I can’t find any info on this - is there a minimum memory / cpu requirement for being able to run DL4j / ND4J?
I built a simple system which only uses ND4J for vector arithmetic - nothing else. When I run that system on my dev machine (64 gb ram) it runs without an issue.
When I try to deploy that on a server w/ 2 GB ram, it gets stuck on startup and seems to get stuck forever. I never see the initial ‘Backend used: CPU, OS: Linux’ messages which ND4J prints on startup.
What’s the minimum spec I’d need on the server?
This is subject to your workload size and how you’re running dl4j. Models can be anything from megabytes to gigabytes.
It’s hard to tell what you are trying to do.
I would recommend learning a bit about application profiling:
There really isn’t a simple answer for this. It depends how you’re running dl4j, eg is it an application server, a standalone uber jar? If you want us to be able to answer any question like this, we’d need to know more about the specifics of your setup.
From a simple standpoint, I wouldn’t try to deploy any server with just 2g of ram. Generally I would recommend deploying on at least something resembling the specs of a cheap laptop if you want to deploy anything for machine learning. You don’t need a 64 gb server, but a mid size machine at least is recommended if possible.
OK, this was due to an application issue. DL4J is running fine on 2g ram. No models loaded, only linear algebra operations being done 
1 Like