Jan Schlüter
Jan Schlüter
(Note, just so it's not overlooked: That PR is from your `master` branch, you may either want to file a new PR from a separate branch or rebase it onto...
> Would there be interest in adding these recipes? I guess so, yes! In what form would these recipes be? Ideally, I guess they'd construct the network in Python code,...
> A separate subdirectory for pre-trained models is not a bad idea. I'm still sad we didn't go for [food-related directory names](https://github.com/Lasagne/Lasagne/issues/215#issuecomment-104888170): `starters` (tutorials), `mains` (papers), `sides` (snippets), and now...
Thanks! I'm not so sure about the way the implementation uses the layer names... I'd start from the residual networks example and either modify the ElemwiseSumLayer to include a `theano.ifelse.ifelse()`...
Actually, the Stochastic Depth code is very clean and easy to read, whether you know Torch or not. I'd suggest to start from this, rather than from the paper, and...
> I'll have a look and try simplify the code a bit. Thanks! Don't have time to review this week... anybody else? > I didn't realize the same layers were...
> I have made some big changes (and more commits than necessary =/)! Don't worry, this can be cleaned up in retrospect if needed. Sounds you've made some good progress,...
@tencia, do you still know what you did there?
Looks good in general! Some comments (the diff is too long for github to display, so I'll comment here): > y_train=T.cast(theano.shared(y_train), 'int32') Why don't you cast the data _before_ turning...
> This is because supposedly you can only store floats on the GPU (the labels would be stored in host memory otherwise). Are you saying this is not the case?...