Skip to content

Pull requests: FlashML-org/FreeToken

New pull request New
Clear current search query, filters, and sorts
113 Open 26 Closed
113 Open 26 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(moe): route fused_topk through the in-repo triton router
#319 by jason-fxz Collaborator was merged Sep 1, 2026 Loading…
fix(kernel): unbreak the nightly kernel-cache wheel build
#310 by jason-fxz Collaborator was merged Aug 31, 2026 Loading…
feat(qwen4_exp): support Qwen3.8-Flash-Next feature New feature or request new-model
#257 by jason-fxz Collaborator was merged Aug 28, 2026 Loading…
3
fix(kernel): avoid row-wise _scaled_mm stall on sm_89 with torch<2.12
#243 by bernimccoy Contributor was merged Aug 29, 2026 Loading…
1
3
feat(qwen4): support Qwen3.8 Flash Next NVFP4, vision, and Windows serving
#232 by MaxKerkula was closed Aug 29, 2026 Loading…
6
feat(models): add Qwen3.8 Flash Next support
#226 by mrcolo was closed Aug 29, 2026 Loading…
5 tasks done
1
fix(server): resolve in-flight generation requests when the backend worker dies
#225 by dalu951 was closed Aug 26, 2026 Loading…
1
feat(qwen4): add Qwen3.8 Flash Next support
#218 by mrcolo was closed Aug 26, 2026 Loading…
4 tasks done
fix(deps): use PyTorch 2.13 to fix SM89 hangs
#184 by endenis was closed Aug 29, 2026 Loading…
1
1
Feat/sm75 cuda128
#157 by JohnieBraaf was closed Aug 24, 2026 Loading…
1
Refactor README for better readability
#149 by murtazaaltaf410-hub was closed Aug 25, 2026 Loading…
feat(server)!: add --gpu to choose the GPU on multi-GPU machines high priority
#117 by jason-fxz Collaborator was merged Aug 24, 2026 Loading…
1
2
feat(moe): per-layer host-bank residency high priority windows
#112 by jason-fxz Collaborator was merged Aug 24, 2026 Loading…
2
feat(kvcache): 8-bit KV cache (q8_0 / fp8_e4m3) behind --kv-cache-dtype
#103 by lucaspirola was closed Aug 30, 2026 Loading…
3
feat(laguna): native GGUF support for poolside Laguna (S/XS)
#102 by lucaspirola was closed Aug 30, 2026 Loading…
1
fix: use current GPU for e4m3 capability
#48 by ousamabenyounes was closed Aug 25, 2026 Loading…
1
3
feat(moe): partial-pin serving to beat the Windows/WSL2 pinned-memory quota (--pin-exempt-layers)
#27 by ervisthegreattt was closed Aug 24, 2026 Loading…
4
fix(checkpoint): FTW truncated shard raises OSError instead of silent garbage load
#25 by RuixiangMa Contributor was merged Aug 24, 2026 Loading…
4
perf(moe): keep the CPU MoE pool and its expert banks on one NUMA node
#18 by gdevenyi was closed Aug 22, 2026 • Draft
3
perf(fp8): run per-tensor fp8 as W8A8 via scaled_mm
#7 by jason-fxz Collaborator was merged Aug 19, 2026 Loading…
1
Build PyPi: pypi-ready wheels, metadata, and publish workflow
#6 by jason-fxz Collaborator was merged Aug 18, 2026 Loading…
1
[Feature] unify reasoning effort on one shared scale
#5 by jason-fxz Collaborator was merged Aug 16, 2026 Loading…
1
Support muse glimmer
#4 by andy-yang-1 Collaborator was merged Aug 18, 2026 Loading…
6
feat(launch): support launch deepseek harness
#3 by jason-fxz Collaborator was merged Aug 15, 2026 Loading…
1
feat(server): enable reasoning_effort on /v1/chat/completions
#2 by jason-fxz Collaborator was merged Aug 14, 2026 Loading…
1
ProTip! no:milestone will show everything without a milestone.