Any reason why BinaryObjectVectorizer is a final class? It would be convenient to be able to extend it for specific subclasses on pre-defined features.
On 22.06.22 18:32, [email protected] wrote:
Responding to my own question in case someone else has a similar question: The solution is to use BinaryObjectVectorizer<K> which gets the labeled features from the binary object! *Question:* Why is BinaryObjectVectorizer a final class? It would be convenient to be able to extend it for specific subclasses on pre-defined features. On 20.06.22 13:22, Thomas Kramer wrote:I am storing my ML training data in a cache with binary data: IgniteCache<UUID, BinaryObject> cache = ignite.cache("cache").withKeepBinary(); I can't seem to understand how to use the Ignite ML classes like Preprocessor and DatasetBuilder with this cache? Do I first have to create another IgniteCache<UUID, MyClass> or Map<UUID, MyClass> and copy the data from the original cache? Should I better not use BinaryObject for ML caches? Or would I write my custom vectorizer that creates MyClass instances of the binary objects?
