Cuda out of memory.

[TomassMa]

Hey, i have similiar issue to #23 , I'm trying to use slightly altered (only a compatibility changes) eval_generic to segment some videos from this dataset. The problem is I keep running out of memory, im using google colab workspace and getting RuntimeError: CUDA out of memory. Tried to allocate 2.97 GiB (GPU 0; 14.76 GiB total capacity; 9.57 GiB already allocated; 319.75 MiB free; 13.13 GiB reserved in total by PyTorch).

It happens at line: processor.interact(with_bg_msk, frame_idx, frames_with_gt[i+1]+1, obj_idx)

I've already increased the mem-every argument, but it didn't really help, can you by any chance think of something that i can change ?

[TomassMa]

Nevermind, i found the cause of my issue, it was similar as in #3.

Anyway thank you for your work :)