Duration: 01.06.2018 – 30.09.2021
Acronym of the project: RI-SI – CLARIN
Applicant of the project: Institut “Jožef Štefan”
Project coordinator: Tomaž Erjavec, PhD
UM FERI coordinator: Darinka Verdonik, Matej Rojc
Other partners: IJŠ, UM FERI, UL FDV
Project funding: 466.000,00 EUR
LINK: DEVELOPMENT OF RESEARCH INFRASTRUCTURE
Project summary
The purpose of this investment project is to purchase new research equipment for improving existing CLARIN.SI hardware. This currently covers a cluster of three computers for the purposes of www.slovenscina.eu (2013 purchase), and a server (2017 purchase) covering all other current CLARIN.SI services (homepage, repository, concordance, web services). The implementation of the operation will enable significantly faster, more intensive, and high quality involvement in major international projects, thereby strengthening the network of research infrastructure of human resources in science, facilitating the free movement of people, ideas and knowledge in the European Research Area. The project defines individual development (general) goals as well as specific project goals.
Development goals (general goals)
To enable the Slovenian research community to access the latest international research infrastructure supporting high quality research and innovation in addressing major contemporary societal challenges;
Strengthen and coordinate national research infrastructure in areas where Slovenia has excellent active research teams;
Enable the optimization of the use of research infrastructures and make them accessible to external users;
To provide Slovenian research communities with an excellent research infrastructure that will help attract Slovenian partners in international research and innovation projects, ensure the employment of top researchers and motivate students to pursue a research career and
Provide support for Slovenian cooperation in upcoming international research infrastructures.
Specific objectives of the operation
To ensure the continued functioning of the infrastructure’s technical services by updating its hardware;
To enable storage of large multimodal language data in infrastructure by upgrading hardware with a new server containing fast and large disk capacities and enabling regular backup of all deposited data;
To enable CLARIN.SI to follow the big data paradigm by purchasing a powerful server and a large disk array that will allow the development of deletions for processing language databases in the way of data and integrated into the national supercomputer infrastructure;
To enable CLARIN.SI to provide publicly available online services for processing large volumes of Slovene texts by purchasing appropriate servers;
To establish a dedicated cluster of computers with GPU accelerators.
In the project DSPLab delivers knowledge from its fields to the consortium, namely: in the field of speech technologies, construction of Slovene language spoken resources, and artificial intelligence. As part of the “Research Infrastructure for Artificial Intelligence and International Competitiveness – RI-SI – CLARIN” project, DSPLab is responsible for NVIDIA DGX-1 supercomputer (HPC-CLARIN). The University of Maribor is a member of CLARIN.SI, the national hub of the European Research Infrastructure CLARIN. UM as a partner in the consortium is a source of knowledge in the field of speech technologies, the construction of Slovene language spoken resources, machine translation, and the creation and processing of large text databases. The high-performance CLARIN DGX-1 supercomputer is designed to research and develop machine learning methods in the field of artificial intelligence at the Digital Signal Processing Laboratory. The DGX-1 with supercomputing capability of 1000 TFLOPS, 40,960 CUDA cores, and 5,120 Tensor cores provides highly parallel machine learning performance and the shortest possible time to generate results. The project undoubtedly represents an additional impetus for work on artificial intelligence in both research and pedagogical terms, since it also opens up opportunities for students to find interesting projects, as well as graduate and master’s theses. Through the project, CLARIN.SI provided upgrades to existing hardware and hardware infrastructure for deep learning and AI artificial intelligence.
While many solutions offer GPU-accelerated performance, only the DGX-1 takes full advantage of the latest NVIDIA V100 GPUs, including innovations, such as: the next-generation NVIDIA NVLink ™ and the new Tensor Core architecture. With its deep learning software pool solutions, the DGX-1 delivers up to three times the learning speed of other GPU systems. With DGX-1 we have access to NVIDIA AI expertise, enterprise support, and extensive training.
PDF dokument.pdf
MIZŠ: SAIAL
Members in the project
UPS for DGX-1
First supercomputer DGX-1 in the DSPLab