Open-source repos to learn from

A step-by-step course on building generative AI apps: prompts, chat, search, function calling and safety.

Classic machine learning through small projects: regression, classification, clustering, NLP and time series.

A broad AI course with notebooks: neural networks, computer vision, NLP and more.

Build a GPT-style model from zero in PyTorch, one chapter at a time. It is the code for the book of the same name.

Runnable examples for the OpenAI API: prompting, embeddings, evals and agents.

Runnable examples for Claude: tool use, RAG, agents and evaluation.

The deep learning framework that most research and courses use: tensors, autograd and training loops.

Load, run and fine-tune pretrained models for text, vision and audio.

Ready pipelines and training tools for diffusion models that make images and audio.

The original Stable Diffusion code. Read it to see how text-to-image generation works.

A model that cuts out any object in an image from one click or a box.

Building blocks that connect LLMs to your data, tools and other services.

Build agents as graphs, with saved state, loops and human approval steps.

Load your documents, index them, and answer questions about them (RAG).

Write LLM programs as code, and let the tool tune the prompts for you.

A framework in which several agents talk to each other to finish a task.

Give each agent a role and a goal, and let them work as a team.

A fast server for LLM inference on GPUs, used in production.

Download and run open models on your own computer with one command.

A chat web app that you host yourself, for local or remote models.