`state`: a dictionary which is carried through the graph construction `opts`: a dict of opts for the current layer

Returns a deep copy of an autograd variable's data. A number of PyTorch operations are perfomed in-place, which makes graph comparison non-tr

print size summary, perform some sanity checks and (by default) pops the leading size

restructure models to be consistent for easier processing

Convert a PyTorch Tensor to a numpy array. (changes the order of dimensions to [width, height, channels, instance])

Convert x to a Python list. x can be a Torch size tensor, a list, tuple or scalar

Flattening is done in the network class, rather than in the moudles with pytorch, so we need to 'catch' this event and reproduce it in th

check for the presence of the LambdaMap block used by the torch -> pytorch converter

check for the presence of the LambdReduce block used by the torch -> pytorch converter

r"""Inception v3 model architecture from `"Rethinking the Inception Architecture for Computer Vision" <http://arxiv.org/abs/1512.00567>`_. Ar

check for the presence of the LambdReduce block used by the torch -> pytorch converter

configure parser for shared keyword arguments