lmstudio-ai / lmstudio-ai/mlx-engine
QK-Normalization Support
Open
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.2k
- Forks
- 133
- Avg merge
- 21h 6m
- Merged PRs (30d)
- 1
Description
The Qwen3 architecture, which introduced new layers (k_norm and q_norm) which is not supported in LM Studio
see here: https://huggingface.co/mlx-community/LightOnOCR-2-1B-8bit/discussions/1
Could this be added?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
No file or test is named. Start by locating the Qwen3 architecture handling in mlx-engine and review the linked Hugging Face discussion for how k_norm and q_norm are represented. Done means the referenced Qwen3-based model loads and runs correctly in LM Studio with both normalization layers supported.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100