It is still not possible to specify the percent of forces to include in training with the python interface

[KMdraijer]

In featurize.py, passing a kwarg to write the forcepercent will result in a TypeError

for arg in kwargs:
            raise TypeError("Unexpected keyword argument '{}'".format(arg))

But to use aenet-PyTorch, one requires the following lines in the generate.in, as specified in the tutorial here:

forces
forcespercent 25

[alexurba]

The reason is that we currently don't support force training. Please use aenet-pytorch.

[KMdraijer]

Thanks for your response, my apologies for misunderstanding the purpose of the aenet-python tools.
However, the aenet-PyTorch github repository explicitly states that

ænet's regular generate.x is used to featurize the atomic structures in the reference data set;

so I expected that the aenet-python AenetAUCFeaturizer would work fine with aenet-PyTorch.
It is also easy to change the generate.x used by aenet-python by changing the config.json and setting the generate key to the modified generate.x shipped with aenet-PyTorch. The only thing that makes the AenetAUCFeaturizer.run_aenet_generate not run with the modified generate.x is the absence of the forcespercent keyword for aenet-PyTorch. If I'm not mistaken, this is easy to implement and facilitate.