public static load(path as Path, maxBytes as ULong = 2147483648ul) as Sequential<T>
Restores an independent model, including all saved training state. The static precision must equal the file dtype; no implicit conversion is performed. Limits, operator configurations, shapes and SHA-256 are checked. No arbitrary code or custom constructors are loaded. Continuation is between fit calls; fit's seed still restarts its shuffle. Device defaults to auto.
| Parameter | Description |
|---|---|
path | Local KNN checkpoint. @param maxBytes Maximum tensor bytes, 2 GiB by default. |
Built model; compiled if training state was saved.
restored = Sequential<Float>.load(FilePath("MnistCNN.knn"))