Democratizing AI: How Hugging Face & KNIME Make It Easier

Hugging Face in KNIME: Increasing the Possibilities of Open-Source AI

Open-source data analytics software meets open-source LLMs

Artificial intelligence , with its advanced language models, is transforming the way we interact with technology and, ultimately, with each other. As these technologies continue to evolve, we have seen significant advances in a wide range of applications, from virtual assistants that help us in our daily lives to recommendation systems that influence our consumer decisions. In this context, the open-source philosophy plays a crucial role in the development and accessibility of these technologies.

The importance of open-source

Organizations like Hugging Face and KNIME have embraced the open-source philosophy by sharing their resources and tools with the global community. This approach not only encourages collaboration and innovation but also democratizes access to artificial intelligence. By providing open-source language models and libraries, Hugging Face enables developers and researchers around the world to build on a common foundation and accelerate progress in the field of AI. KNIME, on the other hand, develops an open-source platform, KNIME Analytics Platform , for the automation of data and analysis processes, making it easier for organizations to implement artificial intelligence solutions in an efficient and personalized way.

Open source not only drives innovation but also ensures greater transparency and trust in AI technology. As these tools continue to impact our society, the collaboration and accessibility that open source promotes are essential for AI solutions to be ethical, fair, and available for the benefit of all.

KNIME

KNIME, named after “Konstanz Information Miner”, develops KNIME Analytics Platform , an open-source data analysis platform that has seen significant growth since its inception. It was developed by a group of researchers at the University of Konstanz, Germany, in the early 2000s.

Over the years, KNIME has evolved and become a versatile tool for data analysis that is used in a wide variety of sectors, including scientific research, industry, bioinformatics etc. Its intuitive graphical interface, which allows users to create visual workflows by connecting nodes representing different data processing operations, has made it accessible to both technical and non-technical users. KNIME focuses on democratizing data analysis and making it accessible to a broader audience.

In addition to that, the KNIME ecosystem is designed to emphasize the swift integration between the platform and external tools and services, such as databases, big data frameworks, APIs, programming languages such as Python or R, cloud services and more. All of that expands the platform’s usefulness and versatility. In short, KNIME strives to make data analysis accessible, efficient, and collaborative, making it a valuable tool in the world of data science and analytics.

Hugging Face

Hugging Face started its journey in 2017 as a simple chatbot app for teenagers. However, over time, it became a beacon of open-source artificial intelligence. Today, it is an innovation hub that embraces open-source collaboration as a mantra. Its evolution has been extraordinary and it has become an essential resource for AI enthusiasts.

Currently, Hugging Face Hub hosts an impressive collection of open-source resources. With more than 120,000 models, 20,000 data sets, and 50,000 demo applications, this platform provides an online space for collaboration and co-building in the field of machine learning. Its goal is to democratize Machine Learning (like KNIME with data analytics altogether), allowing anyone to participate in the development and use of AI technology.

The philosophy behind the Hub is that no company, no matter how large, can solve all machine learning challenges alone. The real solution lies in sharing knowledge and resources in a community-centered approach.

The Hub hosts Git -based repositories containing models, datasets, and demo applications:

• Models: Include a wide variety of open-source models for NLP , computer vision, audio, etc. Each model repository provides information about the limitations and biases of the models and allow direct interaction with them.
• Datasets: Offer a wide variety of datasets in more than 100 languages for NLP, computer vision, audio and multimodal tasks. The data sets are accompanied by detailed documentation and are accessible programmatically.
• Spaces: Allow users to create interactive demo applications to present projects and collaborate in the Machine Learning ecosystem.

What about Hugging Face + KNIME Analytics Platform?

The integration of Hugging Face into KNIME offers a new dimension in data analysis. KNIME, an open-source platform, is a perfect ally for Hugging Face. Together, these two tools enable data and AI professionals to optimize and expand the possibilities of their workflows. From natural language processing to text classification, KNIME and Hugging Face make tasks more efficient and effective.

Let’s have a look at a concrete example.

Prompt Hugging Face Hub LLM

This workflow shows how to connect to an LLM stored on Hugging Face Hub and prompt it to generate text based on the user’s directions.

To run this workflow, you need a free access token for Hugging Face Hub.

Credential Configuration

We start by entering the access token that we have previously obtained in the Credentials Configuration node. Providing the username is optional.

If you don’t already have a free access token, create an account at https://huggingface.co/ , go to https://huggingface.co/settings/tokens and create a new token.

It is recommended to use the Credentials Configuration node in workflows where logging in to an external service is required. This node offers different levels of encryption that make log in safer.

Hugging Face Models

We authenticate to Hugging Face with the HF Hub Authenticator node with the variable that we created before in the Credentials Configuration node.

Next, we choose the model that is hosted on Hugging Face Hub with the HF Hub LLM Connector node. In this case, we choose the HuggingFaceH4/zephyr-7b-alpha . We can parameterize it as we want with all the options that this node offers us.

LLM Prompter

Now that we have our model loaded, let the magic begin ??.

We are leveraging a powerful language model capable of generating text almost like a human being, based on given input. This allows us to tackle a wide range of natural language-related tasks, such as content creation, summaries, translations, classifications, and question answering.

For each row in the input table, the LLM Prompter node sends an instruction to the LLM and receives the corresponding response. The rows are treated independently, that is, the LLM cannot remember the content of previous rows or how it responded to them.

In this case, we create a column with the task and another one with the prompt with the Table Creator node. We can use the KNIME potential and load our table from wherever we want, be it CSV, Excel, table… etc.

Model response

It can be improved by adapting our prompt and parameters to what we ultimately look for. Nevertheless, even with a simple, generic prompt, the result is incredible.

Conclusion

From humble beginnings to the AI revolution, Hugging Face and KNIME have come an amazing path. Their focus on open-source, collaboration and innovation has led to the creation of two platforms that empower everyone, paving the way for a new era of data analytics and AI. This story is a testament to how collaboration and open source are driving innovation in artificial intelligence. Join this exciting revolution!

Don’t miss any more interesting content by subscribing to my Spanish newsletter on LinkedIn. Also follow me on LinkedIn, Twitter , Instagram , YouTube , Tik Tok and Facebook to stay up to date with the latest trends in artificial intelligence and business technology!