Fix for BatchNormalization layers with `center=False` or `scale=False`

fastmachinelearning / hls4ml

Machine learning on FPGAs using HLS

https://fastmachinelearning.org/hls4ml

Apache License 2.0

1.24k stars 402 forks source link

Closed jmduarte closed 1 year ago

jmduarte commented 1 year ago

Fix for batch normalization layers with center=False or scale=False (i.e. beta or gamma are None)

jmitrevs commented 1 year ago

What milestone is this for?

vloncar commented 1 year ago

The current pytorch parser is not developed anymore, we should follow up the changes in the replacement parser in #723.

jicampos commented 1 year ago

@jmduarte I agree with @vloncar for the PyTorch parser. Would you like to add the fix for Keras now and follow up after #723?

JanFSchulte commented 1 year ago

I have updated #723 to support parsing of the affine argument so it will work with these changes.