How to Configure Multiple Video Training in the YAML File

[euminds]

I am currently working on a project that involves training with multiple videos. However, I am unsure about the correct configuration settings in the YAML file to support this. Specifically, I need guidance on the following points:

Dataset Type Configuration:
How should the videos be structured within the folder? Should each video have an associated text prompt file (e.g., video.mp4 with video.txt), or can a fallback prompt be used for all videos?

Path Settings for Multiple Videos:
Should the train_data.path be set to the folder containing all videos?
If using multiple videos, how should single_video_path and single_video_prompt be handled? Should they be removed or left blank?

Best Practices:
What are the recommended practices for setting fps, frame_step, and n_sample_frames when dealing with multiple videos to balance between training efficiency and memory usage?