Generate Responsible AI dashboard with YAML and Python (preview)

APPLIES TO: Azure CLI ml extension v2 (current) Python SDK azure-ai-ml v2 (preview)

The Responsible AI (RAI) dashboard can be generated via a pipeline job using RAI components. There are six core components for creating Responsible AI dashboards, along with a couple of helper components. A sample experiment graph:

Screenshot of a sample experiment graph.

Getting started

To use the Responsible AI components, you must first register them in your Azure Machine Learning workspace. This section documents the required steps.

Prerequisites

You'll need:

  • An AzureML workspace
  • A git installation
  • A MiniConda installation
  • An Azure CLI installation

Installation Steps

  1. Clone the Repository

    git clone https://github.com/Azure/RAI-vNext-Preview.git 
    
    cd RAI-vNext-Preview 
    
  2. Log into Azure

    Az login
    
  3. Run the setup script

    We provide a setup script which:

    • Creates a new conda environment with a name you specify
    • Installs all the required Python packages
    • Registers all the RAI components in your AzureML workspace
    • Registers some sample datasets in your AzureML workspace
    • Sets the defaults for the Azure CLI to point to your workspace

    We provide PowerShell and bash versions of the script. From the repository root, run:

    Quick-Setup.ps1 
    

    This will prompt for the desired conda environment name and AzureML workspace details. Alternatively, use the bash script:

    ./quick-setup.bash <CONDA-ENV-NAME> <SUBSCRIPTION-ID> <RESOURCE-GROUP-NAME> <WORKSPACE-NAME>
    

    This script will echo the supplied parameters, and pause briefly before continuing.

Responsible AI components

The core components for constructing a Responsible AI dashboard in AzureML are:

  • RAI Insights Dashboard Constructor
  • The tool components:
    • Add Explanation to RAI Insights Dashboard
    • Add Causal to RAI Insights Dashboard
    • Add Counterfactuals to RAI Insights Dashboard
    • Add Error Analysis to RAI Insights Dashboard
    • Gather RAI Insights Dashboard

The RAI Insights Dashboard Constructor and Gather RAI Insights Dashboard components are always required, plus at least one of the tool components. However, it isn't necessary to use all the tools in every Responsible AI dashboard.

Below are specifications of the Responsible AI components and examples of code snippets in YAML and Python. To view the full code, see sample YAML and Python notebook

Limitations

The current set of components have a number of limitations on their use:

  • All models must be in registered in AzureML in MLFlow format with a sklearn flavor.
  • The models must be loadable in the component environment.
  • The models must be pickleable.
  • The models must be supplied to the RAI components using the 'Fetch Registered Model' component which we provide.
  • The dataset inputs must be pandas DataFrames in Parquet format.
  • A model must still be supplied even if only a causal analysis of the data is performed. The DummyClassifier and DummyRegressor estimators from SciKit-Learn can be used for this purpose.

RAI Insights Dashboard Constructor

This component has three input ports:

  • The machine learning model
  • The training dataset
  • The test dataset

Use the train and test dataset that you used when training your model to generate model-debugging insights with components such as Error analysis and Model explanations. For components like Causal analysis that doesn't require a model, the train dataset will be used to train the causal model to generate the causal insights. The test dataset is used to populate your Responsible AI dashboard visualizations.

The easiest way to supply the model is using our Fetch Registered Model component, which will be discussed below.

Note

Currently only models with MLFlow format, with a sklearn flavor are supported.

The two datasets should be file datasets (of type uri_file) in Parquet format. Tabular datasets aren't supported, we provide a TabularDataset to Parquet file component to help with conversions. The training and test datasets provided don't have to be the same datasets used in training the model (although it's permissible for them to be the same). By default, the test dataset is restricted to 5000 rows for performance reasons of the visualization UI.

The constructor component also accepts the following parameters:

Parameter name Description Type
title Brief description of the dashboard String
task_type Specifies whether the model is for classification or regression String, classification or regression
target_column_name The name of the column in the input datasets, which the model is trying to predict String
maximum_rows_for_test_dataset The maximum number of rows allowed in the test dataset (for performance reasons) Integer (defaults to 5000)
categorical_column_names The columns in the datasets, which represent categorical data Optional list of strings (see note below)
classes The full list of class labels in the training dataset Optional list of strings (see note below)

Note

The lists should be supplied as a single JSON encoded string forcategorical_column_names and classes inputs.

The constructor component has a single output named rai_insights_dashboard. This is an empty dashboard, which the individual tool components will operate on, and then all the results will be assembled by the Gather RAI Insights Dashboard component at the end.

 create_rai_job: 

    type: command 
    component: azureml:rai_insights_constructor:1 
    inputs: 
      title: From YAML snippet 
      task_type: regression 
      model_info_path: ${{parent.jobs.fetch_model_job.outputs.model_info_output_path}} 
      train_dataset: ${{parent.inputs.my_training_data}} 
      test_dataset: ${{parent.inputs.my_test_data}} 
      target_column_name: ${{parent.inputs.target_column_name}} 
      categorical_column_names: '["location", "style", "job title", "OS", "Employer", "IDE", "Programming language"]' 

Exporting pre-built Cohorts for score card generation

Pre-built cohorts can be exported for use in score card generation. Find example of building cohorts in this Jupyter Notebook example: responsibleaidashboard-diabetes-decision-making.ipynb. Once a cohort is defined, it can be exported to json as follows:

# cohort1, cohort2 are cohorts defined in sample notebook of type raiwidgets.cohort.Cohort 
import json 
json.dumps([cohort1.to_json(), cohort2.to_json) 

A sample json string generated is shown below:

[ 
  { 
    "name": "High Yoe", 
    "cohort_filter_list": [ 
      { 
        "method": "greater", 
        "arg": [ 
          5 
        ], 
        "column": "YOE" 
      } 
    ] 
  }, 
  {
    "name": "Low Yoe", 
    "cohort_filter_list": [ 

      { 
        "method": "less", 
        "arg": [ 
          6.5 
        ], 
        "column": "YOE" 
      } 
    ] 
  } 
] 

Add Causal to RAI Insights Dashboard

This component performs a causal analysis on the supplied datasets. It has a single input port, which accepts the output of the RAI Insights Dashboard Constructor. It also accepts the following parameters:

Parameter name Description Type
treatment_features A list of feature names in the datasets, which are potentially ‘treatable’ to obtain different outcomes. List of strings (see note below)
heterogeneity_features A list of feature names in the datasets, which might affect how the ‘treatable’ features behave. By default all features will be considered Optional list of strings (see note below).
nuisance_model The model used to estimate the outcome of changing the treatment features. Optional string. Must be ‘linear’ or ‘AutoML’ defaulting to ‘linear.’
heterogeneity_model The model used to estimate the effect of the heterogeneity features on the outcome. Optional string. Must be ‘linear’ or ‘forest’ defaulting to ‘linear.’
alpha Confidence level of confidence intervals Optional floating point number. Defaults to 0.05.
upper_bound_on_cat_expansion Maximum expansion for categorical features. Optional integer. Defaults to 50.
treatment_cost The cost of the treatments. If 0, then all treatments will have zero cost. If a list is passed, then each element is applied to one of the treatment_features. Each element can be a scalar value to indicate a constant cost of applying that treatment or an array indicating the cost for each sample. If the treatment is a discrete treatment, then the array for that feature should be two dimensional with the first dimension representing samples and the second representing the difference in cost between the non-default values and the default value. Optional integer or list (see note below).
min_tree_leaf_samples Minimum number of samples per leaf in policy tree. Optional integer. Defaults to 2
max_tree_depth Maximum depth of the policy tree Optional integer. Defaults to 2
skip_cat_limit_checks By default, categorical features need to have several instances of each category in order for a model to be fit robustly. Setting this to True will skip these checks. Optional Boolean. Defaults to False.
categories What categories to use for the categorical columns. If auto, then the categories will be inferred for all categorical columns. Otherwise, this argument should have as many entries as there are categorical columns. Each entry should be either auto to infer the values for that column or the list of values for the column. If explicit values are provided, the first value is treated as the "control" value for that column against which other values are compared. Optional. auto or list (see note below.)
n_jobs Degree of parallelism to use. Optional integer. Defaults to 1.
verbose Whether to provide detailed output during the computation. Optional integer. Defaults to 1.
random_state Seed for the PRNG. Optional integer.

Note

For the list parameters: Several of the parameters accept lists of other types (strings, numbers, even other lists). To pass these into the component, they must first be JSON-encoded into a single string.

This component has a single output port, which can be connected to one of the insight_[n] input ports of the Gather RAI Insights Dashboard component.

  causal_01: 
    type: command 
    component: azureml:rai_insights_causal:1 
    inputs: 
      rai_insights_dashboard: ${{parent.jobs.create_rai_job.outputs.rai_insights_dashboard}} 
      treatment_features: `["Number of github repos contributed to", "YOE"]' 

Add Counterfactuals to RAI Insights Dashboard

This component generates counterfactual points for the supplied test dataset. It has a single input port, which accepts the output of the RAI Insights Dashboard Constructor. It also accepts the following parameters:

Parameter Name Description Type
total_CFs How many counterfactual points to generate for each row in the test dataset Optional integer. Defaults to 10
method The dice-ml explainer to use Optional string. Either random, genetic or kdtree. Defaults to random
desired_class Index identifying the desired counterfactual class. For binary classification, this should be set to opposite Optional string or integer. Defaults to 0
desired_range For regression problems, identify the desired range of outcomes Optional list of two numbers (see note below).
permitted_range Dictionary with feature names as keys and permitted range in list as values. Defaults to the range inferred from training data. Optional string or list (see note below).
features_to_vary Either a string "all" or a list of feature names to vary. Optional string or list (see note below)
feature_importance Flag to enable computation of feature importances using dice-ml Optional Boolean. Defaults to True

Note

For the non-scalar parameters: Parameters which are lists or dictionaries should be passed as single JSON-encoded strings.

This component has a single output port, which can be connected to one of the insight_[n] input ports of the Gather RAI Insights Dashboard component.

 counterfactual_01: 
    type: command 
    component: azureml:rai_insights_counterfactual:1 
    inputs: 
      rai_insights_dashboard: ${{parent.jobs.create_rai_job.outputs.rai_insights_dashboard}} 
      total_CFs: 10 
      desired_range: "[5, 10]" 

Add Error Analysis to RAI Insights Dashboard

This component generates an error analysis for the model. It has a single input port, which accepts the output of the RAI Insights Dashboard Constructor. It also accepts the following parameters:

Parameter Name Description Type
max_depth The maximum depth of the error analysis tree Optional integer. Defaults to 3
num_leaves The maximum number of leaves in the error tree Optional integer. Defaults to 31
min_child_samples The minimum number of datapoints required to produce a leaf Optional integer. Defaults to 20
filter_features A list of one or two features to use for the matrix filter Optional list of two feature names (see note below).

Note

filter_features: This list of one or two feature names should be passed as a single JSON-encoded string.

This component has a single output port, which can be connected to one of the insight_[n] input ports of the Gather RAI Insights Dashboard component.

  error_analysis_01: 
    type: command 
    component: azureml:rai_insights_erroranalysis:1 
    inputs: 
      rai_insights_dashboard: ${{parent.jobs.create_rai_job.outputs.rai_insights_dashboard}} 
      filter_features: `["style", "Employer"]' 

Add Explanation to RAI Insights Dashboard

This component generates an explanation for the model. It has a single input port, which accepts the output of the RAI Insights Dashboard Constructor. It accepts a single, optional comment string as a parameter.

This component has a single output port, which can be connected to one of the insight_[n] input ports of the Gather RAI Insights Dashboard component.

  explain_01: 
    type: command 
    component: azureml:rai_insights_explanation:VERSION_REPLACEMENT_STRING 
    inputs: 
      comment: My comment 
      rai_insights_dashboard: ${{parent.jobs.create_rai_job.outputs.rai_insights_dashboard}} 

Gather RAI Insights Dashboard

This component assembles the generated insights into a single Responsible AI dashboard. It has five input ports:

  • The constructor port that must be connected to the RAI Insights Dashboard Constructor component.
  • Four insight_[n] ports that can be connected to the output of the tool components. At least one of these ports must be connected.

There are two output ports. The dashboard port contains the completed RAIInsights object, while the ux_json contains the data required to display a minimal dashboard.

  gather_01: 
    type: command 
    component: azureml:rai_insights_gather:1 
    inputs: 
      constructor: ${{parent.jobs.create_rai_job.outputs.rai_insights_dashboard}} 
      insight_1: ${{parent.jobs.causal_01.outputs.causal}} 
      insight_2: ${{parent.jobs.counterfactual_01.outputs.counterfactual}} 
      insight_3: ${{parent.jobs.error_analysis_01.outputs.error_analysis}} 
      insight_4: ${{parent.jobs.explain_01.outputs.explanation}} 

Helper components

We provide two helper components to aid in connecting the Responsible AI components to your existing assets.

Fetch registered model

This component produces information about a registered model, which can be consumed by the model_info_path input port of the RAI Insights Dashboard Constructor component. It has a single input parameter – the AzureML ID (<NAME>:<VERSION>) of the desired model.

  fetch_model_job: 
    type: command 
    component: azureml:fetch_registered_model:1 
    inputs:
      model_id: my_model_name:12 

Tabular dataset to parquet file

This component converts the tabular dataset named in its sole input parameter into a Parquet file, which can be consumed by the train_dataset and test_dataset input ports of the RAI Insights Dashboard Constructor component. Its single input parameter is the name of the desired dataset.

  convert_train_job: 
    type: command 
    component: azureml:convert_tabular_to_parquet:1 
    inputs: 
      tabular_dataset_name: tabular_dataset_name 

Input constraints

What model formats and flavors are supported?

The model must be in MLFlow directory with a sklearn flavor available. Furthermore, the model needs to be loadable in the environment used by the Responsible AI components.

What data formats are supported?

The supplied datasets should be file datasets (uri_file type) in Parquet format. We provide the TabularDataset to Parquet File component to help convert the data into the required format.

Next steps