ML workflow
1. ML workflow
SPEAKER: In the previous module, you learned about different options to develop an AI project on Google Cloud from a ready to use approach like pre-trained APIs to low or no-code solutions like AutoML, and to DIY solutions such as custom training. Then how do you build an ML model step by step? You'll find out in this module. You begin with an overview of the ML workflow. Then dive into each of the three workflow stages from data preparation to model development, and finally, model serving. Next, you investigate Machine Learning Operations, or MLOps, which takes ML models from development to production in the back end. You'll be shown an example of how to build a pipeline to automate the production using Vertex AI pipelines. A hands-on lab will then help you walk through the three stages to build an ML model with AutoML on Vertex AI. Gaining a solid grasp of ML terminology requires a clear understanding of how a neural network learns. This module offers an optional lesson that delves into the neural networks learning process, along with the key terminologies. If you're already familiar with the ML theories, feel free to skip this lesson. Let's get started with the ML workflow. Building an ML model is actually not too different from serving food in a restaurant. You start by preparing raw ingredients and finish by serving the dishes on the table. There are three main stages to the ML workflow with Vertex AI. The first stage is data preparation, which includes two steps, data uploading and feature engineering. A model needs a large amount of data to learn from. The quality and quantity of the data decide how much and how well the machine learns. The data used in machine learning can be real-time streaming data or batch data. The data can also be structured or unstructured. Structured data is data that can be easily stored in tables such as numbers and text. Unstructured data is data that cannot be easily stored in tables such as images and videos. The second stage of the ML workflow is model development. A model needs a tremendous amount of iterative training. This is when training and evaluation form a cycle to train a model, then evaluate the model, and then train the model some more. The third and final stage is model serving. A model needs to actually be used in order to predict results. This is when the machine learning model is deployed and monitored. If you don't move an ML model into production, it has no use and remains only a theoretical model. Compare this process to serving food in a restaurant. Data preparation is when you prepare the raw ingredients. Model development is when you experiment with different recipes, and model serving is when you finalize the menu to serve the meal to customers. Now, it's important to note that an ML workflow isn't linear, it's iterative. For example, during model training, you might need to return to the raw data and generate more useful features to feed the model. When monitoring the model during model serving, you might find data drifting or the accuracy of your prediction might suddenly drop. You might need to check the data sources and adjust the model parameters. Fortunately, these steps can be automated with MLOps. You'll learn more about this later in this module. Although the main stages remain the same, you have two options to set up the workflow with Vertex AI. The first choice is to use AutoML, a no-code solution that lets you build an ML model through UI. It's user friendly and doesn't require a lot of ML expertise. Also, no coding skills are needed. Alternatively, you can code the workflow with Vertex AI Workbench or Colab using Vertex AI Pipelines. Vertex AI Pipelines is essentially a toolkit that includes pre-built SDKs, or Software Development Kits, which are the building blocks of a pipeline. Coding the pipeline is a good option if you're an experienced ML engineer or data scientist, and want to automate the workflow programmatically. Let's focus on AutoML first and then explore the code-based approach later in this module.2. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.