Hello,
I am trying to understand the meaning of the parameter "num_dec_block" that is set as 2 in your paper.
From what I understand it is supposed to be the number of decoder blocks.
I tried to find it in your code but it seems that it does not exists.
Also in the original code of perceiver io, the decoder only contains a single cross attention layer as decoder.
Could you tell me about this parameter and whether you tested to put more than a single decoder block ?
Thanks
Hello,
I am trying to understand the meaning of the parameter "num_dec_block" that is set as 2 in your paper.
From what I understand it is supposed to be the number of decoder blocks.
I tried to find it in your code but it seems that it does not exists.
Also in the original code of perceiver io, the decoder only contains a single cross attention layer as decoder.
Could you tell me about this parameter and whether you tested to put more than a single decoder block ?
Thanks