PyTorch

A little PyTorch

Notebook (download/open locally): /kicsigeze/notebooks/03_pytorch_mini_intro.ipynb

What you should get out of this

• Understand tensors vs NumPy arrays (device, dtype, gradients)
• Understand autograd and how computation graphs behave
• Implement a tiny training loop (linear regression / 2-layer MLP)

Exercises (quick list)

• Re-derive MSE gradient and validate with autograd
• Train logistic regression on a synthetic dataset
• Add L2 regularization and observe effect

Readings/videos: Reading / Videos