For more information on uploading a model artifact to Fiddler, see Uploading a Model Artifact.
Suppose you would like to upload a model artifact for a regression model.
Following is an example of what the
package.py script may look like.
import pickle from pathlib import Path import pandas as pd from sklearn.linear_model import LinearRegression PACKAGE_PATH = Path(__file__).parent OUTPUT_COLUMN = ['predicted_quality'] class MyModel: def __init__(self): # Load the model with open(PACKAGE_PATH / 'model.pkl', 'rb') as pkl_file: self.model = pickle.load(pkl_file) def predict(self, input_df): # Store predictions in a DataFrame return pd.DataFrame(self.model.predict(input_df), columns=OUTPUT_COLUMN) def get_model(): return MyModel()
Here, we are assuming that the model prediction column that has been specified in the
fdl.ModelInfo object is called
Updated 4 months ago