Open job-almekinders opened 1 week ago
Oh this is not ideal. Can you post your client usage? Just to make sure I understand it.
I don't think it should be included unless explicitly requested though, so probably the get_historical_features
is the one that's wrong.
@franciscojavierarceo Yes, I've just posted a minimal example in the description!
As you can see, the offline transformation with pandas returns the additional columns: 'REQUEST_FLOAT', 'event_timestamp'
Context
I'm using an
ODFV
with aPandasTransformation
and aRequestSource
input. Furthermore, I'm using Postgres offline and online store.Expected Behavior
I would expect both the
get_online_features_async
method and theget_historical_features
method to return the same set of features. However, when retrieving features withget_online_features_async
, the input data from theRequestSource
is not present in the output response. On the other hand, when retrieving features withget_historical_features
, the input data from theRequestSource
is present in the output response.I'm not sure which behavior is to be expected. However, I think think that both approaches should return the same columns.
Current Behavior
When calling the
get_online_features_async
method I do not see the input from theRequestSource
back in the output response, while I would expect it to be there.One could also argue that it should not be in the output response. That would also be an option. However, I would assume that the online and offline feature retrieval would return the same output in terms of columns.
Steps to reproduce
docker-compose.yml
feature_store.yml
Insert into offline store (postgres) postgres_init/create-offline-store-database.sql
bootstrap.py
materialize.py
inference.py
Specifications
Version: 0.36.0 Platform: macOS - M1 Subsystem: Sonoma 14.1.1
Possible Solution