FAiNDER

What is FAiNDER?

FAiNDER (Find comprehensive AI data and insights) is an open-source web platform designed to transform how researchers and developers navigate the rapidly evolving AI landscape. Developed through a collaboration between the Barcelona Supercomputing Center (BSC) and Micron Technology, FAiNDER provides centralized, up-to-date information on the key system requirements of all major AI models to facilitate exploration and optimize hardware choices.

The platform offers a comprehensive and interactive database that allows users to delve into the software layers of leading AI models to help them understand the hardware resources needed for their efficient execution. FAiNDER addresses a key challenge in the AI field: the difficulty of keeping up with the constant stream of new models, each with different computational requirements, memory footprints, and software dependencies.

With FAiNDER, users choose the depth at which they want to explore the AI world. For students and junior researchers, the tool offers a starting point for exploring any AI model, providing a high-level view of the models' software architecture and their main features, along with a list of reliable sources for future exploration. For advanced users, the tool provides detailed information on all phases of the software pipeline, their main arithmetic operations, data types, memory footprint, and the target hardware platforms, enabling quick comparisons and deep technical insights.

The Story Behind FAiNDER

FAiNDER started as an internal repository for relevant AI studies, created to handle the overwhelming volume of publications in this rapidly evolving domain. The BSC Memory Systems team recognized the challenge researchers faced when trying to determine the right hardware for their AI workloads. Teams often struggled to match their specific AI models with appropriate computing resources, leading to either over-provisioning (wasting resources) or under-provisioning (bottlenecking performance).

The tool gradually evolved and was enriched until it became the must-have tool for any AI-related exploration and discussion between the BSC and their industrial partners, particularly Micron Technology. The collaboration between BSC and Micron, which has been ongoing since 2017 within the framework of a sponsored research agreement, has been instrumental in developing FAiNDER into the comprehensive platform it is today.

FAiNDER represents a commitment to making AI more accessible by democratizing the knowledge needed to run these models effectively. The platform has already been used in key research initiatives undertaken by the BSC, and it is poised to become an essential tool for AI research in the future. Additional joint efforts are now underway to address the challenges associated with memory systems bottlenecks in important application domains and to explore advanced memory systems for high-performance computing and AI.

Key Features

How It Works

FAiNDER provides an intuitive workflow for exploring AI model requirements:

Use Cases

FAiNDER is valuable for a wide range of users and applications:

• Students and Junior Researchers: Get started with exploring AI models through high-level overviews, understand basic requirements, and access reliable sources for further learning.
• Advanced Researchers: Access detailed technical information on software pipelines, memory footprints, and hardware requirements for in-depth analysis and comparison.
• Hardware Architects: Make informed decisions about system design by understanding the memory and computational requirements of different AI models.
• Software Developers: Identify the appropriate hardware platforms and software dependencies needed for deploying specific AI models.
• System Administrators: Optimize resource allocation and avoid over-provisioning or under-provisioning when setting up infrastructure for AI workloads.
• Research Teams: Quickly compare multiple AI models and their requirements to select the most appropriate models for specific research objectives.

Getting Started

FAiNDER is available as a web platform at FAiNDER.eu . No installation or setup is required—simply visit the website and start exploring. The platform is designed to be intuitive for users of all levels, from beginners who need high-level overviews to advanced researchers requiring detailed technical information.

The database is continuously updated with new AI models and research findings, ensuring that users always have access to the latest information. All data is sourced from credible research papers and industry publications, providing reliable information for making informed decisions about AI hardware and software requirements.