Possiblity to download the pretrained models and weights and add in container

[sulantha2006]

Is there a way to download and store the pre-trained weights and the models other than using the URLs? When running multiple jobs in clusters, all jobs will have to download the files and some gets errors from the figshare site for multiple concurrent downloads failing the jobs.
If the models can be added to a shared location the above issue can be mitigated.

[ntustison]

Did you look at antspynet.set_antsxnet_cache_directory?

[sulantha2006]

Yes. That is one approach, but I am looking a way to download all the files directly from GH or some other location.

[ntustison]

Sure, you can download all of the weights/models from Figshare, see get_pretrained_network.

[cookpa]

If you want to manage a shared antsxnet_cache_directory on your cluster, this script might be useful

https://github.com/cookpa/ANTsXContainers/blob/master/docker/ANTsPyNet/get_antsxnet_data.py

I mostly work with containers on the cluster because I've found it the only way to reliably keep the code and models consistent across users. There's docker files to build with and without included data at https://github.com/cookpa/ANTsXContainers/tree/master/docker/ANTsPyNet