Hello i am wondering how to construct the bias tensor for the convolution layer? What parameters should i use for n c h w?
(1, c, 1, 1)
where c is the filter’s first dimension, i.e. the number of channels in the output (a.k.a. “k”)
1 Like