How do you compute and track evaluation metrics like accuracy during PyTorch training?

Tracking metrics correctly requires accumulating values across all batches (not just averaging per-batch metrics naively, which can be biased if the last batch has a different size) and ensuring computations happen without gradient tracking.

import torch
import torch.nn as nn

@torch.no_grad()   # disable gradient tracking for the entire evaluation function
def evaluate(model, loader, criterion, device):
    model.eval()                       # disable dropout, use BN running stats

    total_loss    = 0.0
    total_correct = 0
    total_samples = 0

    for X_batch, y_batch in loader:
        X_batch, y_batch = X_batch.to(device), y_batch.to(device)
        batch_size = X_batch.size(0)

        logits = model(X_batch)
        loss   = criterion(logits, y_batch)

        # Weight by batch_size — correct even if the last batch is smaller
        total_loss += loss.item() * batch_size

        preds = logits.argmax(dim=1)
        total_correct += (preds == y_batch).sum().item()
        total_samples += batch_size

    avg_loss = total_loss / total_samples
    accuracy = total_correct / total_samples
    return avg_loss, accuracy

# WRONG pattern — naively averaging per-batch averages
# is biased if batch sizes are unequal (e.g. last batch is smaller)
def evaluate_wrong(model, loader, criterion):
    losses = []
    for X_batch, y_batch in loader:
        loss = criterion(model(X_batch), y_batch)
        losses.append(loss.item())   # all batches weighted EQUALLY — wrong!
    return sum(losses) / len(losses)  # biased if last batch has fewer samples

# Using torchmetrics for more complex metrics (F1, precision, AUROC)
# pip install torchmetrics
from torchmetrics import Accuracy, F1Score

acc_metric = Accuracy(task="multiclass", num_classes=5).to(device)
f1_metric  = F1Score(task="multiclass", num_classes=5, average="macro").to(device)

for X_batch, y_batch in loader:
    preds = model(X_batch).argmax(dim=1)
    acc_metric.update(preds, y_batch)   # accumulates internally across batches
    f1_metric.update(preds, y_batch)

print(f"Accuracy: {acc_metric.compute():.4f}")  # final correct aggregate
print(f"F1: {f1_metric.compute():.4f}")

Invest now in Acorns!!! 🚀
Join Acorns and get your $5 bonus!

Acorns is a micro-investing app that automatically invests your "spare change" from daily purchases into diversified, expert-built portfolios of ETFs. It is designed for beginners, allowing you to start investing with as little as $5. The service automates saving and investing. Disclosure: I may receive a referral bonus.

Invest now!!! Get Free equity stock (US, UK only)!

Use Robinhood app to invest in stocks. It is safe and secure. Use the Referral link to claim your free stock when you sign up!.

The Robinhood app makes it easy to trade stocks, crypto and more.

Webull! Receive free stock by signing up using the link: Webull signup.