December 2018
Intermediate to advanced
158 pages
3h 58m
English
To get a better understanding of this, let's look at what precisely a computational graph is. We can draw the graph for the function we have been using so far as follows:

Here, the leaves of the graph represent the inputs and parameters of each layer, and the output represents the loss.
Typically, unless retain_graph is set to True, on each iteration of an epoch, PyTorch will create a new computational graph.
Read now
Unlock full access