Categorical context variables

[michaelpollmann]

Is there an easy way to use/implement categorical context variables?
Is my understanding correct that all context_vars will implicitly be treated as continuous, such that I should (?) turn them into dummy variables manually beforehand if they are categorical (and take on more than 2 values)? Or is there a reason to prefer treating categorical context variables as continuous?

[Jonas-Metzger]

In general, your generator should be big enough for it not to make much of a difference, but yes, it's better to include categorical conditioning variables as dummies rather than integers (except if they're ordered categorical, then integers are usually better). The inductive bias of a neural net is towards assuming continuity in it's input dimensions, so it's suboptimal to ask it to learn very discontinuous functions – which you can avoid by using dummies.