Skip to content

Pull requests: pegainfer-project/pegainfer

New pull request New
24 Open 668 Closed
24 Open 668 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

perf(qwen35): align the logits GEMM width so the output projection stops landing on an align-1 sm_75 kernel
#1046 opened Sep 13, 2026 by CAICAIIs Collaborator Loading…
3
feat(frontend): expose cumulative prefix cache metrics through vllm
#1045 opened Sep 11, 2026 by xiaguan Collaborator • Draft
refactor(qwen35): split scheduler step drivers into scheduler/{steps,emit,telemetry}.rs
#1037 opened Sep 7, 2026 by CAICAIIs Collaborator Loading…
4
refactor(qwen35): split TP response contract into tp_executor/responses.rs
#1036 opened Sep 7, 2026 by CAICAIIs Collaborator Loading…
2
refactor(qwen35): split TP worker runtime into tp_executor/worker.rs
#1035 opened Sep 7, 2026 by CAICAIIs Collaborator Loading…
8
feat(qwen3): add DFlash2 Phase 1 path selector
#1027 opened Sep 2, 2026 by RicardoMin Contributor • Draft
7
perf(k3): five cuts on the CP8 128k prefill superstep
#1026 opened Sep 1, 2026 by xiaguan Collaborator Loading…
1
test(kernels): reduce the duplicated CPU operator oracles in the HD256/HD512 QK-RoPE smokes
#1002 opened Aug 30, 2026 by BreezyB1n Contributor Loading…
5 of 8 tasks
1
1
feat(k3): kernel mining + capsule substrate — vendored vLLM KDA/top-k cubins (+7%/+20% decode) stale Automatically marked after inactivity
#982 opened Aug 28, 2026 by xiaguan Collaborator Loading…
4
feat(qwen35): launch only on the step contract stale Automatically marked after inactivity
#981 opened Aug 28, 2026 by xiaguan Collaborator Loading…
4 tasks
3
devops: make serving benchmarks reproducible stale Automatically marked after inactivity
#979 opened Aug 26, 2026 by xiaguan Collaborator Loading…
3
fix(qwen3): preserve explicit stop-token causes
#978 opened Aug 26, 2026 by RicardoMin Contributor Loading…
36
feat(observability): report real prefix-cache query/hit counters in /metrics
#953 opened Aug 23, 2026 by 1571859588 Loading…
19
Restore the star history chart stale Automatically marked after inactivity
#937 opened Aug 19, 2026 by Dessalines39394 Loading…
4
fix(frontend): unify stop-token semantics across model schedulers stale Automatically marked after inactivity
#865 opened Aug 13, 2026 by RicardoMin Contributor Loading…
@xiaguan
9
feat(higgs-audio): add checkpoint loading and validation foundation stale Automatically marked after inactivity
#864 opened Aug 13, 2026 by ywh555hhh Contributor • Draft
@xiaguan
8
feat(qwen35): add opt-in SM120 FlashInfer GDN prefill
#862 opened Aug 10, 2026 by qwzx-qwas Loading…
7 of 9 tasks
1
@xiaguan
32
feat(qwen35): add joint prefix cache
#836 opened Aug 3, 2026 by Ke-Wng Contributor Loading…
1 of 4 tasks
@xiaguan
31
feat(frontend): join upstream W3C trace context into engine request spans stale Automatically marked after inactivity
#791 opened Jul 30, 2026 by xiaguan Collaborator Loading…
59
test(dsv2lite): add retained HTTP soak gate stale Automatically marked after inactivity
#713 opened Jul 18, 2026 by CAICAIIs Collaborator Loading…
1
3
feat(eagle3): Qwen3 EAGLE-3 drafter forward pass (single-step + chain rollout)
#707 opened Jul 18, 2026 by scatyf3 Contributor Loading…
@xiaguan
6
fix(qwen3): wait for async overlap prefill while idle stale Automatically marked after inactivity
#706 opened Jul 18, 2026 by Nyvo-io Contributor Loading…
1
@FeathBow
31
feat(qwen35): add transactional speculative verifier stale Automatically marked after inactivity
#667 opened Jul 13, 2026 by CAICAIIs Collaborator Loading…
5
fix(qwen3): proactive cancellation sweep for queued requests stale Automatically marked after inactivity
#646 opened Jul 10, 2026 by sparkzky Contributor Loading…
7 tasks done
1
@xiaguan
13
ProTip! Type g i on any issue or pull request to go back to the issue listing page.