Why not using the class member (weight, bias)?

[whamza15]

gluon-nlp/gluonnlp/model/sampled_block.py

Line 169 in fb27033

def hybrid_forward(self, F, x, sampled_values, label, weight, bias):

Why pass them from outside?

[szha]

The convention for Gluon is that the class members for parameters will be automatically passed in hybrid_forward as keyword arguments. The name of the keyword argument is from the self.params.get(name).

[whamza15]

Hmm, didn't know that. Not clear from the docs. Thanks for answering

[szha]

Thanks for the feedback. I created an issue in mxnet repo to track this apache/mxnet#12580