macromate-pytorch-model

What problem does it solve?

This Skill provides a clean path to replace the old TensorFlow/Keras final_model.h5 with a PyTorch-based best_model.pt in MacroMate, enabling an updated detection pipeline.

Core Features & Use Cases

• Hybrid detection: prefers the custom PyTorch model for inference and falls back to CLIP for uncertain cases.
• End-to-end integration: places new model files, rewrites detector logic, and updates nutrition lookups and UI badges to reflect the active model.
• Use Case: Devs upgrading MacroMate to PyTorch can follow the steps to swap models, adapt file paths, and ensure results display the correct model source badge.

Quick Start

Place the three new files into the models/ folder and completely rewrite ml_food_predictor.py to load best_model.pt with torchvision, implement hybrid detection (custom PyTorch model first → CLIP fallback), and update nutrition_provider.py and UI badges to reflect the PyTorch integration.