Hi,
so i am trying to make work a convolusionnal SNN with axonal delay, with pytorch.
But i don’t find any info on internet about how to code an axonal delay.
Here i am trying to just make a kind of circular buffer. But i don’t want it to be like a list, because i want to add to it, a tensor like this 10, 0, h, w, the input, a tensor like this h, w, multiplyed by a another mask tensor, like this 10, 0, h, w, filled with ones sparsely distributed to make the axonal delays.
So i need this circular buffer to be a tensor and not a list, because i need to apply to it tensor operations.
Maybe i am doing wrong ? Is there a better way to do it ? I am looking for quickness. Or do you know exactly how am i supposed to make this circular buffer ? Should i use torch.roll ?
Thanks for your help!