-
Notifications
You must be signed in to change notification settings - Fork 62
Pull requests: jjang-ai/vmlx
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
perf: sampler shared-params fast-path + single_batch async pipelining (PR-C of 3)
#164
opened May 12, 2026 by
st-adam
Loading…
2 of 4 tasks
perf: env-gated prefill allocator clear + chunked-loop tightening (PR-B of 3)
#163
opened May 12, 2026 by
st-adam
Loading…
3 of 5 tasks
perf: low-risk hot-path cleanups (PR-A of 3)
#162
opened May 12, 2026 by
st-adam
Loading…
3 of 5 tasks
feat(pflash): importance-scored sparse prefill scaffold (#136)
#161
opened May 12, 2026 by
st-adam
Loading…
4 of 6 tasks
fix: JANG model compat — MiniMax sanitize + MoEGate quantize bypass
#155
opened May 8, 2026 by
pperezrubio
Loading…
feat(spec): draft-model speculative decoding under continuous batching (#135)
#150
opened May 7, 2026 by
st-adam
Loading…
3 tasks
feat(pld): hybrid partial-accept replay for SSM models (#134)
#149
opened May 7, 2026 by
st-adam
Loading…
3 tasks
Fix opencode config error by moving _mlxstudio marker to options
#101
opened Apr 23, 2026 by
dangeReis
Loading…
fix: gracefully handle pixel_values TypeError for text-only Gemma 4 models
#82
opened Apr 15, 2026 by
yelban
Loading…
3 tasks done
Guard default repetition penalty flag for older external engines
#77
opened Apr 14, 2026 by
Rishirandhawa
Loading…
ProTip!
Type g i on any issue or pull request to go back to the issue listing page.