You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
But since the model contains batchnorm layers, it should be training=True to avoid updating the parameters in the inference.
So why training=True here?
The text was updated successfully, but these errors were encountered:
Hi, I have a question about the code of Research mi_lira_2021.
It set training=True at inference:
privacy/research/mi_lira_2021/inference.py
Line 72 in 4dd8d0f
But since the model contains batchnorm layers, it should be training=True to avoid updating the parameters in the inference.
So why training=True here?
The text was updated successfully, but these errors were encountered: