get_acc

Compute top-1 classification accuracy of model on dataset.

Puts the model in eval mode for the pass and restores train mode before returning so the caller's training loop is unaffected.

Parameters:

Name	Type	Description	Default
model	Module	Classifier returning logits of shape [batch, num_classes].	required
dataset	Dataset	Iterable yielding (data, target) pairs.	required
batch_size	int	DataLoader batch size.	128
device	str	Device to run evaluation on.	'cuda'

Returns:

Type	Description
Tensor	Scalar torch.Tensor with the accuracy in [0, 1].