Tutorials

These tutorials walk through the typical workflows on top of the rddac Python surface, from a first download to a PyTorch training loop. Each page is short and code centric: read top to bottom or jump to the section that matches the task at hand.

1. Getting Started: Install, download the small bundle, load the manifest, open one experiment, render the first plot.
2. Build your own view: Use rddac.add_view to compose a custom Croissant RecordSet from the field map.
3. PyTorch training: Stream a view through RDDACDataset, batching with DataLoader, multi worker sharding, DDP, and a reproducibility note.
4. Visualization: Scan images, point clouds, force curves, and sensor traverses on top of rddac.open_h5.
5. Loose HDF5 recipe: Iterate over loose .h5 files after rddac download --extract --remove-zip.
6. Streaming and numpy export: Iterate any view with rddac.streaming.iter_view (no PyTorch, no mlcroissant FileSet walk), and materialise it as flat .npy shards with streaming.export_to_numpy for fast training reads.

If you already know the DDACS tutorials: the flow here is identical by design. The rddac public surface mirrors ddacs, so everything you learned there ports by swapping the import.

Prerequisites

pip install rddac
rddac download --small