How to Extract Video and Audio Features Using VideoLLaMA2?

Hello,

I have a set of pre-chunked video clips, and I would like to use VideoLLaMA2 to extract video and audio feature vectors for each chunk. However, I couldn't find clear documentation on how to achieve this using the pre-trained model.

Could you provide guidance on:

- Extracting video features (frame embeddings) from the vision encoder used in VideoLLaMA2?
- Extracting audio features?
- The recommended process for batch processing multiple video chunks efficiently.

I would appreciate any code references or scripts within the repo that facilitate feature extraction. I appreciate any help you can provide. Thanks, in advance!