StarNet++ Will Not Run on ANY of my Linux (AMD64) Systems

[JonCarleton]

I downloaded from SourceForge the StarNet++ Linux tar.gz file onto a few Linux Ubuntu laptops and desktops (all AMD64 processors) I have and crashed and burned with every attempt.  I am running Ubuntu 20.04 on most of them, but I did try on a 19.x version for grins (also a fail).  When I untar/zip the file and run the test shell script, I get a core dump.  For that matter, I get a core dump if just try to run the executable.

I am not running PixInsight and just want to run the program stand-alone.  If I must, I'll load some other version of Linux on a machine to make it happen.  Anyone have a working version running on some version of Linux that could advise the specs required.  There doesn't seem to be much info on this issue in the mind of Google at present.  If I had to guess, I'd say it is probably a library version issue somewhere.  I have not had time to plow through the core dump to try and decipher that kind of issue and hope to fine a reasonable answer here.

Edited February 22, 2021 by JonCarleton

[JamesF]

I can try it here if that would be helpful.  Probably would be useful if you gave me the exact link you downloaded from, just so we can be certain that I'm running exactly the same thing.

James

[JonCarleton]

The big green Download button on https://sourceforge.net/projects/starnet/ is what I used.  It downloaded a StarNet_Linux.tar.gz that I unfolded into a directory that contained the executable and some shell scripts with demo files to test operation.

Thanks!

[BCN_Sean]

How are you running StarNet, are you dragging the StarNet binary / shell script on to the terminal or going in to terminal and cd-ing to the StarNet path then something like "./rgb_starnet++ in_image.tiff out_image.tiff"?

I know that may sound a bit daft, but I was ripping hair out the other day with nearly the exact same problem but on MacOS (I've been using StarNet a while and couldn't see this wood for the tree) and I was forgetting to cd to the directory and instead just trying to drag and run.

[JamesF]

It appears to work fine for me on an Intel Core i5 machine with 16GB RAM.  There appear to be very few external dependencies and they're very common ones, so I doubt you're missing any of those.

I think the first thing to check would be that your CPU supports the AVX instructions.  You should be able to do that just by running:

$ grep avx /proc/cpuinfo

If that produces any output then it looks like you should be ok for that.  If it doesn't then you'll probably need copies of the tensorflow libraries that don't require AVX support.

James

[JonCarleton]

@BCN_Sean:  Nope, got that right.  I'm an old-school terminal kinda guy. cd is my friend.  The only thing I drag is my feet.

@JamesF:  Yep, no avx support.  I knew it smelled like a library issue.  Odd that -none- of my hardware has it.  Still, I don't generally buy hardware with an eye for graphics support.

 

[BCN_Sean]

@JonCarleton which distribution of StarNet are you using? Is it the one from Github or the one from Sourceforge?

As a bit of a thought on this today, I went through on another machine and set it up from clean using the Github version, and found that at least one of the libraries needed (scipy) has had a couple of bits deprecated which in turn causes StarNet to fall over.  The pre-compiled (this is on MacOS, though) binaries from Sourceforge running over Python 3.8 didn't have any sort of issue there.

Would it be possible to post a grab of what the output is when it's dumping?

As for being old school on the coal face, I completely get that and it surprises me how many people think it's some dark art now-a-days!

[JonCarleton]

I may try to look at the core dumps.  In the past with other issues, I have not found raw core dumps much help.  

The version I am using is the compiled Sourceforge for Linux version.  I haven't tried the raw python version from github...but I may this weekend.