Skip to content

Bump mlx-lm-lora from 5.8.6 to 5.10.2 - #19

Merged
avp1598 merged 2 commits into
mainfrom
dependabot/uv/mlx-lm-lora-5.10.2
Nov 4, 2026
Merged

avp1598 merged 2 commits into
mainfrom
dependabot/uv/mlx-lm-lora-5.10.2

Conversation

@dependabot

@dependabot dependabot Bot commented on behalf of github Nov 4, 2026

Copy link
Copy Markdown
Contributor

Bumps mlx-lm-lora from 5.8.6 to 5.10.2.

Release notes

Sourced from mlx-lm-lora's releases.

mlx-lm-lora 5.10.2

Adds epoch-based training budgets to the MCP workflow.

  • Add the read-only calculate_iters MCP tool to convert dataset size, batch size, and epochs into training iterations.
  • Update the training skill with the epoch calculation workflow and accumulation guidance.
  • Fix completed-run detection after MCP server restarts when process IDs are reused.
  • Handle incomplete optional PyTorch namespaces during Hugging Face dataset loading.

Validation: MCP server test suite and syntax checks.

mlx-lm-lora 5.10.1

Fixes

  • Serialize cold Transformers imports to prevent concurrent lazy-import failures in MCP model validation and dataset inspection.
  • Run model validation and dataset inspection sequentially before training.

v5.10.0

Changes

  • Add MCP inspect_dataset and validate_model_support preflight tools.
  • Require successful dataset and model checks before training; block on errors without script, scraping, or clone bypasses.
  • Clarify that SFT supports causal language-model pretraining, including continued pretraining, as well as supervised fine-tuning.

v5.9.0

MLX-LM-LoRA v5.9.0

A minor release adding an MCP training server for agent-driven fine-tuning, with fixes for SFT evaluation, preference scoring, and training base checkpoints.

Changes:

  • Added an optional MCP server to validate training configurations, start and stop runs, and inspect training status and logs over stdio or HTTP. Install it with pip install "mlx-lm-lora[mcp]" and run mlx-lm-lora-mcp.
  • Added a training skill with algorithm, configuration, quantization, and memory-efficiency guides, and expanded the website's training and MCP documentation.
  • Fixed SFT test evaluation returning only the loss instead of the expected loss and token count, preventing a crash when reporting test results after training.
  • Fixed DPO and CPO preference scoring to exclude prompt tokens and padding, aligned chunked scoring and token counts with next-token targets, and corrected validation reward accumulation.
  • Fixed ORPO response masking to retain the first response token. DPO/CPO reward curves and IPO/DPOP behavior can change with the corrected scoring.
  • Added a fallback chat template for base checkpoints without one, preserving existing tokenizer templates and terminating assistant responses with the EOS token when available.

What's Changed

Full Changelog: Goekdeniz-Guelmez/mlx-lm-lora@v5.8.6...v5.9.0

Commits
  • 75abf66 Release 5.10.2: add MCP epoch budgets and runtime fixes
  • 4fa3c51 fix: serialize MCP tokenizer imports
  • 311058c chore: bump version to 5.10.0
  • ee7ed8c Merge pull request #111 from Goekdeniz-Guelmez/codex/mcp-training-preflight
  • 4881e7c Clarify SFT supports pretraining and supervised fine-tuning
  • 381f145 Add MCP dataset inspection and model support preflights
  • 39277a1 Merge pull request #110 from Goekdeniz-Guelmez/codex/bump-version-5.9.0
  • fce8251 chore: bump version to 5.9.0
  • f037d77 Merge pull request #109 from Goekdeniz-Guelmez/codex/simplify-chat-template-f...
  • e24c0fa docs: expand website training and MCP guidance
  • Additional commits viewable in compare view

Dependabot compatibility score

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.


Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

  • @dependabot rebase will rebase this PR
  • @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
  • @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
  • @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
  • @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
  • @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

Bumps [mlx-lm-lora](https://github.com/Goekdeniz-Guelmez/mlx-lm-lora) from 5.8.6 to 5.10.2.
- [Release notes](https://github.com/Goekdeniz-Guelmez/mlx-lm-lora/releases)
- [Commits](Goekdeniz-Guelmez/mlx-lm-lora@v5.8.6...v5.10.2)

---
updated-dependencies:
- dependency-name: mlx-lm-lora
  dependency-version: 5.10.2
  dependency-type: direct:development
  update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>
@dependabot
dependabot Bot requested a review from avp1598 as a code owner November 4, 2026 09:55
@dependabot dependabot Bot added dependencies Pull requests that update a dependency file python:uv Pull requests that update python:uv code labels Nov 4, 2026
@avp1598
avp1598 merged commit 627e38a into main Nov 4, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dependencies Pull requests that update a dependency file python:uv Pull requests that update python:uv code

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant