Reference · Glossary
PyTorch
Last updated
An open-source deep learning framework — the most widely used one for research and increasingly production — built around tensors (GPU-friendly arrays) and automatic differentiation for training neural networks.
#When to use
Building or fine-tuning a real neural network, especially anything beyond a toy — most published models and Hugging Face checkpoints are PyTorch-first.
#When not to
Classical ML on tabular data (a spreadsheet of rows and columns) — scikit-learn is simpler and usually a better fit than standing up a neural network.
#Example
import torch
import torch.nn as nn
model = nn.Linear(in_features=3, out_features=1)
x = torch.tensor([[0.2, 0.9, 0.4]])
print(model(x))