Halite Home

How to handle large compiled artifact


#1

Hi,

I am looking into trying out reinforcement learning. The thing is, the only library in Java that has some good documentation is deeplearning4j. And adding this dependency will generate me a jar file of ca. 150 MB which is not feasible for upload I guess.
Would it be possible to write an install.sh that just downloads my jar file from a different place in the setup phase which is then used instead?

Any different options?


#2

Can you serialize the trained model and use it without the learning library?


#3

TBH, I have no idea, but I dont think so.

Is this easier in python or a different language? Its not that it has to be Java, java is just the one I am most comfortable with.


#4

We can look into supporting larger file sizes. No promises though.


#5

I thought about it some more. The installation gives me 10 minutes time to setup. I guess it is ok to just download my artifacts from somewhere in this time, including the sources and the libraries, or maybe even the whole jar and overwrite the existing one. Does that make sense?