etienne87

Ah! Agreed. So like we should share a common replay memory to all threads and we recompute forward before large batch update?

hello! where is the draw demo, is it the `main` of the script? i do not see any dijkstra/ astar call to networkx

Thanks a lot for reporting the bug! this is a real bug showing an error in design/ implementation from me, because after exhausting real data from stream dataset workers, somehow...

sure, will do it today

Hi @kaixinbear. The original intent of the repository was to stream from uninterrupted stream readers (e.g hard to seek-to video, online data, etc.) , so by definition 1 process =...

I am very interested in this issue as well as i think moving to GPU makes a lot of sense. However i guess we need to keep at least one...

Ah right! i did not realize there was already cropping happening inside nnUNetDataLoader with the "initial_patch_size" and "final_patch_size" arguments. I will try your variant and benchmark agains't the dumb solution...

I wanted to reproduce your idea on a fork of nnUNet [here](https://github.com/etienne87/nnUNet/blob/feat/gpu_augmentations/nnunetv2/training/nnUNetTrainer/nnUNetTrainer.py#L1491) i have done a fork of batchgenerators-v2 where if fixed the spatial augmentation [code](https://github.com/etienne87/batchgeneratorsv2/blob/feat/gpu_augmentations/batchgeneratorsv2/transforms/spatial/spatial.py#L330) to take into account...

Hello @LukaslWeber i did a P.R [here](https://github.com/alanqrwang/keymorph/pull/24). What i did: - swap the lines as you say - modify torch.bmm for torch.einsum in "get_inverse_transformed_points" (it does not work otherwise because...

you are right the `test_affine_2` does not pass, but for me it gives me error: `FAILED test/test.py::TestAffineAligner::test_affine_2 - torch._C._LinAlgError: linalg.inv: (Batch element 0): The diagonal element 3 is zero, the...