Doubt with the accuracy functión

[chechoreyes]

Video 88
Time: 13:39:04

Hi, i have a question:

Why in this part:

In the accuracy function, you compares y_bloob_test (or train) and test_preds (or train), but i don't undestand. First,y_blob_test is a tensor with the labels and y_pred is the result of torch.softmax(y_logits, dim=1).argmax(dim=1) what are the highest probability indices predicted, then, you compares two different things or PyTorch assumes the value of that index?

Thanks!

[sametucuncu]

Put these commands below after the test_preds = torch.softmax... statement and analyze the outcomes.

print(f"y_blob_test: {y_blob_test}")
print(f"test_preds: {test_preds}")

[mrdbourke]

Hi @chechoreyes ,

Yes, you're correct.

We compare the labels (e.g. y_blob_test) with the predictions (e.g. test_preds) and see where they are equal to each other and sum the outputs.

We then divide by the number of samples to get accuracy as a percentage.

As in the accuracy_fn():

# Calculate accuracy (a classification metric)
def accuracy_fn(y_true, y_pred):
    """Calculates accuracy between truth labels and predictions.

    Args:
        y_true (torch.Tensor): Truth labels for predictions.
        y_pred (torch.Tensor): Predictions to be compared to predictions.

    Returns:
        [torch.float]: Accuracy value between y_true and y_pred, e.g. 78.45
    """
    correct = torch.eq(y_true, y_pred).sum().item()
    acc = (correct / len(y_pred)) * 100
    return acc

Perhaps you could try enter two example tensors in the function above and see what happens?