Skip to content

Inference View

Experimental

The Workbench View classes are currently in experimental mode so have fun but expect issues and API changes going forward.

InferenceView Class: A View that does endpoint inference and computes residuals

InferenceView

InferenceView Class: A View that does endpoint inference and computes residuals

Common Usage 
# Grab a Model
model = Model("abalone-regression")

# Create an InferenceView
inf_view = InferenceView.create(model)
my_df = inf_view.pull_dataframe(limit=5)

# Query the view
df = inf_view.query(f"SELECT * FROM {inf_view.table} where residuals > 0.5")
Source code in src/workbench/core/views/inference_view.py 
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
class InferenceView:
    """InferenceView Class: A View that does endpoint inference and computes residuals

    Common Usage:
        ```python
        # Grab a Model
        model = Model("abalone-regression")

        # Create an InferenceView
        inf_view = InferenceView.create(model)
        my_df = inf_view.pull_dataframe(limit=5)

        # Query the view
        df = inf_view.query(f"SELECT * FROM {inf_view.table} where residuals > 0.5")
        ```
    """

    @classmethod
    def create(
        cls,
        model: Model,
    ) -> Union[View, None]:
        """Create a View that does endpoint inference and computes residuals

        Args:
            model (Model): The Model object to use for the target and features

        Returns:
            Union[View, None]: The created View object (or None if failed)
        """
        # Log view creation
        log.important("Creating Inference View...")

        # Pull in data from the FeatureSet
        fs = FeatureSet(model.get_input())
        df = fs.pull_dataframe()

        # Grab the target from the model
        target = model.target()

        # Run inference on the data
        end = Endpoint(model.endpoints()[0])
        df = end.inference(df)

        # Determine if the target is a classification or regression target
        if model.model_type == ModelType.REGRESSOR:
            df["residuals"] = df[target] - df["prediction"]
            df["residuals_abs"] = df["residuals"].abs()
        elif model.model_type == ModelType.CLASSIFIER:
            class_labels = model.class_labels()
            class_index = {label: i for i, label in enumerate(class_labels)}
            df["residuals"] = df["prediction"].map(class_index) - df[target].map(class_index)
            df["residuals_abs"] = df["residuals"].abs()
        else:
            log.warning(f"Model type {model.model_type} has undefined residuals computation")
            df["residuals"] = 0
            df["residuals_abs"] = 0

        # Save the inference results to an inference view
        view_name = f"inf_{model.uuid.replace('-', '_')}"
        return PandasToView.create(view_name, fs, df=df, id_column=fs.id_column)

create(model) classmethod

Create a View that does endpoint inference and computes residuals

Parameters:

Name Type Description Default
model Model

The Model object to use for the target and features

 required

Returns:

Type Description
Union[View, None]

Union[View, None]: The created View object (or None if failed)
Source code in src/workbench/core/views/inference_view.py 
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
@classmethod
def create(
    cls,
    model: Model,
) -> Union[View, None]:
    """Create a View that does endpoint inference and computes residuals

    Args:
        model (Model): The Model object to use for the target and features

    Returns:
        Union[View, None]: The created View object (or None if failed)
    """
    # Log view creation
    log.important("Creating Inference View...")

    # Pull in data from the FeatureSet
    fs = FeatureSet(model.get_input())
    df = fs.pull_dataframe()

    # Grab the target from the model
    target = model.target()

    # Run inference on the data
    end = Endpoint(model.endpoints()[0])
    df = end.inference(df)

    # Determine if the target is a classification or regression target
    if model.model_type == ModelType.REGRESSOR:
        df["residuals"] = df[target] - df["prediction"]
        df["residuals_abs"] = df["residuals"].abs()
    elif model.model_type == ModelType.CLASSIFIER:
        class_labels = model.class_labels()
        class_index = {label: i for i, label in enumerate(class_labels)}
        df["residuals"] = df["prediction"].map(class_index) - df[target].map(class_index)
        df["residuals_abs"] = df["residuals"].abs()
    else:
        log.warning(f"Model type {model.model_type} has undefined residuals computation")
        df["residuals"] = 0
        df["residuals_abs"] = 0

    # Save the inference results to an inference view
    view_name = f"inf_{model.uuid.replace('-', '_')}"
    return PandasToView.create(view_name, fs, df=df, id_column=fs.id_column)

Questions?

The SuperCowPowers team is happy to answer any questions you may have about AWS and Workbench. Please contact us at workbench@supercowpowers.com or on chat us up on Discord