEMM (fwd + bwd) needs-gpu-ci
#180 opened Jun 22, 2026 by Flink-ddd Collaborator Loading…
1
11
[WIP] [WS1] Add tests to ensure kv cache consistency for prefill & decode
#178 opened Jun 22, 2026 by zhangj1an Collaborator • Draft
1
3
ProTip! Follow long discussions with comments:>50.