PyTorch Essentials for Model Building

Create PyTorch tensors with specified shapes, data types, and devices.

Manipulate tensors using indexing, reshaping, reduction, and broadcasting operations.

Use PyTorch autograd to compute gradients of scalar objectives.

Define a multilayer perceptron by subclassing `torch.nn.Module`.

Compute a loss and update model parameters with a PyTorch optimizer.

Switch correctly between training, evaluation, and gradient-disabled execution modes.

Save and restore model and optimizer state for continued training.

Make a small PyTorch experiment reproducible by controlling random seeds and deterministic settings.