CUDA out of memory

[iBobbyTS]

Is there a way to fix this? I'm using Tesla P100 with 16GB of Vram on Colab, when I tried to process 4K video it will give me this error RuntimeError: CUDA out of memory. Tried to allocate 8.10 GiB (GPU 0; 15.90 GiB total capacity; 9.09 GiB already allocated; 6.09 GiB free; 9.11 GiB reserved in total by PyTorch) I tried to change the code to process thr channels(RGB) separately, but it did,'t work. Is there other way I can process large video with limited Vram? Thank you.

[iBobbyTS]

In the colab notebook in README, it says only 720p can be processed with 16GB of vram, but I found everytime I start, it can interpolate 1 frame, then it stops because vram is not enough. So if the video is 1080p, I make it interpolate one feame each time.

for i in range(png_generated_count-1):
  frame_interpolate(i+1, i+2)

I made the "frame_interpolate()" function before and left "start_frame" and "end_frame" as arguments. This works but it takes much longer then process all frames at once because it needs to initialize it every time. But 4K never works

[tanerbajrovic]

Hey, would you mind providing me with the customized version you made? I want to interpolate a 1080p video and keep running into the same kind of error with CUDA

[iBobbyTS]

Hey, would you mind providing me with the customized version you made? I want to interpolate a 1080p video and keep running into the same kind of error with CUDA

I'm still adding functions, here's a version I archived, it should work. DAIN_Rewrite Here's a link to the tar that has python and dain built. You might need to hange the code little bit for the path of tar and video input. Built_files