There is a limitation with Java that a method cannot have more than 64K btyes information (https://stackoverflow.com/questions/2407912/code-too-large-compilation-error-in-java). Many Random Forest and Boosted Tree models exceed this limit when deploying the models as Java. When attempting to compile a Java model that exceeds the 64K an error exception is thrown, "code too large" and the code is not compiled. This makes it so that the code has to be broken into small chunks manually. It would be very nice to have this automated.
Please vote for this enhancement. It will automate the java deployment so manual changes do not need to be made so the code will compile.