ollama - 💡(How to fix) Fix qwen3.5:35b-a3b-mlx-bf16 posts 500: mlx runner failed: Error: failed to read config.json: [1 comments, 2 participants]

500: mlx runner failed: Error: failed to read config.json:

TL;DR

• The issue is likely due to a missing config.json file in the manifest, which is required by the mlx runner, and can be resolved by ensuring the file is present or modifying the runner configuration.

Guidance

• Verify that the config.json file is included in the model's manifest and that its path is correctly specified.
• Check the model's documentation or configuration options to see if there's a way to specify an alternative configuration file or disable the mlx runner's requirement for config.json.
• Investigate why the usual supporting artifacts are not being pulled during the model pull process, as this might be related to the missing config.json issue.
• Consider testing the model pull and run process on a different platform or with a different version of Ollama to isolate if the issue is specific to Apple Silicon or Ollama 0.19.0.

Example

• No specific code snippet can be provided without more information about the model's configuration or the mlx runner's implementation.

Notes

• The issue seems to be specific to Apple Silicon and Ollama 0.19.0, but it's unclear if this is a bug in Ollama or a configuration issue with the model.
• Without more information about the model's configuration or the mlx runner's requirements, it's difficult to provide a more specific solution.

Recommendation

• Apply workaround: Modify the model's configuration or the mlx runner's settings to either include the config.json file in the manifest or disable its requirement, as upgrading to a fixed version of Ollama is not explicitly implied in the issue.