Model evaluation
With the training loop sorted out, you have trained the model for 1000 epochs, and it is available to you as net
. You have also set up a test_dataloader
in exactly the same way as you did with train_dataloader
before—just reading the data from the test rather than the train directory.
You can now evaluate the model on test data. To do this, you will need to write the evaluation loop to iterate over the batches of test data, get the model's predictions for each batch, and calculate the accuracy score for it. Let's do it!
This exercise is part of the course
Intermediate Deep Learning with PyTorch
Exercise instructions
- Set up the evaluation metric as
Accuracy
for binary classification and assign it toacc
. - For each batch of test data, get the model's outputs and assign them to
outputs
. - After the loop, compute the total test accuracy and assign it to
test_accuracy
.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
import torch
from torchmetrics import Accuracy
# Set up binary accuracy metric
acc = ____
net.eval()
with torch.no_grad():
for features, labels in dataloader_test:
# Get predicted probabilities for test data batch
outputs = ____
preds = (outputs >= 0.5).float()
acc(preds, labels.view(-1, 1))
# Compute total test accuracy
test_accuracy = ____
print(f"Test accuracy: {test_accuracy}")