Skip to main content
Version: v1.0.x

Real time scoring

This example demonstrates how to use the eScorer Python client for real time scoring. You can download the complete example here.

Prerequisites

Install the H2O eScorer Python client.

Real time scoring with the Python client​

After authenticating the client, you can use the realtime_scorer method to perform real time scoring. The realtime_scorer method takes the model name and dataset file path as required input parameters and returns the result of real time scoring.

result = await client.scorer.realtime(
    model_name="<model_name>",
    dataset_filepath="/path/to/dataset.csv",
    skip_header=True,
    num_rows=1000,
    progress=True
)

Get the predictions​

result.predictions
              bad_loan.0           bad_loan.1 
0     0.08621099591255188   0.9137890040874481
1     0.08830222487449646   0.9116977751255035
2     0.03488980233669281   0.9651101976633072
3      0.2397906482219696   0.7602093517780304
4      0.4451645463705063   0.5548354536294937
..                    ...                  ...
995   0.15911102294921875   0.8408889770507812
996   0.03693510591983795    0.963064894080162
997   0.25428231060504913   0.7457176893949509
998    0.1850033551454544   0.8149966448545456
999   0.03011934459209442   0.9698806554079056

[1000 rows x 2 columns]

Get the number of rows scored​

result.rows_produced
1000

Get the scoring latency​

f"{result.time_elapsed}s"
135.317s

Get the number of errors​

result.model_error_count returns the number of errors that occurred during model scoring, such as missing columns or incorrect data types.

result.other_error_count returns the number of errors that occurred during the scoring process but are not related to the model, such as network issues.

Feedback