Danijar Hafner

icon indicating a website link https://danijar.com icon indicating an email [email protected]

icon company DeepMind icon location Berkeley icon location Building AI that makes autonomous decisions using world models, artificial curiosity, and temporal abstraction

Results 87 issues of Danijar Hafner

Use urllib.request.urlretrieve() in download utility?

refactoring

Cache embeddings in dataset and get rid of the disk_cache decorator

Rename Embedding to Encode?

Disk cache for generators

Stream to disk and yield file handle when cache file already exists.

Test argument invalidation of disk_cache decorator

test

Support pre-defined folds in Split

Allow specifying a column and keep rows of the same value together in one split.

feature

Add option for OneHot to encode falsy values by all zeroes

enhancement

@disk_cache does not always detect changed arguments

bug

Speed up loading Glove embeddings

1 comment

enhancement

Constant weight scaling?

You are scaling the weights based on the number of incoming activations: https://github.com/preritj/progressive_growing_of_GANs/blob/04351d69b2e8e9616715f3ac8cb155488953777d/net.py#L42 I think the authors scale adaptively, i.e. change the scale as the learned weights change: https://github.com/tkarras/progressive_growing_of_gans/blob/3e608722f9588ea135c87e972f2909591e22c3ec/network.py#L281-L284