02_pytorch_classification - 89 : model.train() and model.eval()

[dirk-kappel-effectual]

I notice that the book version does not set the model_3.train() as you do in the video version at the beginning of each epoch. Prior to Testing being performed during each epoch the model gets set to eval() mode. How does the model reset itself to train (or does it?) within the for loop (In the Book Version) if the model is never set back to train() mode at the start of each epoch? I ran the model using the Book Version code (without resetting to train()) and my results are the same as the video version.

Book Version Code:
https://www.learnpytorch.io/02_pytorch_classification/#63-training-a-model-with-non-linearity

# Fit the model
torch.manual_seed(42)
epochs = 1000

# Put all data on target device
X_train, y_train = X_train.to(device), y_train.to(device)
X_test, y_test = X_test.to(device), y_test.to(device)

for epoch in range(epochs):
    # 1. Forward pass
    y_logits = model_3(X_train).squeeze()
    y_pred = torch.round(torch.sigmoid(y_logits)) # logits -> prediction probabilities -> prediction labels
    
    # 2. Calculate loss and accuracy
    loss = loss_fn(y_logits, y_train) # BCEWithLogitsLoss calculates loss using logits
    acc = accuracy_fn(y_true=y_train, 
                      y_pred=y_pred)
    
    # 3. Optimizer zero grad
    optimizer.zero_grad()

    # 4. Loss backward
    loss.backward()

    # 5. Optimizer step
    optimizer.step()

    ### Testing
    model_3.eval()
    with torch.inference_mode():
      # 1. Forward pass
      test_logits = model_3(X_test).squeeze()
      test_pred = torch.round(torch.sigmoid(test_logits)) # logits -> prediction probabilities -> prediction labels
      # 2. Calculate loss and accuracy
      test_loss = loss_fn(test_logits, y_test)
      test_acc = accuracy_fn(y_true=y_test,
                             y_pred=test_pred)

    # Print out what's happening
    if epoch % 100 == 0:
        print(f"Epoch: {epoch} | Loss: {loss:.5f}, Accuracy: {acc:.2f}% | Test Loss: {test_loss:.5f}, Test Accuracy: {test_acc:.2f}%")

This is a snippet of what the video version looks like:

# Loop through the data
epochs = 1000

for epoch in range(epochs):
    ### Training
    model_3.train()    <-- This is not in the book version.

    # 1. Forward pass
    y_logits = model_3(X_train).squeeze()
    y_pred = torch.round(torch.sigmoid(y_logits)) # logits -> prediction probabilities -> prediction labels
    
    # 2. Calculate loss
    loss = loss_fn(y_logits, y_train) # BCEWithLogitsLoss
...