TII Falcon-H1 Models Now Available on AWS Marketplace
This post was co-authored with Jingwei Zuo from TII.
We’re excited to announce the availability of Technology Innovation Institute (TII)’s Falcon-H1 models on both the Amazon Bedrock Marketplace and Amazon SageMaker JumpStart. With this launch, developers and data scientists can now leverage six instruction-tuned Falcon-H1 models (ranging from 0.5B to 34B parameters) on AWS, gaining access to a comprehensive suite of hybrid architecture models that combine traditional attention mechanisms with State Space Models (SSMs). This innovative approach delivers exceptional performance alongside unprecedented efficiency.
In this post, we’ll provide an overview of the capabilities of the Falcon-H1 models and demonstrate how to get started with them on both Amazon Bedrock Marketplace and SageMaker JumpStart.
Understanding TII’s Collaboration with AWS
The Technology Innovation Institute (TII) is a leading research institute headquartered in Abu Dhabi. As part of the UAE’s Advanced Technology Research Council (ATRC), TII focuses on advanced technology research and development across various fields, including AI, quantum computing, autonomous robotics, and cryptography. Notably, TII fosters an open and agile environment for international teams of scientists, researchers, and engineers to drive technological innovation and position Abu Dhabi and the UAE as a global hub for research and development—in alignment with the UAE National Strategy for Artificial Intelligence 2031.
Furthermore, TII and Amazon Web Services (AWS) are collaborating to broaden the accessibility of these made-in-the-UAE AI models on a global scale. By combining TII’s expertise in building large language models (Falcon-H1 being a prime example) with AWS’s cloud-based AI and machine learning services, professionals worldwide can now build and scale generative AI applications.
Delving into the Falcon-H1 Architecture
The Falcon-H1 architecture implements a parallel hybrid design, integrating elements from both the Mamba (link to paper) and Transformer (link to paper) architectures. This allows it to combine the faster inference speeds and reduced memory footprint characteristic of SSMs like Mamba with the effectiveness of Transformer attention mechanisms for understanding context and enhancing generalization capabilities. The Falcon-H1 architecture is designed to scale across multiple configurations, ranging from 0.5 to 34 billion parameters, and provides native support for 18 languages.
Key Advantages of the Falcon-H1 Series
- Performance: The hybrid attention-SSM model incorporates optimized parameters with adjustable ratios between attention and SSM heads. Consequently, this leads to faster inference times, lower memory usage, and strong generalization capabilities.
- Efficiency: The innovative architecture facilitates smaller model sizes while maintaining competitive performance, thereby reducing computational costs and simplifying deployment processes.
- Multilingual Support: Falcon-H1’s native support for 18 languages opens up opportunities to build AI solutions that cater to a global audience effectively.
Getting Started with Falcon-H1 on AWS
Accessing and deploying Falcon-H1 models on AWS is remarkably straightforward, thanks to both Amazon Bedrock Marketplace and SageMaker JumpStart.
- Amazon Bedrock Marketplace: Simply browse the marketplace, select your desired Falcon-H1 model variant, and provision it to your account. This streamlined approach simplifies deployment for applications requiring inference endpoints.
- SageMaker JumpStart: Explore the pre-trained models available in JumpStart, choose a Falcon-H1 model, and deploy it with ease. Jumpstart offers various options, including one-click deployments and fine-tuning capabilities for enhanced customization.
For detailed instructions on each deployment method, you can refer to the comprehensive Falcon-H1 technical blog post.
Conclusion
The availability of TII’s Falcon-H1 models on Amazon Bedrock Marketplace and SageMaker JumpStart represents a significant advancement in democratizing access to cutting-edge AI technology. These efficient and high-performance models empower developers and data scientists worldwide to build innovative generative AI applications with unprecedented ease and scalability. We encourage you to explore the potential of Falcon-H1 and contribute to shaping the future of artificial intelligence.
Source: Read the original article here.
Discover more tech insights on ByteTrending.
Discover more from ByteTrending
Subscribe to get the latest posts sent to your email.
