Use ants image in docker to complete registration

[Schwarao]

Use ants image in docker to complete registration

[cookpa]

SyN requires significantly more memory, which can lead to the process being killed if there is not enough available

[Schwarao]

SyN 需要更多的内存，如果没有足够的可用内存，这可能会导致进程被终止

My memory is sufficient, and it can be successfully run by calling the script file locally on my computer. Why can't it be done in docker?

[Schwarao]

Is there any master who has the experience to tell me, this is very urgent.

[Schwarao]

I'm using antsx/ants images

[Schwarao]

Or can anyone tell me about antsx/ants and antsx/antspy in dockerhub: latest
How do I use the containers created by these two images?

[stnava]

Try reading https://docs.docker.com/config/containers/resource_constraints/

[cookpa]

If you're using docker desktop, there is a memory limit

https://docs.docker.com/desktop/settings/windows/#:~:text=Memory%20limit.,decrease%20it%2C%20lower%20the%20number.

[cookpa]

https://docs.docker.com/desktop/settings/mac/

https://docs.docker.com/desktop/settings/windows/

https://docs.docker.com/desktop/settings/linux/

[Schwarao]

It shouldn't be a problem of insufficient memory, could it be that there is a problem with my shell script configuration?

I still have 18GB of RAM available on my computer when I'm not running programs

[cookpa]

I don't recognize these graphs, but the top right one appears to show 11.68 Gb of memory available.

[Schwarao]

That's right, so why is the program being killed?

[stnava]

the docker container does not have enough memory. that's different than the machine on which the container is being run.

[Schwarao]

ok,i will Use it on a different computer

[Schwarao]

Thanks to both of you, I now put it on the linux host to run docker but I found that linux will use more memory, for example: windows only needs more than 10 G, while linux needs more than 100 G, why is this?

[cookpa]

What are the dimensions of the images?

[Schwarao]

moving：

fixed：

[cookpa]

So 400x274x460 and 456x320x528. Those are big images, and the registration needs to store those plus the associated transforms. Simplest solution to reduce footprint would be to run with --float 1.

[Schwarao]

ok，i will try it ,thank you very much.

[stnava]

probably worth trying to blur + downsample