Anant Gupta

Loss goes to infinity on running with default setup

2

I just cloned the repo and ran the `main.py` with the parameters, as listed in README and got the following training progress: ``` [0/100][0/733] Loss_D: 1.38627135754 Loss_G: 1.38978719711 [0/100][1/733] Loss_D:...

This PR fixes all the issue with BahdanauAttnDecoderRNN class and makes the NMT trainable with it as well. The changes consist of the following: * Syntax fixes (wrong class in...

Reproducibility on Task 14

1

Hi Jim, I'm can't seem to get a 100% accuracy on task 14 as reported on README. Here are some stats:

Currently, the height and width of the #lightboxImage and #outerImageContainer are set based on the page height/width and image height/width. Sometimes its producing the following result: ![screen shot 2014-10-21 at...

Geolocation option

3

The option to geolocate the user seems to be missing right now. There can be google's crosshair icon on the top of the map for geolocation, and also an option...

Minor typo bug in models.lua

1

On line 28, it should be `opt.backend`.

Discriminator input

1

In the paper, it's mentioned that the input of discriminator is either joint pairs of q(x\~,z) or p(x,z\^). However, in the [code](https://github.com/IshmaelBelghazi/ALI/blob/master/ali/bricks.py#L49), it seems (z,z\^) and (x,x\~) are concatenated and...