Computational Infrastructure
Providing HUN-REN researchers access to existing infrastructures, such as:
HUN-REN Cloud: A world-class scientific cloud infrastructure created by HUN-REN
SZTAKI and HUN-REN WIGNER RCP, with a decade-long history and continuous updates. Currently equipped with 72 GPU cards (with plans for significant expansion in the near
future), it has supported over 300 research projects from nearly 30 affiliated organisations.
https://science-cloud.hu/en
Komondor: Hungary's most powerful supercomputer dedicated to R&D purposes, equipped with over 200 GPU cards and a dedicated AI partition, operated by KIFÜ since 2023.
https://hpc.kifu.hu/hu/komondor
Running Secured GenAI in Your Own Environment
SZTAKI has developed the GenAI4Science service on the HUN-REN Cloud, which supports
generative AI tasks. This service allows researchers to safely leverage the capabilities of large language models (LLMs) while ensuring secure data storage in compliance with data management guidelines. With this new service, users can access public, open-source models on a user-friendly, chat-based interface, utilising their own data, documents, and other sources. The range of available models is continually expanding. While the service primarily relies on HUN-REN Cloud's internal resources, it also offers access to third-party generative AI services that support OpenAI interfaces.
https://science-cloud.hu/en/genai4science
Facilitating machine learning with distributed multi-GPU support
Research projects often require the development of complex platforms based on the coordinated operation of multiple components. Therefore, the scientific cloud infrastructure of the research network offers customisable, reliable, and scalable resources and templates (known as reference architectures) for users. In the field of machine learning, several widely-used technologies - including TensorFlow, Apache Spark, and HOROVOD - are available to support researchers. https://science-cloud.hu/en/reference-architectures
To ensure smooth usage, a series of instructional videos are available, covering a wide range of topics from project requests and accessing virtual resources to setting up complex AI environments.
https://science-cloud.hu/en/presentations
Upon request from Ambassadors or
researchers, support is also available for
accessing European AI infrastructures
(EuroHPC).
