You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I couldn't totally find where this happens but compile_pymc creates an "obs_id" coordinate automatically. If the user doesn't provide that coordinate it gets populated as None. This causes this line of code to throw an error
coords = {name: pd.Index(vals) for name, vals in model.coords.items()}
as pd.Index(vals) is expecting vals to be an array of some kind.
Should we auto-populate the obs_id coordinate as an array of the length of the number of observations?
The text was updated successfully, but these errors were encountered:
Yeah I realized after posting this that probably what was happening is I specified the "dims" for a distribution as "obs_id" but didn't provide "obs_id" as a coordinate. Pymc automagically handles this behavior. Do we want nutpie to do the same? Or I should just stop doing silly things when specifying the model (in which case we can close this issue)?
I couldn't totally find where this happens but compile_pymc creates an "obs_id" coordinate automatically. If the user doesn't provide that coordinate it gets populated as None. This causes this line of code to throw an error
coords = {name: pd.Index(vals) for name, vals in model.coords.items()}
as pd.Index(vals) is expecting vals to be an array of some kind.
Should we auto-populate the obs_id coordinate as an array of the length of the number of observations?
The text was updated successfully, but these errors were encountered: