Wei Wu

@xslittlegrass yes, you are right, the later is the old server. would you fix this and give a pr? thanks.

solved this problem by: ```pip uninstall prettytensor``` ```pip install prettytensor==0.6.2```

solved this problem with suggest form @pmiller10 : ```pip uninstall prettytensor``` ```pip install prettytensor==0.6.2```

just use tf version `r0.9rc0-devel` from readme: https://github.com/openai/InfoGAN#running-in-docker

yes, you're right. because no align in training. you man consider using align during training.

@kaishijeng not have one currently, welcome to contribute for this, it can borrow some code from lfw.py

hello, do you use my face branch of mxnet(https://github.com/tornadomeet/mxnet/tree/face)? moon operator files is in https://github.com/tornadomeet/mxnet/blob/face/src/operator/

@ZHAIXINGZHAIYUE it should be the Eq.(4) and set Ti+ = Ti- = 0.5, as paper describe in page 12.

sorry for not familiar with android. the number parmaters in src_dist.txt should be the class number. so could you print data.shape_[1] and src_dist_.size(), and provide more info?

i don't know, because no detail about your experiment.