GPU Runtime
A notebook session backed by an NVIDIA GPU for fast deep-learning training.
- Free tiers: T4 (16 GB) or P100 (16 GB).
- Paid: A100, L4, H100.
- Python code unchanged; only runtime differs.
When to use: Anything beyond toy MLP training that would otherwise be unbearably slow on CPU.
Example: In Colab: Runtime > Change runtime type > T4 GPU. torch.cuda.is_available() → True.