Where is the batch_normalization implementation specified in TVM?


I would like to see the implementation of batch_normalization used by Relay. However, I searched all the source code, but I can’t get any information about the implementation.

Is there an implementation in another location?

1 Like

I think it is not implemented per se. There is a BatchNormToInferUnpack function, part of the SimplifyInference pass defined later in that file. Similarly, dropout is deleted.

I do not think we currently have a training equivalent (which one would need to position in a way that it can be overridden for people with their own batch norm implementation), I’m using my own pass for the *Norm and to generate the dropout masks externally.

Best regards


1 Like

Thank you for answer. It is defined in pass, not direct implementation.

Thank you for good information.