search

k4yt3x / video2x

A machine learning-based lossless video super resolution framework. Est. Hack the Valley II, 2018.
https://video2x.org
GNU Affero General Public License v3.0
10.69k stars 1k forks source link

Unable to fully process mpeg2 encoded MKVs #842

Closed stevezemlicka closed 4 weeks ago

stevezemlicka commented 1 year ago

I am able to run the docker container but for some reason, it never gets past 80% for short videos and 81% for long videos. I've tried multiple input videos and everything seems to proceed well except that it stops at precisely 80% on nearly all. The only exception is when the input video is very long (~45min). In those cases, it does the same except at 81%. The time continues to count and it never errors out but no further progress is made. And while it seems to hold on to the VRAM, GPU utilization goes down to normal indicating no further GPU processing is occurring.

It seems to me that there is some process, perhaps encoding or something similar that occurs at around 80% that is failing. It's not time based since long videos will proceed normally for 20+ hours until they get to 81% and short videos will run normally for an hour or so until they get to 80%.

I've tried the beta6 image as well as the beta5 with no changes. I'm on a relatively recent nvidia 520.56.06-2 driver. I did try podman but was unable to get this to work with nvidia (though my laptop seems to run it with its AMD chipset just fine).

Any thoughts on what might be happening here? Is there any additional info I can gather to help diagnose? Is there a way to increase the verbose level of the terminal output? I saw someone posted strace logs but I don't know how to gather those. Any suggestions or help would be appreciated.

stevezemlicka commented 1 year ago

I tried the Google Colab interation of this with the same result. This leads me to think there is an issue when this is processing mpeg2 encoded MKV files. Is it possible that there is something preventing this from working with mpeg2 MKVs?