Skip to content

Pull requests: waybarrios/vllm-mlx

New pull request New
53 Open 106 Closed
53 Open 106 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

fix: report prompt_tokens correctly for LLM models in SimpleEngine
#236 opened Mar 30, 2026 by sjswerdloff Loading…
3 tasks
1
perf(reasoning): O(1) state-machine streaming parser (13-19x faster at 2k+ tokens)
#234 opened Mar 29, 2026 by penumbraforge Loading…
1
Add TurboQuant KV cache compression for prefix cache (4.6x)
#233 opened Mar 29, 2026 by arozanov Loading…
9 tasks done
fix: suppress tool call XML from streaming text content (#129)
#232 opened Mar 29, 2026 by sjswerdloff Loading…
1
3
feat: add MiniMax tool call parsing support
#231 opened Mar 29, 2026 by sjswerdloff Loading…
fix: add missing return in load_model_with_fallback
#230 opened Mar 29, 2026 by sjswerdloff Loading…
fix: populate tokens field in BatchedEngine.generate()
#229 opened Mar 28, 2026 by mmcaulif Loading…
3 tasks done
fix: bump mlx-lm minimum to 0.31.0 for hybrid model batching
#227 opened Mar 25, 2026 by krystophny Loading…
1
test: make Python 3.13 async suite pass and cover it in CI
#226 opened Mar 25, 2026 by krystophny Loading…
fix: MLLM hybrid batching + message normalization
#224 opened Mar 25, 2026 by Thump604 Loading…
1
1
feat: MTP per-request routing in BatchedEngine
#223 opened Mar 24, 2026 by Thump604 Loading…
2 of 3 tasks
simple-engine: keep tool chat on the streaming execution path
#222 opened Mar 24, 2026 by krystophny Loading…
1
scheduler: preserve prompt checkpoints in chunked prefill resume path
#221 opened Mar 24, 2026 by krystophny Loading…
2
engine: keep SimpleEngine serialized across cancellation
#220 opened Mar 24, 2026 by krystophny Loading…
2
chat: forward chat_template_kwargs on simple-engine paths
#218 opened Mar 24, 2026 by krystophny Loading…
prefix_cache: preserve hybrid recurrent state across blocks
#217 opened Mar 24, 2026 by krystophny Loading…
2
cli: expose harmony and gpt-oss tool parsers
#216 opened Mar 24, 2026 by krystophny Loading…
tokenizer: return successful mlx-lm load result
#215 opened Mar 24, 2026 by krystophny Loading…
server: add OpenAI-compatible /v1/responses endpoint
#214 opened Mar 24, 2026 by krystophny Loading…
feat: full sampling parameter support (top_k, min_p, presence_penalty, repetition_penalty)
#213 opened Mar 23, 2026 by Thump604 Loading…
5 tasks done
3
fix: respect tool_choice="none" by excluding tools from template
#210 opened Mar 23, 2026 by awanawana Loading…
1
1
fix: Don’t truncate base64 images before hashing
#206 opened Mar 22, 2026 by BelieveDiffusion Loading…
1
2
feat: add lifecycle-managed residency for the default server model
#205 opened Mar 22, 2026 by lyonsno Loading…
fix: skip RNN snapshots in MTP optimistic mode to prevent memory leak
#196 opened Mar 21, 2026 by Thump604 Loading…
4 tasks
2
fix: streaming detokenizer for UTF-8-safe incremental decode
#195 opened Mar 21, 2026 by Thump604 Loading…
5 tasks
1
2
ProTip! Mix and match filters to narrow down what you’re looking for.