📏 supervised_dataset_evaluator.py

Evaluates a trained supervised ViT on an entire dataset. The script loads the configuration stored inside the checkpoint, builds the dataloader and reports classification metrics.

Usage

python evaluators/supervised_dataset_evaluator.py checkpoint=path/to/model.pth

Predictions and the computed Top‑1 accuracy are written to evaluation_results.json and predictions.csv in the current directory. Pass eval.confusion_matrix=true in the Hydra command to additionally store a confusion matrix.