Question concerning usage of Sequential model.

[errhernandez]

Hello! I am using Sequential to construct my model by concatenating different types of layers. If I wanted to freeze parameters associated to different layers, or apply different learning rates to different layers, how can this be done?

In pytorch one can use an ordered dict to name different layers in Sequential, but this appears not to be possible in PyG's version of Sequential.

Thanks!

[errhernandez]

Ok, I think I now know; it is simple enough:

for layer in model.children():
     for name, parameter in layer.named_parameters():
          'do whatever with them here, for example:'
           print( name, parametr.shape )

Sorry for asking trivialities; I'm still learning ;-) I hope this is useful to someone.

[rusty1s]

You are right that PyTorch supports an OrderedDict as input, which is something we could potentially add as well. However, your current workaround works as well :)