lmstudio-ai / lmstudio-ai/mlx-engine
MLX vision models do not currently support KV cache quantization for qwen3.5
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 1.2k
- Forks
- 133
- Avg merge
- 21h 6m
- Merged PRs (30d)
- 1
Description
Title
MLX vision model fails to load when KV cache quantization is enabled
Description
When trying to load a vision model with KV cache quantization enabled, the model fails to load and throws an error indicating that MLX vision models do not currently support KV cache quantization.
Currently, the application allows KV cache quantization to be configured even when loading a vision model, which leads to a runtime error during model initialization.
KV Cache Settings
KV cache quantization bits: 8
Group size strategy: Balanced
Start quantizing when ctx reaches this length: 5000
Error Message
🥲 Failed to load the model
Failed to load model
Error when loading model:
ValueError: MLX vision models do not currently support KV cache quantization
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
Start by reproducing the failure while loading the qwen3.5 vision model with KV cache quantization enabled, then trace model initialization and the KV cache settings path that raises the reported ValueError. Done means the intended handling of KV cache quantization for vision models is defined and loading no longer fails unexpectedly; the issue names no files or tests to begin with.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 52/100