3611452afac53b53f3e41ee83d7fc7bd811ffb81,thinc/neural/_classes/batchnorm.py,BatchNorm,begin_update,#BatchNorm#Any#Any#,43

Before Change

        // Should we consider a sample be the instance, or the batch?
        diff = X - self.m
        incr = (1-alpha) * diff
        self.m += incr.mean(axis=0)
        self.v += (diff * incr).mean(axis=0)
        self.v *= alpha

        Xhat = _forward(self.ops, X, mu, var)

After Change

    def begin_update(self, X, drop=0.):
        if drop is None:
            return self.predict(X), None
        assert X.dtype == "float32"
        X, backprop_child = self.child.begin_update(X, drop=0.)
        N, mu, var = _get_moments(self.ops, X)

Instances