IntelLabs / bayesian-torch

A library for Bayesian neural network layers and uncertainty estimation in Deep Learning extending the core of PyTorch

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

local variable 'kl' referenced before assignment when using flipout in dnn_to_bnn()

barghit opened this issue · comments

The problem seems to be in line 544 of conv_flipout.py, the line is self.kl = kl inside the forward() method, i get the error "local variable 'kl' referenced before assignment" as an UnboundLocalError.

Thank you in advance.

@barghit can you give more info on how you are defining the model with code snippet or log to understand your issue?

Hi! I attached the exact code snippet and the error I get. I don't get the same error with the "Reparameterization" option, it works perfectly. (I also thought it was a problem of the summary method but it does the same when i try to run the training).

for_github.txt

fixed with commit 1180b87