Lumia101/Flowers102-EfficientNet-B2
This model is a fine-tuned model of EfficientNet-B2, pre-trained on ImageNet-1K, for the Flowers 102 dataset.
Test Results
- Loss: 1.2519 (With label_smoothing=0.1)
- Accuracy: 90.19%
Training Hyperparameters
- Batch size: 128
- Optimizer: AdamW
- Learning Rate: 3e-4
- Weight Decay: 1e-4
- LR Scheduler Type: CosineAnnealingLR
- Epochs: 20
- Automated Mixed Precision: Native AMP
Training Results
See TensorBoard.
- Downloads last month
- 28
Model tree for Lumia101/Flowers102-EfficientNet-B2
Base model
google/efficientnet-b2