Skip to content

Pull requests: openinfer-project/openinfer

New pull request New
43 Open 505 Closed
43 Open 505 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(kernels): QK-norm and partial RoPE prep at head_dim 512
#838 opened Aug 3, 2026 by FeathBow Collaborator Loading…
1 task done
1
feat(qwen35): add joint prefix cache
#836 opened Aug 3, 2026 by Ke-Wng Contributor • Draft
1 of 4 tasks
feat(glm52): kernel_lab fp8 GEMM lab — CUTLASS sm_103 units + CuTe DSL tcgen05 line
#835 opened Aug 3, 2026 by xiaguan Collaborator Loading…
3
test(gemma4): record the HF reference for the 12B text tower
#832 opened Aug 2, 2026 by FeathBow Collaborator Loading…
1
feat(glm52): full migration to openinfer-kv-store — resolver pipeline, keyed tail verbs, 1P1D e2e verified (#824)
#830 opened Aug 2, 2026 by xiaguan Collaborator Loading…
13
refactor(qwen3): migrate prefix/offload path to openinfer-kv-store
#829 opened Aug 2, 2026 by xiaguan Collaborator Loading…
8
feat(gemma4): validate and load the 12B text tower
#827 opened Aug 2, 2026 by FeathBow Collaborator Loading…
1 task done
1
perf(kernels): budget cublasLt tuning repeats by weight traffic
#822 opened Aug 1, 2026 by FeathBow Collaborator Loading…
feat(core): split the RoPE frequency denominator from the rotary extent
#821 opened Aug 1, 2026 by FeathBow Collaborator Loading…
1 task done
1
feat(kernels): add sliding-window paged attention for head_dim 256
#819 opened Aug 1, 2026 by FeathBow Collaborator Loading…
1 task done
1
PR: wire prefix-cache hit counters into /metrics
#814 opened Aug 1, 2026 by 1571859588 Loading…
3 tasks
8
docs(gemma4): specify the two-group KV cache contract
#809 opened Jul 31, 2026 by FeathBow Collaborator • Draft
1 task done
1
2
test(sim): add tool-call round-trip e2e tests
#808 opened Jul 31, 2026 by 1571859588 Loading…
1
test(gemma4): gate tokenizer and chat-template parity against a reference
#794 opened Jul 30, 2026 by FeathBow Collaborator Loading…
1
9
feat(frontend): join upstream W3C trace context into engine request spans
#791 opened Jul 30, 2026 by xiaguan Collaborator Loading…
58
feat(metrics): spec-decode acceptance counters for the DFlash draft path
#787 opened Jul 29, 2026 by scatyf3 Contributor Loading…
1
3
feat(qwen35): overlap Unified prefill with active decode
#786 opened Jul 29, 2026 by CAICAIIs Collaborator Loading…
1
fix(kv-offload): support LLVM 22 RDMA bindings
#785 opened Jul 29, 2026 by ClSlaid • Draft
feat(kv-offload): import the server-allocated KV arena over CUDA IPC
#764 opened Jul 26, 2026 by xiaguan Collaborator Loading…
3
perf(kimi-k2): compact local top-1 readback into one pinned transfer
#734 opened Jul 21, 2026 by LanluZ Contributor Loading…
4 of 8 tasks
1
1
feat(logprobs): batched device-side logprobs reduction for Qwen3 stale Automatically marked after inactivity
#723 opened Jul 19, 2026 by n-WN Contributor Loading…
1
3
fix(logprobs): preserve None/Some(0)/Some(k) contract end-to-end (#720)
#722 opened Jul 19, 2026 by n-WN Contributor Loading…
1
6
perf(qwen): reuse PrefillPagedPlan for uncompiled-GQA decode
#714 opened Jul 18, 2026 by Nyvo-io Contributor Loading…
1
@FeathBow
22
test(dsv2lite): add retained HTTP soak gate
#713 opened Jul 18, 2026 by CAICAIIs Collaborator Loading…
1
2
perf(glm52): connect remote rank hosts concurrently
#712 opened Jul 18, 2026 by Nyvo-io Contributor Loading…
6
ProTip! Exclude everything labeled bug with -label:bug.