Best way to use models mapped to tutorials

[dbericat]

Hi @wyli @Nic-Ma @ericspod ,

As we think about how to put together MONAI Deploy examples and tutorials, we'd like to reuse some of the models mapped to these tutorials.

In particular, we couldn't find the model for https://github.com/Project-MONAI/tutorials/blob/master/3d_segmentation/spleen_segmentation_3d.ipynb

New area for me, so how does this work?

Thanks in advance!

[Nic-Ma]

Hi @dbericat ,

Excited to see the MONAI deploy progress.
May I know what do you mean could't find the model? You want the network or trained model weights?
The network is UNet in MONAI source code.
We don't maintain some trained weights so far.

Thanks.

[dbericat]

I think we were trying to find the trained model weights. Otherwise we can't package it in a MONAI Deploy App. @MMelQin @gigony @rahul any comments?

[rijobro]

the tutorial uses torch.save, so as long as you run the tutorial, you'll save the network weights under os.path.join(root_dir, "best_metric_model.pth"), where:

directory = os.environ.get("MONAI_DATA_DIRECTORY")
root_dir = tempfile.mkdtemp() if directory is None else directory

so make sure MONAI_DATA_DIRECTORY is set as an environmental variable.

[dbericat]

Thanks Richard!