Describe the workflow you want to enable

In lines like:

model.predict_proba(df)

I know that certain rows do not need the probability to be predicted. So I would need to:

• Filter the dataframe
• Store the indexes at which I do not want evaluation
• Evaluate the filtered dataframe
• Put back the whole dataframe with the probabilities of the dropped data as -1, NaN or some other reasonable value.

Describe your proposed solution

I would like to have a skip_at argument like:

indexes=numpy.array[1, 20, 40])

probabilities = model.predict_proba(df, skip_at=indexes)

Such that probabilities is NaN at 1, 20 and 40 do not get added and specially the model does not waste time evaluating the probability there.

Describe alternatives you've considered, if relevant

No response

Additional context

No response