Open adrhill opened 1 year ago
Hello, i am also encountering this issue. How do have a work around it? I saw that you can try to change to Dense layer. Is there a code that i could reference from?
The workaround using a Dense layer is described here: https://github.com/albermax/innvestigate/issues/283#issuecomment-1276112045
On iNNvestigate
v2.0.1
, creating an analyzer inheriting fromAnalyzerNetworkBase
errors when the model contains aBatchNormalization
layer, e.g.:This might be due to batch normalisation layers keeping moving averages of the mean and standard deviation of the training data, causing problems with the Keras history when reversing the computational graph in iNNvestigate's
create_analyzer_model
.Minimal example reproducing the issue
Full stacktrace