No Positional Embedding for Temporal Attention?

Hi,

I noticed that eventually, you guys didn't apply any positional embedding for each frame in temporal attention layer, but you have the RelativePosition implemented. I am just curious why you decide not using any positional embedding in temporal attention layer?

Thanks!