Course contentsShow
← Machine Learning and Deep Learning
Lesson 837 of 3,53819. PyTorch FundamentalsFree lesson

Training Progress Monitoring

Using tqdm for progress bars and logging libraries to track training metrics in real-time.

Training Progress Monitoring

What you'll learn: How to use tqdm for visual progress bars and logging libraries to track your training metrics in real-time, making your model training transparent and debuggable.

Why Real-Time Monitoring Matters

Imagine baking a cake but never opening the oven or checking the timer—you'd have no idea if it's burning or undercooked until it's too late. Training neural networks is similar: without monitoring progress in real-time, you're flying blind. You need to see how loss decreases, how accuracy improves, and how long each epoch takes.

Using tqdm for Progress Bars

tqdm is a Python library that wraps iterables and displays a smart progress bar. When training with PyTorch DataLoaders, it shows:

  • Current iteration and total iterations
  • Time elapsed and estimated time remaining
  • Processing speed (iterations per second)
  • Custom metrics you want to track
from tqdm import tqdm

# Wrap your DataLoader with tqdm
for epoch in range(num_epochs):
    train_bar = tqdm(train_loader, desc=f"Epoch {epoch+1}/{num_epochs}")
    
    for batch_idx, (data, target) in enumerate(train_bar):
        # Your training code here
        loss = compute_loss(data, target)
        
        # Update progress bar with current metrics
        train_bar.set_postfix(loss=loss.item())

Logging for Persistent Records

While progress bars are great for watching training live, logging libraries (like Python's built-in logging) save metrics to files for later analysis:

import logging

logging.basicConfig(filename='training.log', level=logging.INFO)

for epoch in range(num_epochs):
    train_loss = train_one_epoch()
    val_loss = validate()
    
    logging.info(f"Epoch {epoch+1}: Train Loss={train_loss:.4f}, Val Loss={val_loss:.4f}")

Key Takeaway: Use tqdm to visualize training progress interactively and logging to create permanent records of your experiments—both help you understand what's happening during training and debug issues quickly.