A upgrade to PytorchLightning 2.0 and PyTorch 2.0

[YuyangXueEd]

Hello all!

Since PyTorch and PytorchLightning is upgraded to 2.0 and above, the fastmri_examples code such as unet is now deprecated. I am wondering that whether there is a plan to update the code to compatible to 2.0.

There are some changes like:

• used LightningDataModule.parse_argparser(), LightningDataModule.from_argparse_args() and Trainer.from_argparse_args() method now switch to using LightningCLI.
• had any logic except reducing the DP outputs in LightningModule.validation_step_end hook now port it to LightningModule.on_validation_batch_end hook
• had any logic except reducing the DP outputs in LightningModule.test_step_end hook now port it to LightningModule.on_test_batch_end hook
• no resume_from_checkpoint for the Trainer but changed to ckpt_path
• ...

I will try to pull a request when ready.

Please:

• [*] Check for duplicate requests.
• [*] Describe your goal, and if possible provide a code snippet with a motivating example.

[mmuckley]

Hello @YuyangXueEd, are you still working on this?

[amirshamaei]

@mmuckley Hi, I worked on this. However, it wasn't possible to make it using the same structure. I added a yaml file as a config file. I have tested only on single-coil knee images