Videos of all of the datasets

[EsamGhaleb]

Dear all,

First of all, thanks a lot for the amazing work and the benchmarks. The dataset on the Huggingface website contains text, labels, and dimensions. How can I access the audio-video files for each of the dataset rows? This would be a big contribution.

Best,

Esam

[HanleiZhang]

Hi, thanks for your attention. You can obtain audio-video files from the compressed archive file (xxx.tar.gz) in the directories of each dataset.

[EsamGhaleb]

That is great.

Do you have a requirements file for the Python packages used? Also, instructions on how to use the datasets for the different setups (zero-shot, fine-tuning, and instruction fine-tuning), for the models used?