Andy Brock

Hi guys, I'm using the dev versions of both libraries (so Theano 0.9.0dev2 and lasagne 0.2dev1), and I've got the standard theano flags set to floatX=float32 and time_once for all...

I'll email the celebA authors and ask their permission to temporarily host it through git-LFS. Thanks for letting me know!

The authors have put up a Google Drive link on the celebA project page: https://drive.google.com/drive/folders/0B7EVK8r0v71pWEZsZE9oNnFzTm8 Let me know if anything else like this happens, I'm checking the issues fairly regularly.

Git-LFS is hitting me with a bandwidth limit, here's a drive mirror for the simple model: https://drive.google.com/open?id=0B3_iVBZsC4GGck5WWWc0R0dvT1U

Thanks, this is definitely a discrepancy, likely born of the fact that I threw this layer together in a couple minutes right after seeing the paper. Also because cuDNN's batchnorm...

^ I'll see about putting that together sometime this month

Excellent question; "self.layer_index" is actually the number of layers in the network overall. This is only called "layer_index" and not "num_layers" or "L" because I quickly adopted the init() code...