Huadong Liao

A same issue [here](https://github.com/naturomics/CapsLayer/issues/5)

The following table is the approximate training time | num x GPU device | other configurations | speed(min/epoch) | |:--------------------:|:--:|:---:| | 1 x 1080ti | default | 3 | |...

You can check it with the command `nvidia-smi` to see how many GPU the program is using under Linux. For windows user, find it out in Task Manager(任务管理器). But by...

BTW, which tesla GPU and CPU u'r using?

Both should work. Training for 30min on a 1080ti is enough to have a good result, but on CPU, it might take for hours. Prashant Desai 于2019年2月27日周三 下午9:06写道： > So...

@jackalwoods Looks like it's your tf env problem. please make sure your tf is installed correctly or a right version (the code was developed with tf 1.3, a newer one...

Hello guys, I found a solution for this 'not invertible' problem. During the training, the weighs of invertible 1x1 conv keeps increase to balance the log-determinant terms generated by invertible...

It doesn't matter. Squeeze2d is followed by 1x1 conv layer. After squeezing, both implementations have the same data within each channel group but a different order, they look same for...

Hope my explanation will help you: 1. as the latent space is constrained to this Gaussian distribution p(z; mean, scale), in order to calculate logp(z), we need z, mean and...

pz.logp(z) calculates p(z)~N(z;mean, scale), not p(z)~N(z;0, I), so there are no more transformation.