-
Notifications
You must be signed in to change notification settings - Fork 933
Pull requests: ml-explore/mlx-lm
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Raise clear error when LoRA config uses 'alpha' instead of 'scale'
#1671
opened Aug 7, 2026 by
arjunneema09
Loading…
Save the best-validation-loss adapter, and add optional early stopping
#1665
opened Aug 5, 2026 by
axiom-of-choice
Contributor
Loading…
Add optional gradient clipping and non-finite guard to the LoRA trainer
#1659
opened Aug 5, 2026 by
axiom-of-choice
Contributor
Loading…
fix(generate): guard per-sequence logits_processors against None in batched generation
#1657
opened Aug 4, 2026 by
chronoshift
Loading…
Enable readline for the chat prompt and persist history
#1654
opened Aug 3, 2026 by
Vlor999
Loading…
Add opt-in q4_0 lattice calibration to quantization
#1646
opened Aug 2, 2026 by
GoodOlClint
Loading…
4 tasks done
fix(tool_parsers/mistral): parse the classic [TOOL_CALLS] JSON-array format
#1643
opened Jul 30, 2026 by
Anai-Guo
Loading…
Fix per-token Metal buffer-object leak in
ArraysCache.advance() (qwen3_5/qwen3_next batch decode crashes at the resource limit on long completions)
#1642
opened Jul 29, 2026 by
thinkroth
Loading…
fix(server): build /v1/models body before committing 200 status (#1636)
#1638
opened Jul 29, 2026 by
Anai-Guo
Loading…
fix(qwen3_coder): accept float-formatted integers in tool-call args
#1635
opened Jul 29, 2026 by
Anai-Guo
Loading…
Materialize shards incrementally in save_model to avoid GPU watchdog timeout
#1634
opened Jul 29, 2026 by
axiom-of-choice
Contributor
Loading…
Fix unbounded ArraysCache metadata graph during decode
#1632
opened Jul 28, 2026 by
LukaAfonin
Loading…
Fix logprobs being silently dropped from streamed responses
#1630
opened Jul 28, 2026 by
slipstr34m
Loading…
Previous Next
ProTip!
Adding no:label will show everything without a label.