Is this implementation finished?

[mjack3]

Hello @chrischute, I hope you are good.

I would like to learn from the FLOW ++ paper. Unfortunately, the original implementation is unintelligible to me. I would say it's because I'm not a tensorflow programmer but, the truth is that the code is horribly organized. So thank you for your work in translating the original implementation to Pytorch. I would like to ask you if you think there are any points in the code where you think it might be wrong. Just to be sure that the code I will study here is ok.

Regards

[mjack3]

For example. Maybe this code wrong regarding the original version?

https://github.com/chrischute/flowplusplus/blob/master/models/flowplusplus/log_dist.py#L78

Original:

class Sigmoid(Flow):
    def forward(self, x, **kwargs):
        y = tf.sigmoid(x)
        logd = -tf.nn.softplus(x) - tf.nn.softplus(-x)