myron

> For what it's worth I have this code for defining a gaussian kernel of arbitrary size: > > ```python > def gaussian_kernel(*shape, device, dtype, sigma=2.0): > dims = [torch.arange(-s...

is this hybrid loss the same as Dice + CrossEntropy loss? https://github.com/Project-MONAI/MONAI/blob/dev/monai/losses/dice.py#L609

> Sure, I don't think it's a blocker, if we use torchrun and the import will be within each process. Lets follow @YanxuanLiu's process to upgrade these... @wyli @Nic-Ma I...

This is great effort, to interpolate only once, it would avoid the repeated interpolation artifacts and should speed up the transforms pipeline, and use less memory, in theory. At the...

@atbenmurray thank you for the reply. I understand the idea of lazy resampling and fusing transforms. And I understand the idea of hiding this from the user, to simplify things...

We also added SlidingWindowInfererAdapt class to automatically manage memory without OOM, which you can use as a replacement for SlidingWindowInferer.

@razorx89 thank you for a great example/issue and the code with evaluation. So, SlidingWindowInfererAdapt() simplifies memory management by attempting to run optimally within GPU budged, and it will use try/except....

thank you for the response, I can see it's useful for your case, and some other certain cases. I will let other people to comment.

hi @atbenmurray I don't understand the issue well ``` . Noise will undergo spatial distortion that was not intended . Non-lazy spatial transforms will output the wrong result ``` which...