v1.0.0 – Initial Model Release

This release includes the pre-trained model weights (model.pth).

Model Details

• Architecture: EfficientNet_B2 (IMAGENET1K_V1)
• Test Accuracy: 80.0% (expect 78% - 82% for this architecture)
• Dataset: Food101
• Sagemaker Instance Type: ml.g4dn.xlarge

Training Hyperparameters

• seed: 42
• batch-size: 128
• num-epochs-phase1: 8
• num-epochs-phase2: 10
• lr-head: 4e-3
• lr-backbone: 5e-4
• patience: 3
• num-workers: 2
• img-size: 224

What's Included

• model.pth: Pre-trained weights ready for inference or further fine-tuning
• All supporting code and dependencies as in previous versions

How to Use

1. Download model.pth from the Assets section below.
2. Place it in the appropriate directory as described in the README.