Iris Classification Project using Support Vector Machine (SVM) and BentoML

This project focuses on the development of a machine learning model using the Iris dataset. We use Support Vector Machine (SVM) as our model of choice which we implemented using sklearn. The model is built, trained, and then served using BentoML which provides an incredible way of handling models in production.

Project Set Up

The project setup is simple and straightforward. You need to have Python, BentoML and Scikit-Learn installed. To install BentoML and Scikit-learn, run the following command:

pip install bentoml scikit-learn

Building the Model

The model is built using the Iris dataset. SVM is chosen for this multi-class classification problem. The dataset contains 150 instances of iris flowers from three different species.

Here's an overview of how the model was saved:

import bentoml
from sklearn import datasets, svm

# Load the dataset
X, y = datasets.load_iris(return_X_y=True)

# Model Training
clf = svm.SVC(gamma="scale")
clf.fit(X, y)

# Create a BentoService
saved_model = bentoml.sklearn.save_model("iris_svm_clf", clf)
print(f"Model saved: {saved_model}") 

The save_model function is used to save the trained model instance.

Loading and Making Predictions with the Model

To load the model back into memory, following BentoML's load_model function is used:

import bentoml

iris_clf_runner = bentoml.sklearn.get("iris_svm_clf:latest").to_runner()

iris_clf_runner.init_local()

print(iris_clf_runner.predict.run([[5.9, 3.0, 5.1, 1.8]]))

You can then use this model for predicting unseen examples.

Deploying the Model with BentoML

BentoML is a fantastic library that simplifies the process of serving and deploying ML models. One of its features is the ability to automatically generate a Swagger API from your PyTorch model.

By using BentoML, you can easily serve your model as a high-performance API endpoint and consume it from a web frontend, mobile applications or in a microservices architecture.

Conclusion

This project is a demonstration of how one can build and deploy models with BentoML. You can access the complete code in this repository.

If you find it helpful, feel free to clone, download, or contribute to this project.