Plans for MeMViT code release and a question on training

Hi,

Do you have plans for releasing the code for [MeMViT]()?

And I have a specific question about the sequential training with MeMViT on AVA: How do you do the batching? for a batch size of 128, do you train with 128 unique AVA videos sequentially at the same time until they are finished, and move on to the next 128 (there are only 239 videos in the AVA training set)?

Thanks!

Best,
Junwei