Skip to content

Pull requests: ml-explore/mlx-swift-lm

New pull request New
49 Open 131 Closed
49 Open 131 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

Add continuous batching and Gemma 4 text support
#263 opened May 2, 2026 by Gajesh2007 Loading…
3 tasks done
2
Share the fused gated-delta kernel between MLXLLM and MLXVLM
#257 opened May 1, 2026 by john-rocky Loading…
Add Gemma 4 video tower support
#256 opened May 1, 2026 by john-rocky Loading…
Fix SmolVLM2 tiles() upscaling small images, ~9x slowdown (#208)
#255 opened May 1, 2026 by john-rocky Loading…
2 of 3 tasks
1
Fix Qwen35 VLM sanitize for bare 'model.*' weight keys (#143)
#254 opened May 1, 2026 by john-rocky Loading…
2 of 3 tasks
1
Fix UserInput init not setting self.images / self.videos (#182)
#253 opened May 1, 2026 by john-rocky Loading…
2 tasks done
1
JSONToolCallParser: accept stringified arguments field (#169, Granite 4)
#252 opened Apr 30, 2026 by john-rocky Loading…
2 of 3 tasks
1
2
Fix Gemma 4 text K-eq-V attention: don't double-process v from already-transposed k (#231)
#250 opened Apr 30, 2026 by john-rocky Loading…
2 of 3 tasks
1
Optimize Gemma 4 text decode: compile-fused RMSNorm-add and gelu-mul (+23.8%)
#249 opened Apr 30, 2026 by john-rocky Loading…
3 of 4 tasks
Add LLM generation benchmark helper
#248 opened Apr 30, 2026 by john-rocky Loading…
1 of 2 tasks
Fix Gemma4 attention crash when vProj is nil (attentionKeqV)
#247 opened Apr 29, 2026 by jfranknichols Loading…
Add tool-call id round-trip for OpenClaw / Hermes-style tool calling
#246 opened Apr 28, 2026 by ronaldmannak Contributor Loading…
3 of 4 tasks
2
Make Gemma 4 ScaledLinear quantizable
#244 opened Apr 27, 2026 by kr1s0404 Loading…
2 of 4 tasks
1
Fix Qwen2.5-VL preprocessing to match Python mlx-vlm reference
#243 opened Apr 25, 2026 by NivDvir Loading…
Fix Qwen2-VL chat template: emit vision tokens before text
#242 opened Apr 25, 2026 by NivDvir Loading…
Fix: tolerate 1D token inputs in Gemma 4 VLM forward pass
#241 opened Apr 25, 2026 by brdcastro Loading…
1
Fix MROPE, rope_deltas, and invFreq loading in Qwen2.5-VL
#239 opened Apr 24, 2026 by NivDvir Loading…
Fix vision-encoder attention mask being ignored in Qwen2.5-VL
#238 opened Apr 24, 2026 by NivDvir Loading…
Fix Gemma4Text quantized KV cache attention
#237 opened Apr 24, 2026 by aleroot Contributor Loading…
4 tasks done
Add coherence integration tests
#235 opened Apr 23, 2026 by DePasqualeOrg Contributor Loading…
Remove temporal coupling in applyRotaryPosition
#234 opened Apr 22, 2026 by DePasqualeOrg Contributor Loading…
1
add TurboQuant KV cache compression
#232 opened Apr 22, 2026 by TheTom • Draft
17
add kvScheme parameter for extensible KV cache compression
#230 opened Apr 21, 2026 by TheTom Loading…
4 tasks done
fix segsum dtype promotion -- 2x memory waste on hybrid SSM models
#229 opened Apr 21, 2026 by TheTom Loading…
4 tasks done
fix Gemma 4 MoE router -- softmax order + fuse norm dispatches
#228 opened Apr 21, 2026 by TheTom Loading…
4 tasks done
1
ProTip! Add no:assignee to see everything that’s not assigned.