-
Notifications
You must be signed in to change notification settings - Fork 110
Open
Description
I see you have qconv2d_batchnorm
layer which folds the weights of the two layers and then quantizes.
We're bringing support for that to hls4ml, and it should help us save some resources & latency.
I'm wondering, do you plan to add the equivalent combined QDense
+ BatchNormalization
layer to QKeras
?
Metadata
Metadata
Assignees
Labels
No labels