In this article, I’ll introduce Hugging Face — a company that provides tools and platforms for building machine learning applications. We’ll explore what Hugging Face offers, how to use its resources, and how it’s transforming the field of artificial intelligence. Whether you’re an experienced data scientist or a curious beginner, this guide will help you understand and utilize Hugging Face’s capabilities. What is Hugging Face? Let’s start with the basics — what is Hugging Face? Hugging Face initially emerged as a chatbot company that later pivoted to focus on developing cutting-edge open-source NLP technologies. Its flagship library, Transformers, is a game-changer. It simplifies the complex tasks associated with NLP by providing easy access to pre-trained models. This library is built on transformer architectures, celebrated for their ability to handle quantum leaps in processing natural language at scale and with unprecedented accuracy. The beauty of Hugging Face is its democratization of AI technology. By offering accessible tools and models, Hugging Face allows practitioners of various levels to tap into the potential of transformers without needing extensive computational resources or deep expertise in machine learning. How to Get Started with Hugging Face We are going to explore multiple ways to work with Hugging Face. The first way will be through https://huggingface.co/ website. Before you start using it, you must create an account there. https://huggingface.co/ There are three main sections you should know about: Models


Datasets


Spaces Models Models Datasets Datasets Spaces Spaces To use models and datasets, you would need to use the Python language, transformer library, and one of the machine learning frameworks. But if you don’t have programming skills, you can use Spaces to play with different AI models. Hugging Face Models The Hugging Face Model Hub is a repository where you can find pre-trained models for a wide range of tasks, such as natural language processing (NLP), computer vision, audio processing, and more. Hugging Face Model Hub These models are contributed by the community and Hugging Face itself, covering various architectures like BERT, GPT, T5, and others. Users can access thousands of models that have been pre-trained on large datasets, allowing for efficient fine-tuning on their custom tasks. Each model comes with a model card, providing important information such as its intended use case, limitations, and performance metrics. Keep in Mind: Keep in Mind: High-performance models may require significant computational resources and GPU memory to run effectively.
Not all models are free to use for commercial purposes. You should check the specific licensing information provided for each model. High-performance models may require significant computational resources and GPU memory to run effectively. Not all models are free to use for commercial purposes. You should check the specific licensing information provided for each model. Hugging Face Datasets The Datasets library at Hugging Face is designed to provide a simple, efficient way of accessing a wide array of datasets that are useful for machine learning and data-driven projects. Datasets There are datasets for text, audio, image, and tabular data across multiple domains and languages. All of the datasets seamlessly integrate with Hugging Face’s other tools and libraries like Transformers and Tokenizers. Keep in mind: Keep in mind: Some datasets are vast, making them challenging to handle without ample disk space and memory.
Some datasets may have restrictions regarding their use, especially for commercial applications. Check for any licenses before use.
Data might not always be perfect, and additional cleaning or processing may be necessary to fit specific use cases. Some datasets are vast, making them challenging to handle without ample disk space and memory. Some datasets may have restrictions regarding their use, especially for commercial applications. Check for any licenses before use. Data might not always be perfect, and additional cleaning or processing may be necessary to fit specific use cases. Hugging Face Spaces Spaces on Hugging Face are a recent addition, providing an easy-to-use platform for users to deploy their machine learning models and showcase interactive AI applications. Spaces Hugging Face Spaces offers both free and paid options. Free Spaces come with default hardware resources of 16GB RAM, 2 CPU cores, and 50GB of non-persistent disk space. Many models have interactive demos, sharing them with the community without needing your server. You can create public Spaces accessible to all or private ones restricted to selected collaborators or team members. Keep in Mind: Keep in Mind: Restrictions on computational resources may apply, impacting the performance of heavy models or datasets hosted in Spaces.
There might be limitations based on your account level (e.g., free vs. paid subscriptions) that determine the number of Spaces you can maintain and the resources they can consume. Restrictions on computational resources may apply, impacting the performance of heavy models or datasets hosted in Spaces. There might be limitations based on your account level (e.g., free vs. paid subscriptions) that determine the number of Spaces you can maintain and the resources they can consume. How to Use Hugging Face Spaces To explore existing applications on Hugging Face Spaces, follow these steps: 1. Visit the Hugging Face Spaces Directory: Navigate to the Spaces page, where a variety of machine learning applications are showcased. 1. Visit the Hugging Face Spaces Directory: Spaces 2. Browse Applications: On the Spaces page, applications are organized by categories such as Image Generation, Text Generation, Language Translation, and more. 2. Browse Applications: 3. Explore Featured and Trending Spaces: Click on any application name to access its dedicated page, where you can interact with the demo and view additional details. 3. Explore Featured and Trending Spaces: 4. Interact with Applications: Many Spaces offer interactive demos. Follow the on-screen instructions to use the application. 4. Interact with Applications: How to Use Hugging Face Models To use models, we would need to install the transformers library, which offers access to numerous pre-trained models. models, What is Hugging Face Transformers? Transformers are a type of deep learning model architecture that excels at understanding the context and nuances of language. The library provides a plethora of pre-trained models and fine-tuning tools that are invaluable for various tasks such as text classification, tokenization, translation, summarization, and much more. With just a few lines of code, you can integrate these advanced models into your projects, significantly reducing the time and effort typically required to train models from scratch. This accessibility lowers the barrier to entry, fostering a more inclusive environment where more people can innovate with AI. How to Use Hugging Face Transformers Before diving into the specific applications, ensure your development environment is set up correctly. You’ll need it installed on your system: IDE (VS Code or any other)
Python language
Transformers library
Machine Learning Framework (PyTorch or TensorFlow) IDE (VS Code or any other) VS Code Python language Transformers library Machine Learning Framework (PyTorch or TensorFlow) Step 1: Install Necessary Libraries We are going to use the terminal. Use these commands: Install Python: Install Python: sudo apt updatesudo apt install python3 sudo apt updatesudo apt install python3 Use a Virtual Environment Use a Virtual Environment Instead of installing packages globally, create a virtual environment: python3 -m venv venv

source venv/bin/activate python3 -m venv venv

source venv/bin/activate Install Transformers and a few other libraries: Install Transformers and a few other libraries: pip install transformers datasets evaluate accelerate pip install transformers datasets evaluate accelerate You’ll also need to install your preferred machine learning framework. PyTorch and TensorFlow are two of the most popular open-source frameworks for deep learning. PyTorch was developed by Facebook’s AI Research lab (FAIR) and released in 2016. It has gained immense popularity due to its ease of use and flexibility. PyTorch TensorFlow was developed by the Google Brain team and open-sourced in 2015. It’s one of the oldest and most widely adopted frameworks for building deep learning models. TensorFlow Install PyTorch: Install PyTorch: pip install torch pip install torch (Optional) For GPU acceleration, install the appropriate CUDA drivers. Simply follow the instructions on the NVIDIA website https://developer.nvidia.com/cuda-downloads (Optional) https://developer.nvidia.com/cuda-downloads CUDA is a parallel computing platform and application programming interface (API) model created by NVIDIA specifically for its line of GPUs (Graphics Processing Units). It allows developers to utilize NVIDIA GPU hardware for general-purpose processing (GPGPU), extending beyond the traditional use in graphics and enabling significant acceleration of computational tasks in areas like machine learning, scientific computing, and data analysis. Step 2: Explore the Model Hub Navigate to the Hugging Face model hub (https://huggingface.co/models) and explore the available models. https://huggingface.co/models Once you find the one you want to try, click on it and copy the transformer’s code into your IDE (keep in mind that there could be code not only for transformers. For instance, the model https://huggingface.co/Salesforce/blip-image-captioning-base is designed to generate descriptive captions for images. https://huggingface.co/Salesforce/blip-image-captioning-base The code: import requests  
from PIL import Image  
from transformers import BlipProcessor, BlipForConditionalGeneration  
  
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")  
  
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")  
  
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'   
  
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')  
  
# conditional image captioning  
text = "a photography of"  
  
inputs = processor(raw_image, text, return_tensors="pt")  
  
out = model.generate(**inputs)  
  
print(processor.decode(out[0], skip_special_tokens=True))  
  
# >>> a photography of a woman and her dog  
  
# unconditional image captioning  
  
inputs = processor(raw_image, return_tensors="pt")  
  
out = model.generate(**inputs)  
  
print(processor.decode(out[0], skip_special_tokens=True)) import requests  
from PIL import Image  
from transformers import BlipProcessor, BlipForConditionalGeneration  
  
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")  
  
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")  
  
img_url = 'https://storage.googleapis.com/sfr-vision-language-research/BLIP/demo.jpg'   
  
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert('RGB')  
  
# conditional image captioning  
text = "a photography of"  
  
inputs = processor(raw_image, text, return_tensors="pt")  
  
out = model.generate(**inputs)  
  
print(processor.decode(out[0], skip_special_tokens=True))  
  
# >>> a photography of a woman and her dog  
  
# unconditional image captioning  
  
inputs = processor(raw_image, return_tensors="pt")  
  
out = model.generate(**inputs)  
  
print(processor.decode(out[0], skip_special_tokens=True)) And here is the result: In this way, you can try over 1000 different models. But keep in mind, to effectively run Hugging Face’s Transformers library with PyTorch, it’s recommended to have a system with at least 8 GB of RAM and a GPU with 4 GB of VRAM; however, for optimal performance with larger models, 64 GB of RAM and a GPU with 24 GB of VRAM are advisable. Video Tutorial I also created a detailed video tutorial. Watch it now or save it for later. Watch on YouTube: Hugging Face Tutorial Watch on YouTube: Hugging Face Tutorial Hugging Face Tutorial Conclusion By engaging with Hugging Face’s resources, you can enhance your AI projects and contribute to the broader AI community’s growth and innovation. Please give it a short and share with me your thoughts and learnings below! Cheers! :)

Before

Brain

Salesforce

Google

Keep

Facebook

Mind

NVIDIA

scale

YouTube

Google AI Studio Tutorial for Beginners

You Can Now Access Gemini Code Assist for Free

Read My Blog

Support my work

Too Long; Didn't Read

Here's How the Pros Use Hugging Face to Run Countless AI Models on their PCs

Here's How the Pros Use Hugging Face to Run Countless AI Models on their PCs

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

10 Git Commands Senior Devs Should Know (and Use Daily)

The Noonification: Use This 7-Step McKinsey Framework to Solve Any Problem (1/10/2023)

The Noonification: A Taxonomy of Inclusiveness (1/11/2024)

The Noonification: What is the InfiniteNature-Zero AI Model? (11/19/2022)

10 Ways AI Has Changed Our Lives

100 Days of AI, Day 8: Experimenting With Microsoft's Semantic Kernel Using GPT-4

10 Git Commands Senior Devs Should Know (and Use Daily)

The Noonification: Use This 7-Step McKinsey Framework to Solve Any Problem (1/10/2023)

The Noonification: A Taxonomy of Inclusiveness (1/11/2024)

The Noonification: What is the InfiniteNature-Zero AI Model? (11/19/2022)

10 Ways AI Has Changed Our Lives

100 Days of AI, Day 8: Experimenting With Microsoft's Semantic Kernel Using GPT-4

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps