Artifex Example Overview¤
Artifex publishes docs/examples/ pages only when a real source pair exists
under examples/generative_models/: one runnable .py file and one paired
.ipynb notebook. Not every published pair is a canonical runtime-backed
tutorial, though. The examples catalog now separates runtime-backed tutorials,
standalone pedagogy, exploratory workflows, and validation utilities so the
public teaching surface matches the current code honestly. Still-relevant topics
without runnable pairs live in Planned Example Topics.
Published Example Buckets¤
- Runtime-backed tutorials: examples that instantiate shipped Artifex owners end to end, such as VAE on MNIST, Simple GAN, Diffusion on MNIST, DiT Demo, and the retained protein and geometric examples in the main catalog.
- Standalone pedagogy: raw JAX/Flax NNX concept walkthroughs that are clearly labeled as not instantiating shipped Artifex runtime owners, such as Simple Diffusion, Simple Audio Generation, Simple Text Generation, and Simple Image-Text.
- Exploratory workflows: useful but non-canonical example tiers that currently rely on lower-level components or custom orchestration, such as Advanced GAN and Protein Diffusion.
- Validation utilities: quick environment or technology-stack checks, such as Protein Diffusion Technical Validation.
Running Examples From A Source Checkout¤
source ./activate.sh
uv run python examples/generative_models/image/vae/vae_mnist.py
uv run python examples/generative_models/protein/protein_extensions_example.py
uv run python examples/generative_models/audio/simple_audio_generation.py
Working With Example Pairs¤
- The Python source is the review surface. Regenerate the notebook from it with
uv run python scripts/jupytext_converter.py sync examples/path/to/example.py. - Reader-facing example docs belong under
docs/examples/only when the pair exists. The catalog taxonomy then decides whether that pair is runtime-backed, standalone, exploratory, or validation-oriented. - Verification or maintenance scripts such as
examples/verify_examples.pyare utilities, not canonical tutorial examples.
Choosing A Starting Point¤
- New to Artifex: start with VAE on MNIST or Simple GAN.
- Interested in current training workflows: start with Advanced Training and the Training Guide.
- Interested in lightweight concept demos: use the standalone pages for text, audio, multimodal, or simple diffusion ideas.
- Interested in lower-level or partially retained surfaces: use the exploratory and validation sections in All Examples with the status labels in mind.