use MONAI Core for Swin UNETR training and evaluation on HiperGator

This directory hosts scripts for runnning Swin UNETR on HiperGator, adapted from Swin UNETR BRATS21 MONAI research contribution.

Note

This tutorial assumes you have downloaded the repository monai_uf_tutorials following this section.
If you have no experience running MONAI Core using Singularity as container runtime on HiperGator before, I strongly recommend going through tutorial monaicore_singlegpu and making sure it's working before moving on to this tutorial.
If you have no experience running distributed training with MONAI Core on HiperGator before, I strongly recommend going through tutorial monaicore_multigpu and making sure it's working before moving on to this tutorial.
In all following commands, replace hju with your HiperGator username; change the path to files according to your own settings on HiperGator.
In all following SLURM job scripts, alter the #SBATCH settings for your own needs.
Please read the comments at the beginning of each script to get a better understanding on how to tune the scripts to your own needs.

Read the original description Swin UNETR BRATS21 MONAI research contribution. Download and put required data on your HiperGator storage space.
Go to directory /monai_uf_tutorials/monaicore_swinUNETR/
```
cd ~/monai_uf_tutorials/monaicore_swinUNETR/
```
To launch training/inference on a single GPU,
```
sbatch launch_single.sh
```
To launch training/inference on multiple GPUs,
```
sbatch launch_multi.sh
```
Note
- Launch scripts launch_single.sh and launch_multi.sh includes many training options, e.g., whether use AMP, whether use gradient checkpointing, train from scratch or train from a pretrained/checkpointed model. Please read comments in the scripts.
- In training scripts e.g., main.py, trainer.py, some commented out code is from original research-contribution scripts and the replacement code is usually underneath, some is for debugging purpose. Besides, there're other code modification.
- When using validation in training with gradient checkpointing, you might see warning output None of the inputs have requires_grad=True. Gradients will be None which can be ignored.