def main(args):
# Get dataset
test_files = glob.glob(os.path.join(args.test_path, '*.pickle'))
datasource = DataSource(None,
test_files,
shape=tuple(args.eye_shape),
data_format=args.data_format,
heatmap_scale=args.heatmap_scale)
# Get model
learning_schedule = [{
'loss_terms_to_optimize': {
'heatmaps_mse': ['hourglass'],
'radius_mse': ['radius'],
},
'learning_rate': 1e-3,
}]
model = CNN(datasource.tensors, datasource.x_shape, learning_schedule)
# Get evaluator
evaluator = Trainer(model, model_checkpoint=args.model_checkpoint)
# Evaluate
avg_losses = evaluator.run_eval(datasource)
print('Avarage Losses', avg_losses)