Skip to content

Pull requests: ml-explore/mlx-lm

New pull request New
89 Open 1,095 Closed
89 Open 1,095 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/mlx_lm.evaluate ignoring max_gen_toks and --batch-size during generation
#1812 opened Sep 1, 2026 by dasashreeya Loading…
1
fix: plumb evaluate batch-size to generate_until, accept max_gen_toks
#1810 opened Aug 31, 2026 by axiom-of-choice Contributor • Draft
1
Add Spark2.5 model support new_model
#1809 opened Aug 31, 2026 by simple-zheng Loading…
Feat/server kv cache quant partial
#1804 opened Aug 29, 2026 by soobrosa Loading…
Add pipeline parallelism for Qwen3 MoE await verification This pull request is non-trivial and requires a human expert to verify its correctness.
#1801 opened Aug 29, 2026 by azamamirza Contributor Loading…
1
fix: apply 1/sqrt(head_v_dim) readout scale in gated_delta_update
#1799 opened Aug 28, 2026 by axiom-of-choice Contributor Loading…
1
2
Add DeepSeek-V4-Flash without MTP (yet)
#1797 opened Aug 28, 2026 by nh13 Contributor Loading…
1
Return 503 from /health when the generation thread exits await response This pull request is waiting for response from the author.
#1791 opened Aug 27, 2026 by alexis-ag Loading…
2
Add Qwen3.8-Flash-Next (qwen4_exp) model support await verification This pull request is non-trivial and requires a human expert to verify its correctness.
#1788 opened Aug 26, 2026 by eauchs Loading…
9
Make "state" of cache return full state
#1778 opened Aug 24, 2026 by zcbenz Member Loading…
2
Add Intern-S2-Mobius (interns2_mobius) model new_model
#1771 opened Aug 21, 2026 by nightscape • Draft
2
Add M-RoPE support so vision embeddings keep their grid positions enhancement
#1768 opened Aug 21, 2026 by lpalbou Contributor Loading…
1
Feature/refactor generate enhancement
#1722 opened Aug 13, 2026 by michalk8 Collaborator Loading…
@michalk8
2
fix(granitemoehybrid): accept nested rope_parameters and tied lm_head weight await response This pull request is waiting for response from the author. bug
#1605 opened Jul 22, 2026 by lkrapf Loading…
1
Fix deepseek_v32 Indexer evicting attention sinks from sparse top-k bug
#1552 opened Jul 11, 2026 by robertlangdonn Contributor Loading…
5 tasks done
1
1
Add Nemotron-H Puzzle model support new_model
#1535 opened Jul 10, 2026 by sxuff Loading…
5
Fail fast when the generation thread stops bug
#1514 opened Jul 10, 2026 by dogukanveziroglu Contributor Loading…
Keep the server generation loop alive when a batched request fails bug
#1513 opened Jul 10, 2026 by rajanshxrma Loading…
1
3
Add support for Talkie (talkie-1930-13b) new_model
#1492 opened Jul 7, 2026 by ethanelasky Loading…
Fix float32 promotion in BatchKVCache/BatchRotatingKVCache extend() bug
#1491 opened Jul 7, 2026 by ethanelasky Loading…
1
Add support for ESMC and ESMFold2 models new_model
#1484 opened Jul 6, 2026 by faustomilletari Loading…
3
Fix mixed-bit quantized load for sanitize()-derived MLA projections bug
#1482 opened Jul 6, 2026 by rajanshxrma Loading…
1
fix(vl): strip model.visual.* in qwen2_vl/qwen3_vl/qwen3_vl_moe sanitize bug
#1473 opened Jul 4, 2026 by Jonathangadeaharder Loading…
Add LongCat-2.0 new_model
#1464 opened Jul 4, 2026 by kernelpool Contributor Loading…
DeepSeek-V3.2/GLM DSA: fix silent >128k top-k corruption + sparse-gather prefill bug
#1454 opened Jul 2, 2026 by aidiffuser Loading…
1
ProTip! Follow long discussions with comments:>50.