Hello, I am trying to set up a Julia kernel in mybinder.org with a few packages. Here is the GitHub repo for the kernel. I find the precompilation failing due to the process overrunning the maximum amount of RAM (2GB) provided by the Binder.
How can I reduce the usage of memory during the precompilation process?
2GB is ridiculously little. I doubt that mybinder container could be useful with so little memory. Get a bit more, perhaps?
I’m not sure but maybe this flag
--heap-size-hint=<size> Forces garbage collection if memory usage is higher than that value.
The memory hint might be specified in megabytes(500M) or gigabytes(1G)
when starting Julia can help?
Thanks a lot! May I please know how much memory would be recommended?
I find that 8GB is a decent lower bound.
On Linux you can use zram to get 50% more RAM for free… And you can always use a swap file if you are patient…
I would try requesting mybinder for memory limit increase, but I am doubtful they will support it. What would be the recommended alternative? A private BinderHub deployment?
@ufechner7 Since mybinder is a public cloud service, I am not sure if I can fiddle with the configuration myself. I will give @Tortar’s solution a try!
I eventually made up a postBuild file for binder to precompile the required packages with “heap-size-hint=1G” , and made a minimal Julia environment to begin with! This seems to help! Thanks everyone !
@pvarshinee, I would greatly appreciate it if you could show me how to generate a postBuild file for binder to precompile packages. Thanks a lot!
A possible solution specific to precompilation is this enviroment variable: JULIA_NUM_PRECOMPILE_TASKS.