Saving runs#

The hybrid_aif.data package reads and writes episodes on disk in two formats, a folder of arrays and HDF5. It is pure numpy. A run’s record can be written with a writer and read back with load_dataset; the arrays it returns are what the replay facility takes. The neural training buffers read the same formats.

Reading and writing#

load_dataset

Open path as a ReaderDataset, through the registered format named format or, when it is None, the first registered format that detects the path.

detect_format

Return the first registered format whose detect matches, else None.

Dataset

Base class for episode-based datasets.

Writer

Streaming episode writer — append episodes, finalize on exit.

Format

Declarative format spec.

For authors of a new format#

Reader

Random-access episode reader over flat per-column storage (a protocol).