Where can I find the token and model_id for the Aporia SDK?
Go to the "versions" page, which can be found in the model sidebar. Once there, clicking any of the integrations will show you a code snippet that contains your token and model_id.
Does the Aporia SDK affect the performance of my application?
For the most part, no - the SDK works in a separate thread in the background, and shouldn't have a noticeable effect on the performance of your application.
There are a few exceptions to this rule:
create_model_versionfunction is blocking.
- Functions that log training/test data are blocking, and might involve a noticeable computational overhead, depending on the size of the dataset your are logging.
Is it possible to report historical predictions?
Yes - you can report historical predictions (i.e. predictions that happened in the past) using the
occurred_at parameter in the
Can I modify a prediction after I report it?
No - The fields that are reported by
log_prediction (namely features and predictions) cannot be changed after they are reported.
This also applies to raw_inputs or actuals that have been reported with the same prediction ID.
Can I add raw_inputs or actuals to a prediction after I report it?
Yes - You can report raw_inputs and actuals for a prediction after you report the features and prediction values by using the same prediction ID.
I created a new monitor - when will it run?
Monitor scheduling is based on the inspection period you defined when creating the monitor.
As a rule of thumb, monitor scheduling is calculated as ~
inspection_period / 12.
For example, if you configure a monitor to inspect the last hour, it will run roughly every ~5 minutes.
How do I invite a new user to my organization?
Go to the team page in the sidebar.
How much does Aporia cost?
See our pricing.