public static read(path as Path, maxBytes as ULong = 2147483648ul) as ModelCheckpoint<T>
Loads and validates all architecture, ranges and checksums before publication. Limits and expected parameter shapes are checked before tensor allocation. The returned model is independent; no existing model can be partly mutated.
| Parameter | Description |
|---|---|
path | Local KNN 1.0 file. @param maxBytes Maximum tensor bytes, default 2 GiB. |
Private restored model state with current optimizer/random state.
checkpoint = ModelArchive<Float16>.read(Path("half.knn"))