How to deploy machine learning on google cloud platform

August 10, 2021

Google Cloud Platform is one of the latest cloud technology that has been able to capture a large market share in cloud computing. GCP has a wide variety of basic services like computing, storage, processing, data warehouses and database management systems. What Google has focused mainly is to provide services in the domain that has real time application for users and has integration of latest technologies. Google provides many services in the field of Artificial Intelligence. It has focused on creating specific services like google vision that is mainly used for image related datasets. It has text-to-speech-to-text service that comes handy when someone is trying to make a voice command based product.

The Artificial Intelligence services provided by Google Cloud Platform are:-

Vertex AI- A service that helps in deploying machine learning models, and scaling which also helps accelerating data preparation.

Virtual Agent- It is a very advanced ML service which is very useful in creating a virtual assistant or a chat-bot that can interact with customers 24/7 without any human intervention and is fast to deploy.

Vision OCR- It is a Image based service that parses text data from handwritten or printed pdf or image and converts into a machine understandable text that can be used for processing or digitizing the documents.

These services are trained and ready-to-use, but what if you want to build a product that needs a custom Machine learning algorithm with your dataset provided. One day you will need to deploy your machine learning model so for deploying this you have platforms like GCP, Azure, AWS.

Google Cloud Platform is one of them and often used. Google Cloud Platform GCP is very friendly as compared to others. Only little change or work has to be done. The Data lifecycle in a MLOps environment is shown in the diagram.

The procedure to train and deploy machine learning model using Google Cloud Services is shown below:

Step 1: Firstly, we should pack our machine learning model properly your model properly

Step 2: Build a bucket on Google Cloud Storage

Step 3: Now in this Cloud Storage Bucket add packed machine learning model

Step 4: Building an AI Platform Prediction Model Resource

Step 5: Building an AI Platform Prediction Version Resource

Step 1: Firstly, we should pack our model machine learning model properly your model properly

GCP expects our model should be packaged as a single file or sometimes multiple files. There are different ways to pack your models also you can pack or deploy your machine learning model by Packaging your machine learning model with joblib or with pickle file like  model.joblib, model.pkl,…

Step 2: Build a bucket on Google Cloud Storage

In order to access our packaged models by google cloud platform resources. We should upload  to GCP and store properly. Gcp offers a standard solution to store

your model files in buckets. An easy way to do so is to pass the following gsutil command:

$gsutil mb -l gs://

Step 3: Now in this Cloud Storage Bucket add packed machine learning model