CogVideo
CogVideo copied to clipboard
position embedding
I noticed that position embedding is 3d_sincos_pos_embed, but the position embedding mentioned in the paper is rope. Will the code using rope be released?
Better model will come soon