MADE generative pass, tanh layer for scaling params, moons example

[ekrim]

Feel free to incorporate any or none of these fixes/additions. Let me know if you'd like me to break up the PR. Thanks!

1. Implement the generative pass (mode == 'inverse') for MADE module. This iteratively builds the output of the generative model given the input (e.g. the Gaussian latent data). The shift and scale parameters for the first element are independent of the data, and are determined simply by the bias weights.

2. Add the HalfTanh module, which applies tanh to the second half of the dimensions of a given tensor. This is used in MADE, since tanh adds stability to the scale parameters.

3. Add a 2D moon dataset example, including a visualization of generated samples. The command

python main.py --dataset MOONS --flow maf --epochs 50

results in the following plot:

4. Changing sign of scale parameter in 'inverse' mode to correspond to the MAF paper.

[ikostrikov]

Looks good to me!

I'm ready to merge.

One minor thing: there is a dependency between halftanh and the coupling layer. If one of the layers changes, it might break the other. I would prefer to have HalfTanh implemented using .chunk function so the connection with the coupling layer is easier to see in the code. Or even to move tanh to the implementation of MADE (with a flag).

[ekrim]

Ah, that does seem cleaner. I put a use_tanh flag for MADE, to directly apply tanh to the a.

[ikostrikov]

Great! Merging now.