Apologies for the double-submission, am not familiar with best practice yet but pull request #747 is a suggested fix for this
In calling GroupNorm from mapchildren (line 386) I believe "gn,G" should read "gn.G" to call G as an attribute rather than a (not defined) variable. I discovered this by using a Chain containing GroupNorm and trying to move it to the gpu, where it raised the error "ERROR: LoadError: UndefVarError: G not defined". This is also the case when trying to use GroupNorm in a neural_ode in DiffEqFlux
DhairyaLGandhi
commented
5 years ago
Apologies for the double-submission, am not familiar with best practice yet but pull request #747 is a suggested fix for this
In calling GroupNorm from mapchildren (line 386) I believe "gn,G" should read "gn.G" to call G as an attribute rather than a (not defined) variable. I discovered this by using a Chain containing GroupNorm and trying to move it to the gpu, where it raised the error "ERROR: LoadError: UndefVarError: G not defined". This is also the case when trying to use GroupNorm in a neural_ode in DiffEqFlux