Skip to main content

🔥 PyTorch Basics

PyTorch is the undisputed king of Deep Learning frameworks. It's essentially just numpy but with two massive superpowers:

  1. It runs on GPUs (which are 100x faster than CPUs for matrix math).
  2. It automatically calculates derivatives (Autograd).

🐍 Python Implementation: Tensors​

Everything in PyTorch is a Tensor (the word for Matrix/Vector).

import torch

# Create a Tensor
x = torch.tensor([[1.0, 2.0], [3.0, 4.0]])
print("Tensor:\n", x)

# Superpower 1: Move it to the GPU!
# (This code won't run unless you actually have a GPU)
if torch.cuda.is_available():
    x = x.to('cuda')
    print("Tensor is now on the GPU!")
    
# Superpower 2: Autograd
y = torch.tensor(2.0, requires_grad=True)
z = y ** 3
z.backward()
print("Derivative of y^3 at y=2 is:", y.grad) # Should be 12!