Opening local video triggers Killed

[tteresi7]

The following code triggers an exit with the message Killed.

import torch

from cotracker.utils.visualizer import Visualizer, read_video_from_path


video = read_video_from_path('path/to/local/video.mp4')
video = torch.from_numpy(video).permute(0, 3, 1, 2)[None].float()

The message is triggered after the torch call.

[nikitakaraevv]

Hi @tteresi7, you need to provide a path to your video

[tteresi7]

Hi @tteresi7, you need to provide a path to your video

I mean obviously. I was simply providing example code. The problem is the float call absolutely rips through memory. There's probably more memory efficient ways to handle the video in the call. This snippet was taken from the examples you provided. Basically right now, it opens the video, stores it in memory, then creates a torch copy of it and then rips through memory going to float. I got it to run using swap memory, but then when you go to write the video per the snippets in the repository, it creates yet another copy and rips through the remaining swap memory.