Open
Description
The paper mentions that the time to train the pipeline is about 6 hours on a single RTX 4090 . Is this time for end to end or just the strands reconstruction stage ?
I tried training it on a video of (1058 x 1887) on a single A5000 and it takes significantly longer than 6 hours (~17-18 hours) for the entire pipeline . May I know what was the resolution of the video you trained on ? I don't believe such a huge difference can be attributed to different GPU models. Maybe I am making a mistake somewhere
Metadata
Metadata
Assignees
Labels
No labels