About Video checkpoint #11
Description
First, congratulations on the inspiration your work has given me. However, while re-debugging this section of your code, I encountered some issues. In your paper, you mentioned that the video modal pre-training files were obtained from the MAE-Face project. However, from what I can see, the pre-training files in this project appear to be only those that have undergone MAE pre-training. Could you please clarify whether you performed any fine-tuning or similar work afterward? When I attempted to adjust the single-modal settings, the training results were not optimal. If you could assist me in resolving this issue, I would be deeply grateful.