# Explainability#

Arthur can automatically calculate explanations (feature importances) for every prediction your model makes. In order to make this possible we package up your model in a way that allows us to call its predict function, which allows us to calculate explanations. We require a few things from your end:

• A python script that wraps your models predict function

• For Image models, a second function, load_image is also required (see CV Explainability).

• A directory containing the above file, along with any serialized model files, and other supporting code

• A requirements.txt with the dependencies to support the above

This guide will walk through setting everything up, and then using the SDK to enable explainability.

## Setting up Project Directory#

### Project Structure#

Here is an example of what your project directory might look like.

|-- model_folder/
|   |-- data/
|   |  |-- training_data.csv
|   |  |-- testing_data.csv
|   |-- requirements.txt
|   |-- model_entrypoint.py
|   |-- utils.py
|   |-- serialized_model.pkl


### Requirements File#

Your project requirements and dependencies can be stored in any format you like, such as the typical requirements.txt file, or another form of dependency management.

This should contain all packages your model and predict function need to run.

Note

You do not need to include the arthurai package in this requirements file, we supply that.

# example_requirements.txt

pandas==0.24.2
numpy==1.16.4
scikit-learn==0.21.3
torch==1.3.1
torchvision==0.4.2


It is advised to pin the specific versions your model requires. If no version is pinned we will use the latest version. This can cause issues if the latest version is not compatible with the version used to build your model.

### Prediction Function#

We need to be able to send new inferences to your model to get predictions and generate explanations. For us to have access to your model, you need to create an entrypoint file that defines a predict() method.

The exact name of the file isn’t strict, so long as you specify the correct name when you enable explainability (see below). The only thing that does matter is that this file implements a predict() method. In most cases, if you have a previously trained model, this predict() method will likely just invoke the prediction from your trained model.

# example_entrypoint.py

def predict(x):
return sk_model.predict_proba(x)


See the SparkML Integration Guide for an example using a SparkML model.

This predict method can be as simple or complicated as you need, so long as you can go from raw input data to a model output prediction. Specifically, in the case of a binary classifier, we expect a 2-d array where the first column indicates probability_0 for each input, and the second column indicates probability_1 for each input. In the case of a multiclass classifier with n possible labels, we expect an n-d array where column i corresponds to the predicted probability that each input belongs to class i.

### Preprocessing for Prediction#

Commonly, a fair amount of feature processing and transformation will need to happen prior to invoking your actual model.predict(). This might include normalizations, rescaling, one-hot encoding, embedding, and more. Whatever those transformations are, you can make them a part of this predict() method. Alternatively, you can wrap all those transformations into a helper function.

# example_entrypoint.py
from utils import pipeline_transformations

def predict(x):
return sk_model.predict_proba(pipeline_transformations(x))


## Enabling Explainability#

Enabling explainability can be done using the SDK function arthur_model.enable_explainability, which takes as input a sample of your model’s data (to train the explainer), and which takes as input the files that contain your model’s predict function and necessary environment.

arthur_model.enable_explainability(
project_directory="/path/to/model_folder/",
requirements_file="requirements.txt",
user_predict_function_import_path="model_entrypoint",
ignore_dirs=["folder_to_ignore"] # optionally exclude directories within the project folder from being bundled with predict function
)


The above provides a simple example. For a list of all configuration options and details around them, see the explainability entry in the Enrichment guide.

1. joblib is a Python library that will allow you to reconstruct your model from a serialized pickle file.

2. X_train is your trained model dataframe.

3. user_predict_function_import_path is the Python path to import the entrypoint file as if you imported it into the python program that is running enable_explainability.

### Configuration Requirements#

When going from disabled to enabled, you will need to include the required configuration settings. Once the explainability enrichment has been enabled, you can update the non-required configuration settings without re-supplying required fields. When disabling the explainability enrichment, you are not required to pass in any config settings.

Configuration

Setting

Required

Description

df

X

The dataframe passed to the explainer. Should be similar to, or a subset of, the training data. Typically small, ~50-100 rows.

project_directory

X

The path to the directory containing your predict function, requirements file, model file, and any other resources need to support the predict function.

user_predict_function_import_path

X

The name of the file containing the predict function. Do not include .py extension. Used to import the predict function.

requirements_file

X

The name of the file containing pip requirements for predict function.

python_version

X

The Python version to use when executing the predict function. This is automatically set to the current python version when using model.enable_explainability().

sdk_version

X

The arthurai version used to make the enable request. This is automatically set to the currently installed SDK version when using model.enable_explainability().

explanation_algo

The explanation algorithm to use. Valid options are 'lime' or 'shap'. Default value of 'lime'.

explanation_nsamples

The number perturbed samples used to generate the explanation. For a smaller number of samples, the result will be calculated more quickly but may be less robust. It is recommended to use at least 100 samples. Default value of 2000.

inference_consumer_score_percent

Number between 0.0 and 1.0 that sets the percent of inferences to compute an explanation score for. Only applicable when streaming_explainability_enabled is set to true. Default value of 1.0 (all inferences explained).

streaming_explainability_enabled

If true, every inference will have an explanation generated for it. If false, explanations are available on-demand only.

ignore_dirs

List of paths to directories within project_directory that will not be bundled and included with the predict function. Use to prevent including irrelevant code or files in larger directories.

## CV Explainability#

Note

Explainability is currently available as an enrichment for classification, multi-labeling, and regression CV models, but not object detection CV models.

In your model_entrypoint.py for Multiclass Image models, in addition to the predict() function, there is a second function which is required: load_image(). This function should take in a string, which is a path to an image file. The function should return the image in a numpy array. Any image processing, such as converting to grey scale, should also happen in this function. This is because Lime (the explanation algorithm used behind the scenes) will create variations of this array to generate explanations. However any transformation resulting in a non-numpy array, should happen in the predict function, such as converting to a Tensor.

No image resizing is required. As part of onboarding an image model, pixel_height and pixel_width are set as metadata on the model. When ingesting, Arthur will automatically resize the image to the configured size, and pass this resized image path to the load_image function.

Below is a full example file for an Image model, with both load_image and predict defined. Imports and class definitions are omitted for brevity.

# example_entrypoint.py
import ...

class MedNet(nn.Module):
...

# load model using custom user defined class
net = MedNet()
path = pathlib.Path(__file__).parent.absolute()

# helper function for transforming image
def quantize(np_array):
return np_array + (np.random.random(np_array.shape) / 256)

"""Takes in single image path, and returns single image in format predict expects
"""
return quantize(np.array(Image.open(image_path).convert('RGB')) / 256)

def predict(images_in):
"""Takes in numpy array of images, and returns predictions in numpy array.

Can handle both single image in numpy array, or multiple images.
"""
batch_size, pixdim1, pixdim2, channels = images_in.shape
raw_tensor = torch.from_numpy(images_in)
processed_images = torch.reshape(raw_tensor, (batch_size, channels, pixdim1, pixdim2)).float()
net.eval()
return net(processed_images).numpy()


## NLP Explainability#

Enabling explainability for NLP models follows the same process for Tabular models.

Note

An important choice for NLP explainability is the text_demiliter parameter, since this delimiter determines how tokens will be perturbed when generating explanations.

Here is an example entrypoint.py file which loads our NLP model and defines a predict function that the explainer will use:

model_path = os.path.join(os.path.dirname(__file__), "model.pkl")

def predict(fvs):
# our model expects a list of strings, no nesting
# if we receive nested lists, unnest them
if not isinstance(fvs[0], str):
fvs = [fv[0] for fv in fvs]
return model.predict_proba(fvs)


## Troubleshooting#

While this can be an issue with any model type, it is common to see when using sk-learn objects that take in custom user functions. We will use TfidfVectorizer as an example, which is a commonly used vectorizer for NLP models, that often utilizes custom user functions.

A TfidfVectorizer accepts a user defined tokenize function, which is used to split a text string into tokens.

Problem

Say this code was used to create your model.

# make_model.py

def tokenize(text):
# tokenize and lemmatize
doc = nlp(txt)
tokens = []
for token in doc:
if not token.is_stop and not token.is_punct \
and not token.is_space and token.lemma_ != '-PRON-':
tokens.append(token.lemma_)

def make_model():
# here we pass a custom function to an sklearn object
vectorizer = TfidfVectorizer(tokenizer=tokenize)
vectorizer.fit(X_train)
model = LogisticRegression()
model.fit(vectorizer.transform(X_train))

pipeline = make_pipeline(vectorizer, model)
joblib.dump(pipeline, 'model.pkl')

if __name__ == "__main__":
make_model()


Now you create this entrypoint file to enable explainability:

# entrypoint.py

def predict(fv):
return model.predict_proba(fv)


Now when the SDK imports entrypoint to test the function, the following error gets thrown:

AttributeError: module '__main__' has no attribute 'tokenize'

What happens is that Python failed to serialize the custom function, only the reference to how it was imported. Which in this case, it was just top level in the model creation script (hence __main__.tokenize in the error). This function doesn’t exist in entrypoint, and so the error is thrown.

Solution

To solve, you need to pull out tokenize into its own module, that can be imported from both create_model.py and also in entrypoint.py.

# model_utils.py

def tokenize(text):
# tokenize and lemmatize
doc = nlp(txt)
tokens = []
for token in doc:
if not token.is_stop and not token.is_punct \
and not token.is_space and token.lemma_ != '-PRON-':
tokens.append(token.lemma_)

# create_model.py
from model_utils import tokenize

def make_model():
# here we pass a custom function to an sklearn object
vectorizer = TfidfVectorizer(tokenizer=tokenize)
vectorizer.fit(X_train)
model = LogisticRegression()
model.fit(vectorizer.transform(X_train))

pipeline = make_pipeline(vectorizer, model)
joblib.dump(pipeline, 'model.pkl')

if __name__ == "__main__":
make_model()

# entrypoint.py
from model_utils import tokenize


Now, when Python serializes the model, it stores the reference as model_utils.tokenize, which is also imported within entrypoint.py and therefore no error is thrown.
Now everything will work, but both model_utils.py AND entrypoint.py must be included in the directory passed to enable_explainability().