Batch learning with unconditional inputs

[RomanKlus]

Hi,

I am trying to implement a hypernetwork that only has unconditional inputs and train the models in batches.
Nevertheless, the
Weights = hnet.forward(train_features)
returns the ret_format as 'sequential' and
hidden = mnet.forward(train_features, weights=Weights)
outputs the "AssertionError" message that "assert(len(weights) == len(self.param_shapes))".
I understand where the issue lies, but following your documentation, I am not able to convince "mnet.forward" to input weights in 'sequential' format.

Thanks for the help!
Roman