Lecture Notes 8

This document discusses deploying a machine learning model as a web service using Streamlit. Streamlit allows embedding HTML directly in Python files, avoiding separate templates. The document walks through importing libraries like Streamlit, loading a pretrained linear regression model, defining functions for predictions and running the web app, and obtaining predictions by entering values and clicking a button. When run, Streamlit launches a local web page where users can interact with the model by inputting values and receiving predictions.

Uploaded by

fgsfgs

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

53 views

Lecture Notes 8

Uploaded by

fgsfgs

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 4

Chapter 3 Machine Learning Deployment as a Web Service

eploying a Machine Learning Model Using

D
Streamlit
Streamlit is an alternative to Flask for deploying the machine learning
model as a web service. The biggest advantage of using Streamlit is that it
allows you to use HTML code within the application Python file. It doesn’t
essentially require separate templates and CSS formatting for the front-end
UI. However, it is suggested that you create separate folders for templates
and style guides for a more complex application. To install Streamlit, we
can simply use pip to install Streamlit in our terminal.
We are going to use the same dataset that we used when building the
Flask app model. The only content that is going to change is in the app.py
file. The first set of commands is to import the required libraries such as
joblib and streamlit.

[In]: import pandas as pd

[In]: import numpy as np
[In]: import joblib
[In]: import streamlit

In the next step, we import the trained linear regression model to be

able to predict on the test data.

[In]: model=open("linear_regression_model.pkl","rb")
[In]: lr_model=joblib.load(model)

The next step is to define a function to make predictions using the

trained model. We pass the five input parameters in the function and do
a bit of reshaping and data casting to ensure consistency for predictions.
Then we create a variable to save the model predictions result and return it
to the user.

[In]: def lr_prediction(var_1,var_2,var_3,var_4,var_5):

pred_arr=np.array([var_1,var_2,var_3,var_4,var_5])

76
Chapter 3 Machine Learning Deployment as a Web Service

      preds=pred_arr.reshape(1,-1)
      preds=preds.astype(int)
      model_prediction=lr_model.predict(preds)
          return model_prediction

In the next step, we create the most important function. We accept the
user input from the browser and render the model’s final predictions on
the web page. We can name this function anything. For example, I have
used run (since it does the same thing as Flask’s app.run). In this function,
we include the front-end code as well such as defining the title, theme,
color, background, etc. For simplicity purposes, I have kept it basic, but this
can have multiple levels of enhancements. For more details, you can visit
the Streamlit website. Next, we create five input variables to accept the user
input values from the browser. This is done using Streamlit’s text_input
capability. The final part contains the model prediction, which gets the
input from our lr_prediction function defined earlier and gets rendered
in the browser through streamlit.button.

[In]: def run():

streamlit.title("Linear Regression Model")
html_temp="""

"""
streamlit.markdown(html_temp)
var_1=streamlit.text_input("Variable 1")
var_2=streamlit.text_input("Variable 2")
var_3=streamlit.text_input("Variable 3")
var_4=streamlit.text_input("Variable 4")
var_5=streamlit.text_input("Variable 5")

77
Chapter 3 Machine Learning Deployment as a Web Service

prediction=""

            if streamlit.button("Predict"):
                  prediction=lr_prediction(var_1,var_2,var_3,
var_4,var_5)
      streamlit.success("The prediction by Model : {}".
format(prediction))

Now that we have all the steps mentioned in the application file, we
can call the main function (run in our case) and use the streamlit run
command to run the app.

[In]: if __name__=='__main__':
run()
[In]: streamlit run app.py

Once we run the previous command, we will soon see the app up and
running on port 8501, as shown in Figure 3-6. We can simply click the link
and access the app.

Figure 3-6. Access running app

Once we are at http://localhost:8501/, we will see the screen shown

in Figure 3-7.

78
Chapter 3 Machine Learning Deployment as a Web Service

Figure 3-7. User input page

As you can see, we have nothing fancy here in the UI, but it serves the
overall purpose for us to be able to interact with the model/app behind the
scenes. It is similar to what we had built using Flask, but it needed much
less HTML and CSS code. We can now go ahead and fill in the values, as
shown in Figure 3-8, and get the model prediction. For comparison sake,
we fill in the same values that we filled in the Flask-based app.