write_record

write_record(spec, name, *, derived=None, history=None)

Write <output_dir>/<name>/run.json and return its path.

derived is whatever the run worked out that the specification does not already say - fitted anchors, the target survey. For a segmentation run it is usually empty, and that is the honest answer rather than an omission: the specification already carries the colormap, the input shape and the window, so re-running it reproduces the model.

Parameters

Name	Type	Description	Default
spec		The specification the run was built from. It is written out whole, so the record is enough to reproduce the run without the file it came from.	required
name	str	The model’s name, which is also the directory the record goes in.	required
derived	dict[str, Any] | None	What the run worked out for itself, if anything.	None
history	Any	The per-epoch numbers, as the trainer produced them.	None

Returns

Name	Type	Description
	Path	The path written: <output_dir>/<name>/run.json.

The record stamps pyplatypus.__version__, which is why that literal and the one in pyproject.toml are held together by a test: a release reporting the wrong version writes wrong provenance into every file it produces, and those files outlive it.

No doctest: this writes a directory and wants a whole specification to do it, and the engine’s own tests cover it with one.