Thank you for sharing your work. Sincerely want a clarification about mask radio. I refer the closed issue "About 'compute_trick_metric'", adjusting seed to 42 and mask ratio to 0.5 for msrvtt, but R1 still can't not reach to 48.2 metioned in the paper.
Thank you for sharing your work. Sincerely want a clarification about mask radio. I refer the closed issue "About 'compute_trick_metric'", adjusting seed to 42 and mask ratio to 0.5 for msrvtt, but R1 still can't not reach to 48.2 metioned in the paper.