fix: run model in eval mode with inference_mode for predictions in ce…

[bandomatteo]

What was changed

• Updated cell 13 in 02_pytorch_classification.ipynb to:
  • Set the model to eval() mode before prediction
  • Wrap prediction code with torch.inference_mode()

Why this change

Previously, predictions were made while the model was still in training mode and gradients were tracked, which could lead to:

• Inconsistent outputs during inference
• Unnecessary memory and compute usage

By setting the model to evaluation mode and using inference_mode, predictions are now:

• Consistent with expected inference behavior
• More efficient (no gradient tracking)

[bandomatteo]

@mrdbourke