Skip to content

Pull requests: lightseekorg/tokenspeed

New pull request New
44 Open 344 Closed
44 Open 344 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

[WIP] perf(gdn): remove gdn prefill unnecessary h_state copy
#409 opened Jun 10, 2026 by minedec Contributor Loading…
[CI] Use new label for gfx950 benchmark jobs
#408 opened Jun 10, 2026 by antiagainst Member Loading…
2
fix(runtime): use scattered token counts for MoE RSAG
#407 opened Jun 10, 2026 by Williams500 Loading…
1
2
[WIP] feat: add DeepSeek V4 L2 KV offload
#406 opened Jun 10, 2026 by SimonCqk Contributor Loading…
4
[WIP] Split AMD Gluon kernels package
#404 opened Jun 9, 2026 by antiagainst Member Loading…
10
Add small-M Gluon warp-decode MoE path for GPT-OSS
#403 opened Jun 9, 2026 by panditsa Contributor Loading…
DRAFT
#399 opened Jun 9, 2026 by netanel-haber • Draft
perf(deepseek-v4): pre-compile deep_gemm JIT kernels at startup
#398 opened Jun 9, 2026 by dongjiyingdjy Contributor Loading…
3 of 4 tasks
4
feat(engine): Sleep / Wake Up API (release/resume_memory_occupation)
#393 opened Jun 9, 2026 by HJSang Collaborator Loading…
7
[WIP] Try new Triton package 20260608
#392 opened Jun 9, 2026 by antiagainst Member Loading…
2
[WIP] refactor(spec-decode): simplify Llama Eagle3 attention path for #217 (1/3)
#390 opened Jun 9, 2026 by rjzhb Contributor Loading…
3
[WIP] feat(config): runtime config decoupling(design for reference)
#383 opened Jun 8, 2026 by rjzhb Contributor • Draft
perf(gdn): fuse causal_conv1d and QKV split for GDN prefill
#382 opened Jun 8, 2026 by elwhyjay Contributor Loading…
6
fix(scheduler): publish prefix to radix tree during prefill for non-hybrid models
#381 opened Jun 8, 2026 by qywu Collaborator Loading…
12
[WIP] feat(kernel): introduce moe kernel api
#374 opened Jun 7, 2026 by borontion Contributor Loading…
11
fix(cache): Coarsely fence the compute stream behind the host loadback stream on.
#370 opened Jun 6, 2026 by LorrinWWW Contributor Loading…
2
[Bugfix] fix smg command argument cannot be used multiple times
#366 opened Jun 6, 2026 by lengrongfu Loading…
1
3
feat(mla): decode-context-parallel (DCP) support in the MLA decode kernel
#364 opened Jun 5, 2026 by RomaA2000 Loading…
6
[Bugfix] fix miss help command
#362 opened Jun 5, 2026 by lengrongfu Loading…
[Bugfix]fix cli logger not set level
#360 opened Jun 5, 2026 by lengrongfu Loading…
2
feat(video) Generalize multimodal runtime support and add Qwen3.5 video
#354 opened Jun 4, 2026 by yechank-nvidia Collaborator • Draft
2
Add initial GLM 5.1 support
#348 opened Jun 3, 2026 by FlamingoPg Contributor • Draft
27
feat(rl): weight-transfer control plane for RL online weight sync
#342 opened Jun 2, 2026 by qywu Collaborator • Draft
feat(mla): support custom tree masks in decode
#338 opened Jun 2, 2026 by mesaleh Loading…
feat(logprobs): vLLM-style output logprobs (LogprobParams), spec-decode support
#337 opened Jun 2, 2026 by HJSang Collaborator Loading…
45
ProTip! Filter pull requests by the default branch with base:main.