How to Use Hugging Face API

Last Updated : 23 Jul, 2025

Hugging Face is one of the best platforms for machine learning, and artificial intelligence (AI) models. Using the Hugging Face API, we can easily interact with various pre-trained models for tasks like text generation, translation, sentiment analysis, etc.

In this article, we are going to discuss how to use the Hugging Face API with simple steps and examples.

Table of Content

Using HuggingFace APIs for Computer Vision Tasks

Why Use Hugging Face API?

Ease of use: We don't have to spend hours building models from scratch. The API streamlines the process, making it accessible even to beginners.
Pre-trained Models: Hugging Face offers a treasure trove of pre-trained models, which means we can pick one that suits what we need and start using it right away.
Time-Saving: We can integrate AI into our projects with the Hugging Face API and we can focus on what really matters that is developing our application rather than getting bogged down in the intricacies of machine learning.

Setting Up the Hugging Face API

We have to first setup the Hugging Face API before using it. Now let's see how we can

Step 1: Create a Hugging Face Account and Get API Token

Create an account on Hugging Face. After creating an account, go to your account settings and get your HuggingFace API token.

Step 2: Install the Hugging Face Hub Library

The Hugging Face Hub library helps us in interacting with the API. This library provides an easy-to-use interface for interacting with the Hugging Face models and making requests. We can easily install it using pip:

pip install huggingface_hub

Now that we have our API token and have installed the library, now we can start making requests to the API.

Step 3: Import the Necessary Library and Authenticate

In our Python environment, we start by importing the InferenceClient class from the Hugging Face Hub. Then, we can authenticate using our API token:

Python

from huggingface_hub import InferenceClient

# Replace 'your_hf_api_token_here' with your actual API token
client = InferenceClient(token="your_hf_api_token_here")

Once authenticated, we are ready to use the API.

Using HuggingFace API for NLP Tasks

Now, we are going to see different Natural Language Processing (NLP) tasks using the Hugging Face API, focusing on Text Generation, Named Entity Recognition (NER), and Question Answering. These tasks demonstrate the capabilities of advanced models like GPT-2 and BERT, which can significantly enhance our understanding and interaction with text.

Text Generation with GPT-2

Using GPT-2 for text generation is straightforward with Hugging Face's API. By sending an input prompt, we can generate coherent, engaging text for various applications.

Here’s how to get started:

Setup: Import the requests library and set up API details with the GPT-2 model URL and authorization headers.
Query Function: Define a function that sends a POST request to the API with a prompt, receiving generated text in JSON format.

Python

import requests

API_URL = "https://api-inference.huggingface.co/models/gpt2"
headers = {"Authorization": "Bearer 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"}  # Replace with your valid token

# Define the query function to send a request to the Hugging Face API
def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

# Define the text generation function using the query function
def generate_text(prompt):
    response = query({"inputs": prompt})
    
    # Print the raw response to inspect its structure
    print(response)
    
    # Safely access the response, handling errors
    if isinstance(response, list) and len(response) > 0:
        return response[0].get('generated_text', "No text generated.")
    else:
        return "Error: Unable to generate text."

# Provide a text prompt
prompt = "Once upon a time"
print(generate_text(prompt))

Output:

[{'generated_text': 'Once upon a time it seemed like Lightsaber on Coruscant would be a terrible threat. to Spike they decided . . . .and they continued their'}]

Named Entity Recognition (NER) with BERT

Named Entity Recognition (NER) identifies entities like names, locations, and organizations in text, a valuable tool for data extraction and information retrieval.

API Setup: Use the bert-large-cased-finetuned-conll03-english model and authenticate with a Bearer token.
NER Function: Define a query function that sends text as JSON and processes the API's response.

Python

import time
import requests

API_URL = "https://api-inference.huggingface.co/models/dbmdz/bert-large-cased-finetuned-conll03-english"
headers = {"Authorization": f"Bearer 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"}  # Replace YOUR_API_KEY with your actual key

def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

def ner(text):
    while True:
        response = query({"inputs": text})
        if 'error' in response and 'loading' in response['error']:
            print(response['error'])
            time.sleep(10)  # Wait for 10 seconds before retrying
        else:
            return response

text = "Hugging Face Inc. is based in New York City."
print(ner(text))

Output:

'score': 0.9958662986755371, 'word': 'Hugging Face Inc', 'start': 0, 'end': 16}, {'entity_group': 'LOC', 'score': 0.9992396235466003, 'word': 'New York City', 'start': 30, 'end': 43}]

Question Answering with RoBERTa

Using RoBERTa for Question Answering enables contextualized responses based on provided information, ideal for chatbots and Q&A applications.

API Setup: Use roberta-base-squad2 and set up a function that takes a question and context.
Response Handling: The model returns the best answer based on the input context.

Python

API_URL = "https://api-inference.huggingface.co/models/deepset/roberta-base-squad2"

def answer_question(question, context):
    response = query({
        "inputs": {
            "question": question,
            "context": context
        }
    })
    return response['answer']

question = "What is Hugging Face?"
context = "Hugging Face is a company based in New York."
print(answer_question(question, context))

Output:

a company based in New York

Using HuggingFace APIs for Computer Vision Tasks

For Computer Vision, Hugging Face provides image-processing models that facilitate classification, object detection, and segmentation.

Image classification

Using image classification, we can identify the contents of an image.

Here’s how to set it up:

API Setup: Use an image classification model with authentication details.
Classification Function: The function reads the image, encodes it, and sends it to the API, returning label predictions.

Python

import requests
import base64
from PIL import Image, ImageDraw, ImageFont
import io

API_URL = "https://api-inference.huggingface.co/models/facebook/detr-resnet-50"
headers = {"Authorization": f"ENTER YOUR API KEY"}  # Replace with your actual API key

def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

def detect_objects(image_path):
    # Open the image file in binary read mode
    with open(image_path, "rb") as f:
        image_data = f.read()

    # Encode the image data to base64
    image_base64 = base64.b64encode(image_data).decode('utf-8')

    # Prepare the request payload with the base64 encoded image
    response = query({"inputs": image_base64})

    # Handle errors in the response
    if 'error' in response:
        print(f"Error: {response['error']}")
        return None

    return response

def draw_boxes(image_path, detections, output_path="output_with_boxes.jpg"):
    # Open the image
    image = Image.open(image_path)
    draw = ImageDraw.Draw(image)

    # Define font (adjust font size as necessary)
    try:
        font = ImageFont.truetype("arial.ttf", 20)
    except IOError:
        font = ImageFont.load_default()

    # Draw each bounding box on the image
    for detection in detections:
        box = detection['box']
        label = detection['label']
        score = detection['score']
        
        # Bounding box coordinates
        xmin, ymin, xmax, ymax = box['xmin'], box['ymin'], box['xmax'], box['ymax']
        
        # Draw the bounding box
        draw.rectangle([(xmin, ymin), (xmax, ymax)], outline="red", width=3)
        
        # Add label and score above the bounding box
        text = f"{label} ({score:.2f})"
        
        # Get text size using textbbox (works in Pillow 8.0+)
        text_bbox = draw.textbbox((xmin, ymin), text, font=font)
        text_width = text_bbox[2] - text_bbox[0]
        text_height = text_bbox[3] - text_bbox[1]
        
        # Draw background rectangle for text
        draw.rectangle([(xmin, ymin - text_height), (xmin + text_width, ymin)], fill="red")
        
        # Draw the label text
        draw.text((xmin, ymin - text_height), text, fill="white", font=font)

    # Save the image with bounding boxes
    image.save(output_path)
    print(f"Image saved with bounding boxes as {output_path}")

# Path to your image file
image_path = "/content/animal images.jpg"
result = detect_objects(image_path)

# Draw boxes and save the result if detection is successful
if result is not None:
    draw_boxes(image_path, result)

Output:

Image saved with bounding boxes as output_with_boxes.jpg

Object Detection with Facebook DETR

Object detection allows precise identification and localization of objects within images.

API Setup: Use the DETR model for detecting objects in photos.
Object Detection Function: Reads, encodes, and sends the image to the API, returning coordinates for detected objects.

Python

import requests
import base64
from PIL import Image, ImageDraw
import io

API_URL = "https://api-inference.huggingface.co/models/facebook/detr-resnet-50"
headers = {"Authorization":  f"Bearer API key"}  # Replace with your actual API key

def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

def detect_objects(image_path):
    # Open the image file in binary read mode
    with open(image_path, "rb") as f:
        image_data = f.read()

    # Encode the image data to base64
    image_base64 = base64.b64encode(image_data).decode('utf-8')

    # Prepare the request payload with the base64 encoded image
    response = query({"inputs": image_base64})

    # Handle errors in the response
    if 'error' in response:
        print(f"Error: {response['error']}")
        return None

    return response

def draw_boxes(image_path, detections, output_path="output_with_boxes.jpg"):
    # Open the image
    image = Image.open(image_path)
    draw = ImageDraw.Draw(image)

    # Draw each bounding box on the image
    for detection in detections:
        box = detection['box']
        label = detection['label']
        score = detection['score']
        
        # Bounding box coordinates
        xmin, ymin, xmax, ymax = box['xmin'], box['ymin'], box['xmax'], box['ymax']
        
        # Draw the bounding box
        draw.rectangle([(xmin, ymin), (xmax, ymax)], outline="red", width=3)
        
        # Add label and score
        draw.text((xmin, ymin), f"{label} ({score:.2f})", fill="red")

    # Save the image with bounding boxes
    image.save(output_path)
    print(f"Image saved with bounding boxes as {output_path}")

# Path to your image file
image_path = "dog.jpg"
result = detect_objects(image_path)

# Draw boxes and save the result if detection is successful
if result is not None:
    draw_boxes(image_path, result)

Output:

Image saved with bounding boxes as output_with_boxes.jpg

Conclusion

We can use the Hugging Face API for machine learning models for performing various tasks. It is a powerful tool . With only just a few lines of code, we can integrate pre-trained models into our applications without worrying about the heavy lifting of training our own model. Whether we are building a chatbot, performing language translation, or generating text, Hugging Face offers a simple and effective solution through its API. You can explore different models in the Hugging Face Model Hub and experiment with the code provided above to start using these models in your projects.

bug8wdqo

Improve

Article Tags :

How to Use Hugging Face API

Why Use Hugging Face API?

Setting Up the Hugging Face API

Step 1: Create a Hugging Face Account and Get API Token

Step 2: Install the Hugging Face Hub Library

Step 3: Import the Necessary Library and Authenticate

Using HuggingFace API for NLP Tasks

Text Generation with GPT-2

Named Entity Recognition (NER) with BERT

Question Answering with RoBERTa

Using HuggingFace APIs for Computer Vision Tasks

Image classification

Object Detection with Facebook DETR

Conclusion

Explore

Deep Learning Basics

Neural Networks Basics

Deep Learning Models

Deep Learning Frameworks

Model Evaluation

Deep Learning Projects

Thank You!

What kind of Experience do you want to share?