Average model result

[wagpa]

Is the prediction accuracy of the evaluator consistent? Do we see reasons why it wouldn't be (for example a badly generated graph)

[wagpa]

With tensorboard it seems like the evaluation is consistent

Later could create tabular data with different embedders.