All WEKA packages have a lib folder in which dependencies (jar files) for that package are stored. DL4J pulls in many jar files (through the Maven pom.xml script) and right now all of them are naively copied to the lib folder. Because the DL4J classifier only exposes a small part of DL4J itself (just CNNs and MLPs), it seems weird to be carrying around more jars than we should. For example, jars like opencv-3.1.0-1.2-macosx-x86_64.jar or cleartk-opennlp-tools-2.0.0.jar. I'm not sure what the best way to start culling some of these jars, short of doing some sort of process by elimination or maybe asking the JVM what jars have been loaded at runtime. @agibsonccc do you have any suggestions?
All WEKA packages have a lib folder in which dependencies (jar files) for that package are stored. DL4J pulls in many jar files (through the Maven pom.xml script) and right now all of them are naively copied to the lib folder. Because the DL4J classifier only exposes a small part of DL4J itself (just CNNs and MLPs), it seems weird to be carrying around more jars than we should. For example, jars like opencv-3.1.0-1.2-macosx-x86_64.jar or cleartk-opennlp-tools-2.0.0.jar. I'm not sure what the best way to start culling some of these jars, short of doing some sort of process by elimination or maybe asking the JVM what jars have been loaded at runtime. @agibsonccc do you have any suggestions?