Closed alexpiehler closed 4 years ago
The implementation of the catvae is still missing the annealing of the temperature parameter, which is used in the gumbel softmax function. The problem is that we need to be able to access a attribute batch_idx from somewhere. The code is currently commented out. https://github.com/MoritzWag/Representation-Learning/blob/e1bf9a75934527ab51b1d8af923ad40d0da5af4a/library/library/models2/cat_vae.py#L159-L169
batch_idx
The implementation of the catvae is still missing the annealing of the temperature parameter, which is used in the gumbel softmax function. The problem is that we need to be able to access a attribute
batch_idx
from somewhere. The code is currently commented out. https://github.com/MoritzWag/Representation-Learning/blob/e1bf9a75934527ab51b1d8af923ad40d0da5af4a/library/library/models2/cat_vae.py#L159-L169