Decoder. plain-English AI glossary

PyTorch

● Core

Facebook's deep learning framework; dominant choice for LLM research and development.

Think of it like

The researcher's choice; flexible, debuggable, dominates academia.

Example

Write a custom training loop in PyTorch; adjust backprop or loss dynamically.

How it actually works

PyTorch uses dynamic computation graphs; more flexible than static graphs (TensorFlow). Strong ecosystem: Hugging Face, PyTorch Lightning, DeepSpeed all built on PyTorch.

For product teams

Market leader in research; drives recruitment and technical culture at AI labs.

For engineers

PyTorch is the default for LLM training and fine-tuning; ecosystem is mature.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome