New answers tagged implementation
0
With the existing frameworks PyTorch, Tensorflow you can easily implement this functionality, by keeping some of the intermediate computations inside forward or call method and passing them as an input to the given layer. For example:
x[i] = layer[i-1](x[i - 1])
...
x[j] = layer[j-1](x[j - 1] + x[i]) # resnet-like skip connection
or
x[j] = layer[j-1](concat(...
Top 50 recent answers are included
