Skip to content

Pull requests: Blaizzy/mlx-vlm

New pull request New
56 Open 991 Closed
56 Open 991 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 dedicated Poolside Laguna tool parser
#1802 opened Aug 5, 2026 by IronBCC Loading…
Add qwen3 text backbone support to llmjpvl (Jagle-VL)
#1801 opened Aug 5, 2026 by takuyaomi Contributor Loading…
Fix Qwen3.5 ragged decode on GPUs that reject a 1024-thread threadgroup
#1799 opened Aug 5, 2026 by neil0306 Contributor Loading…
1
feat(server): route transcription vocabulary by backend capability
#1796 opened Aug 5, 2026 by LauraGPT Loading…
1
Improve LFM continuous batch decode throughput
#1795 opened Aug 5, 2026 by Blaizzy Owner • Draft
1
Fix wrapped mixed-quantization path matching
#1793 opened Aug 5, 2026 by ft-ioxcs Loading…
1
fix(quant): honor per-layer quantization overrides for wrapped text models
#1791 opened Aug 4, 2026 by Lazarus-931 Collaborator Loading…
server: add --max-num-seqs to bound concurrent batch size
#1782 opened Aug 3, 2026 by Lazarus-931 Collaborator Loading…
1
Fix APC extra_hash to use prefix-invariant semantic inputs
#1779 opened Aug 2, 2026 by f1yn Loading…
3 tasks done
1
1
fix(tokenizer): keep incomplete UTF-8 tail buffered at streaming flush boundary (#1771)
#1778 opened Aug 2, 2026 by FreazeMood Loading…
1
fix: tolerate null vision_config in quantization skip_vision path
#1777 opened Aug 2, 2026 by garnetlyx Loading…
1
2
Stop Idefics3 generation at end of utterance
#1774 opened Aug 2, 2026 by jrp2014 • Draft
embedding_loader: quantize pre-quantized weights before loading
#1773 opened Aug 2, 2026 by Lazarus-931 Collaborator Loading…
Fix image-valued Responses function output handling
#1761 opened Jul 31, 2026 by Rajeev-SG Loading…
1
feat(speculative): validate Laguna DFlash contract
#1753 opened Jul 29, 2026 by Thump604 Contributor Loading…
5
Add native GLM-4.7 attention with TurboQuant APC support
#1742 opened Jul 27, 2026 by Blaizzy Owner • Draft
1
Flatten image_grid_thw collation for multi-image SFT codex codex-automation
#1734 opened Jul 27, 2026 by Blaizzy Owner Loading…
1
Honor explicit trust_remote_code in processor from_pretrained
#1731 opened Jul 27, 2026 by camparker Loading…
1
Match MODEL_CONFIG's lowercase convention when detecting text-only models
#1723 opened Jul 25, 2026 by danmackinlay Loading…
1
1
Preserve APC prefix reuse for growing prepared prompts
#1713 opened Jul 25, 2026 by byoan Loading…
1
fix(rope): apply declared rope_scaling in models that ignored it (Llama-3.2-Vision and 9 others)
#1710 opened Jul 24, 2026 by 1qh Loading…
fix(gemma3): apply rope_scaling to global attention layers
#1709 opened Jul 24, 2026 by 1qh Loading…
3
Shared transformer block
#1704 opened Jul 24, 2026 by Lazarus-931 Collaborator Loading…
1
Add DRY (Don't Repeat Yourself) repetition sampler
#1662 opened Jul 22, 2026 by Lazarus-931 Collaborator Loading…
fix(server): row-level per-request sampling in the continuous batch (supersedes #1353)
#1647 opened Jul 21, 2026 by mmmugh Loading…
14
ProTip! What’s not been updated in a month: updated:<2026-07-05.