Is your feature request related to a problem? Please describe.
ZooModels from Deeplearning4j such as UNet, NASNet etc. have a graph-based, non-linear architecture.
Since the Dl4jMlpClassifier holds all layers in a linear/sequential manner, it is not capable of parsing these models.
Idea
Create a Dl4jZooModel classifier that mimics the Dl4jMlpClassifier behavior and can load arbitrary zoo models but does not allow to change the model architecture, except for the output and input.
Is your feature request related to a problem? Please describe. ZooModels from Deeplearning4j such as
UNet,NASNetetc. have a graph-based, non-linear architecture. Since theDl4jMlpClassifierholds all layers in a linear/sequential manner, it is not capable of parsing these models.Idea Create a
Dl4jZooModelclassifier that mimics theDl4jMlpClassifierbehavior and can load arbitrary zoo models but does not allow to change the model architecture, except for the output and input.