Support for Converting Maestro-Fine-Tuned Models to Core ML for On-Device iOS Deployment

[maxwell44]

Search before asking

• I have searched the Multimodal Maestro issues and found no similar feature requests.

Question

I am writing to inquire about the possibility of converting models fine-tuned using Maestro into the Core ML format for deployment on iOS devices. Specifically, I am interested in understanding whether the models trained with Maestro can be exported to Core ML to facilitate on-device inference within iOS applications.

Context:

Model Training: I have utilized Maestro to fine-tune vision-language models such as Florence-2 for specific tasks.

Deployment Objective: To integrate these fine-tuned models into an iOS application, leveraging Core ML for efficient on-device processing.

Questions:

Export Functionality: Does Maestro currently support exporting fine-tuned models to the Core ML format? If not, are there recommended workflows or tools to achieve this conversion?

Compatibility: Are there any known compatibility considerations or limitations when converting Maestro-fine-tuned models to Core ML, particularly concerning model architecture or supported operations?

Future Support: Are there plans to incorporate direct support for Core ML export within Maestro in future releases?

Additional

No response