Gradio: A Library for Building Machine Learning Web Apps

In this tutorial, you will learn how to build a machine learning web app using the most popular interface, Gradio with the help of Python Programming.


These days, creating Machine Learning models is becoming easier and easier thanks to several open-source and proprietary-based administrations (for example Python, R, SAS). Even though experts may frequently believe that it is difficult to create interfaces that may be tested and shared with partners or colleagues.

Gradio, a free open-source Python package that helps you create model UIs that you can quickly share to connect to friends and partners, is one possible solution to this problem.

What is Gradio?

Gradio is an open-source Python package that allows you to quickly create simple-to-use, configurable UI components for any ML model, any API, or any subjective capability in only a few lines of code. It allows you to interact with your models in your web browser by dropping and dragging your photos, text, or recordings of your voice, for example, and viewing the outcome in real-time. You may integrate the GUI directly into your Python notebook or share the URL with anyone.

Gradio assists in the creation of an online GUI in a few lines of code, which is useful for displaying model presentation exhibits. It is quick, simple to set up, and ready to use, as well as shareable as the public connection via which anybody may remotely and concurrently execute the model in your system. Gradio works with a variety of media, including text, images, video, and sound. Aside from ML models, it can also be used for python code embeddings, as intended.

Features of Gradio

  •  With the help of Gradio, you may develop demos of your machine learning scripts that your clients/users/team members may find beneficial.
  • You may debug your models interactively while they are being developed.
  • Users can provide input on the model’s performance. As a consequence, you may develop your model more easily and quickly.


Gradio requires Python 3.7 or above, that’s it!. You can install Gradio on your  Python IDE by entering the following :

pip install gradio

Import Library

The Gradio library can be imported by entering the following:

import gradio as gr

In Code

Let’s first build a Hello World program using the function hell that takes a name as input.


import gradio as gr

def hell(name):
    return "Hello " + name + "!!"

hellop = gr.Interface(fn=hell, inputs="text", outputs="text")


The example below will automatically appear within the Python notebook, or in a browser at http://localhost:7860 if executed from a script.


Interface Class in Gradio

You’ll see that we defined a gradio.Interface to generate the demo. This Interface class may provide a user interface for practically any Python function. In the preceding example, we saw a basic text-based function. However, the function might be anything from a music generator to a tax calculator to (most typically) a pre-trained machine learning model’s prediction function.

The interface class contains three properties. These are listed below:

  • fn: the function for wrapping a user interface around.
  • inputs: which input element(s) to use, e.g. “text” or “picture” or “audio”.
  • outputs: which elements(s) to use for output like “text” or “picture” or “label”.

Gradio contains around 20 distinct components, the majority of which may be used as inputs or outputs. For more information please check the Manuals.)

Components Attributes 💻

We can develop bespoke components based on our specifications. For example, if we want to display huge text and a text suggestion, we may specify various options to the Interface class’s inputs parameter. Gradio provides a wide range of customizing options. Check the following code:


import gradio as gr

def hell(name):
    return "Hello " + name + "!"

hellp = gr.Interface(
    inputs=gr.Textbox(lines=2, placeholder="Name Here..."),



The following are some examples of customizations:

Check out this link for additional information on modifications.

Multitudinous Inputs and Outputs

Let us assume that we had a much more complicated function with various information sources and outputs. The code model below accepts a string, boolean, and number and outputs a string and a number. Examine how we pass a list of input and output components.


import gradio as gr
def begin(name, morn_ng, temp):
            msg = "Good morning" if morn_ng else "Good evening"
            greeting = "%s %s. It is %s degrees today" % (msg, name, temp)
            celsius = (temp - 32) * 5 / 9
            return greeting, round(celsius, 2)
a = gr.Interface(fn=begin, inputs=["text", "checkbox", gr.inputs.Slider(0, 100)], outputs=["text", "number"])


Working with Pictures

Let’s talk about images. The Image Input Interface takes any size NumPy array of the type (width, height, 3), with the last dimension representing the RGB colors. It will return a NumPy array with a picture in it. Furthermore, the Input interface contains an EDIT button that launches a tool for cropping, rotating, flipping, and adding filters to a picture. Isn’t it wonderful? We’ve noticed that manipulating images in this manner typically reveals hidden flaws in a model. The code below demonstrates how this tool works:


import numpy as np

import gradio as gr

def img(input_img):
    img_filter = np.array(
        [[0.393, 0.769, 0.189], [0.349, 0.686, 0.168], [0.272, 0.534, 0.131]]
    sepia_img =
    sepia_img /= sepia_img.max()
    return sepia_img

demo1 = gr.Interface(img, gr.Image(shape=(200, 200)), "image")










Furthermore, our Image input interface includes an ‘edit’ option that provides tools for cropping and zooming into photographs. We discovered that altering photographs in this manner can assist in revealing biases or hidden faults in a machine learning model!.

Using Machine Learning

After you’ve mastered the basics of the Gradio library, you’ll probably want to try it out like an ML(machine-learning) model.

  • Image Classification using TensorFlow

To begin, we will use image classification as an example of machine learning. We begin with the Inception Net Image Classifier, which is loaded using TensorFlow. We know it’s an image classification challenge because we’re using the Image input interface. This Image input interface provides a good test Inception Net interface where you can drag and drop images and also use the Alter button to edit images. The result would be a dictionary of labels and the confidence scores associated with them. Look at the image below:

  • Text Generation with Transformers (GPT-2)

Let’s begin with some text. We are employing the GPT-2 text generation model. In this example, we’re employing Text Text Interface. Simply enter some data, and the output will be displayed in the Output box. Look at the image below:

  • Question Answering with Robert-Base

The Roberta-based model has been fine-tuned using the SQuAD2.0 dataset. It has been trained for the job of Question Answering on question-answer pairings, including unanswerable questions. Look at the image below :


Check out the official documentation for further details: Link

Leave a Reply

Your email address will not be published. Required fields are marked *