3.2 Part 2: Deploy with configuration step

3.2.1 Intoduction

The previous part showed the main concepts of the platform and how to use the basics of the SDK. With what you already know you are able to put into production really simple applications. But in order to build more realistic applications, using more complex code with dependencies such as Python libraries, it is needed to learn more advanced functionnalities and especially how to configure the execution context of a step.

This page will present the same commands as the previous ones going through more available functionalities offered by the platform, with a real Machine Learning use case. We will improve this Machine Learning application later in Part 3 and Part 4.

You can find all the code used in this part and its structure here

Prerequisites

• Python 3.8 or higher is required to be installed on your computer.

• Have done the previous Part of this tutorial (Part 1: Deploy a simple pipeline).

3.2.2 First ML application

Here we will build a pipeline to train a simple ML model on the iris dataset.

Dataset and use case introduction

The iris dataset describes four features (petal length, petal width, sepal length, sepal width) from three different types of irises (Setosa, Veriscolour, Virginica).

The goal of our application is to classify flower type based on the previous four features. In this part we will start our new application by building a first pipeline to train a simple ML model on the iris dataset and deploy it.

We will use the following code that trains a sklearn KNN classifier on this dataset and make a prediction on the test set.

from sklearn import datasets
from sklearn.neighbors import KNeighborsClassifier
import numpy as np


def TrainPredictIris():

   iris_X, iris_y = datasets.load_iris(return_X_y=True, as_frame=True)

   np.random.seed(0)
   indices = np.random.permutation(len(iris_X))

   iris_X_train = iris_X.loc[indices[0:90], :]
   iris_y_train = iris_y.loc[indices[0:90]]
   iris_X_test = iris_X.loc[indices[90:], :]
   iris_y_test = iris_y.loc[indices[90:]]

   knn = KNeighborsClassifier()
   knn.fit(iris_X_train, iris_y_train)

   result = knn.predict(iris_X_test)

   print(result)

3.2.3 Cleaning object

Before we really start building our new application, we might want to clean the platform from the objects we created in Part 1. To do this, we need to use the functions associated with each object, here the endpoint, the pipeline and the step.

Warning

These objects have dependencies on each other, we have to delete them in a certain order. First the endpoint, then the pipeline and then the step.

sdk.delete_deployment(deployment_name="part-1-hello-world-endpoint")

sdk.delete_pipeline(pipeline_name="part-1-hello-world-pipeline")

sdk.delete_step(step_name="part-1-hello-world-step")

In the rest of this part we will follow the same workflow as in the previous one:

3.2.4 In depth step configuration

Now it is time to use the create_step() method of craft-ai-sdk object to create step like before. This time though, we will see how we can define a bit more the step and its execution context. We are going to focus on two parameters.

3.2.4.1 Python libraries

As you might have noticed, the code above uses external Python libraries (numpy and scikit learn). In the previous step we conveniently built an application that didn’t require any external dependency but this time if we want this code to work in the platform we have to inform it that this step requires some Python libraries to run properly.

To do so we create a requirements.txt file at the root of the repo, containing the list of Python libraries used in our step function:

scikit-learn==1.2.1
numpy==1.19.5
pandas==1.4.2

In this case we placed it at the root of the repo but you can put it whereever you want.

As for the code, the platform only sees what’s on your github repository so don’t forget to push your requirement file on github.

You can now specify the path of this file in the Libraries & Packages section of your project settings using the web interface and all the steps created in this project will have these libraries installed and ready to be used.

3.2.4.2 Included folder

When creating your step, you may not need to include all the files of your project repositories in your step. You can specify the files and folder(s) to include from the GitHub repository to prevent the step from accessing all code available in the repository by default, using the included_folders option.

Once again you can set this option in your project settings page in the web interface.

🎉 Now all the steps created in this project will have the relevant libraries installed and only the necessary file will be included

3.2.5 Configuration Hierarchy

When you created your project in the platform (cf Part 0),, you have set up different parameters (like the repository URL, the deploy key or the Python version you are using) and we also set up new parameters in the previous section.

By default, your step will apply those parameters during its creation (that is why you didn’t need to add any parameters when creating your step in Part 1). However, sometimes you want to define them only at the step level and keep the previous ones at the project level. This is the role of the create_step() function’s parameter container_config . You can pass as a dictionary the set of parameters you want to use for the step creation. It allows you to be really specific in the execution context of your step.

💡 For example, if you need to build a step embedding a function from another repository, you can specify the new Github repository URL and deploy key at the step level using the container_config parameter. Your project parameters will remain unchanged.

Warning

The execution context of a step is non persistent. It means that as soon as the execution is done, everything written in memory and on disk during the execution will be deleted. Therefore, a step is not the right place to store data or any information. We will learn more about persistence in Part 4

Here we will specify the requirements_path and the included_folders.

Note: You can specify them in the user interface at the project level as well. Just remember that they will be used by default if you don’t specify others at the step level.

3.2.5.1 included_folders

3.2.5.2 requirements_path

In order for our requirements.txt file to be taken into account, we must also add it to the container_config parameter with the requirements_path key.

sdk.create_step(
    step_name="part-2-irisclassifier-step",
  function_path="src/part-2-irisModel.py",
    function_name="TrainPredictIris",
    description="This function creates a classifier model for iris and makes prediction on test data set",
    container_config = {
        "requirements_path" : "requirements.txt",
        "included_folders" : ["src"]
    }
)

It may also be useful to describe precisely the steps created to be able to understand their purpose afterward. To do so, you can fill in the description parameter during the step creation.

🎉 Now your step has been created. You can now create your Pipeline.

From here, we reproduce the same steps as before with the creation of the pipeline, the endpoint and the execution of the objects we have created.

3.2.6 Create a pipeline

Create a pipeline with the create_pipeline() method of the SDK.

sdk.create_pipeline(
    pipeline_name="part-2-irisclassifier-pipeline",
        step_name="part-2-irisclassifier-step"
)

3.2.7 Deploy your Pipeline through an Endpoint

Now you can deploy your pipeline with an endpoint and execute it with an HTTP call as in Part 1.

Note : Here we get the execution ID directly from the endpoint return call instead of using sdk.list_pipeline_executions().

endpoint = sdk.create_deployment(
    pipeline_name="part-2-irisclassifier-pipeline",
    deployment_name_name="part-2-irisclassifier-endpoint",
    execution_rule="endpoint"
)

import requests

endpoint_URL = sdk.base_environment_url + "/endpoints/" + endpoint["name"]
headers = {"Authorization": "EndpointToken " + endpoint["endpoint_token"]}
request = requests.post(endpoint_URL, headers=headers)

request.status_code

pipeline_executions = sdk.list_pipeline_executions(pipeline_name="part-2-irisclassifier-pipeline")

pipeline_executions

logs = sdk.get_pipeline_execution_logs(pipeline_name="part-2-irisclassifier-pipeline",
    execution_id=pipeline_executions[-1]['execution_id'])
print('\n'.join(log["message"] for log in logs))

The output is a list of iris categories :

>> [2 0 2 0 2 2 0 0 2 0 0 0 1 2 2 0 0 0 1 1 0 0 1 0 2 1 2 1 0 2 0 2 0 0 2 0 2
1 1 1 2 2 2 1 0 1 2 2 0 1 1 2 1 0 0 0 2 1 2 0]

🎉 You can now execute a pipeline with an endpoint. Your Machine Learning application is deployed and ready to use!

Now that we can have more complex code in our steps and we know how to parametrize the execution context of our steps, we would like to be able to give it input elements to vary the result and receive the result easily. For this, we can use the input/output feature offered by the platform.

Next step: Part 3: Deploy with input and output