meganset

Infogan batchnorm2d layers have eps=0.8 ?

2

infogan.py creates most batchnorm layers with a 2nd positional parameter of 0.8, e.g. in the generator: ``` self.conv_blocks = nn.Sequential( nn.BatchNorm2d(128), nn.Upsample(scale_factor=2), nn.Conv2d(128, 128, 3, stride=1, padding=1), nn.BatchNorm2d(128, 0.8), .....

Unable to run minimal example on Ubuntu

4

My first try failed because a numpy .h file couldn't be located: ``` g++ minimal.cpp -std=c++11 -I ~/anaconda3/include/python3.7m -L ~/anaconda3/lib -lpython3.7m In file included from minimal.cpp:1:0: matplotlibcpp.h:17:12: fatal error: numpy/arrayobject.h:...

The method for calculating a running standard deviation is not giving the correct answer (in utils.py) e.g. for one big batch: ``` >>> import torch >>> import torchvision >>> import...