5 lessons, in order. Each one runs in the editor above; for the concepts behind them, read the PyTorch guide.
Tensor basicsLearn the PyTorch Tensor concept with NumPy. Pick up the core ideas in an environment without a GPU.
Autograd — backpropagationImplement backpropagation by hand in NumPy to understand how PyTorch Autograd works.
Neural network structureImplement the nn.Module pattern in NumPy and understand the forward and backward flow.
The training loopImplement the loss function, optimizer, and epoch training in NumPy. In PyTorch the same thing takes just a few lines.
Advanced. Real-world input validation and result contractsEven in browser lab code, separating input validation, success responses, and error responses makes it easy to grow into real API/batch code.