Introduction
Researchers conducting scientific work using supercomputers and large-scale storage resources have diverse requirements and working methods, and therefore need various types of access to computers and other IT infrastructure resources. In response to users' needs, we provide a range of advanced IT platforms and dedicated services. Together with the computing infrastructure, we offer a set of tools that enables researchers to carry out large-scale complex experiments and to manage their results efficiently and easily.
All tools, services, and platforms are provided within the PLGrid infrastructure (https://www.plgrid.pl), allowing Polish scientists and their international collaborators to access resources and data in the most convenient way for them.
LLM Lab platform
The LLM Lab Platform is available to PLGrid infrastructure users as a dedicated service, enabling the use of large language models tailored to various needs such as data analysis, modeling, and automation of research processes. Users can apply for access through computational grants via the PLGrid Portal.
The platform offers access to various language models. One of the key models available on the platform is Bielik, optimized for Polish morphology and syntax, making it particularly useful for research related to the Polish language and local data.
The LLM Lab Platform is accessible from anywhere in the world, allowing researchers to use its resources regardless of their location. Additionally, the API interface complies with the OpenAI API standard, facilitating integration with existing applications and research tools.
Access to LLM Lab: https://portal.plgrid.pl
JupyterHub
JupyterHub is a platform that enables launching JupyterLab servers as computational tasks on Cyfronet clusters.
Users gain access to JupyterLab instances via a web browser. This solution significantly simplifies working with Jupyter notebooks - an interactive coding method especially popular among users involved in data analysis and artificial intelligence methods. The platform also allows direct execution of computations within an active task. Additionally, through JupyterLab servers, it is possible to launch a remote instance of the Visual Studio Code editor running on the cluster, with a browser-based interface for remote access.
Access to JupyterHub at Cyfronet and documentation: https://portal.plgrid.pl
Pro-viz tool
The pro-viz tool enables users of computing clusters to run applications in graphical mode (GUI) based on the XFCE environment directly on the cluster’s compute nodes, with optional GPU acceleration. It operates using TurboVNC - one of the fastest VNC clients - and allows comfortable use of scientific software even over low-bandwidth networks. Moreover, the tool is integrated with a convenient web-based access service, so cluster resources in graphical mode can be used without installing a dedicated VNC client, simply through an HTML5-compatible web browser.
More information about the service: https://guide.plgrid.pl/computing/pro_viz/
OneData
Onedata is the flagship product of the Cloud Technologies Laboratory. It is a globally scalable data management system that unifies access to data stored across distributed systems. Onedata effectively meets the needs of both small user groups and large international scientific communities. The system allows users to utilize a unified data management framework for storing both personal data and work-related content, such as research results, and provides efficient access from any device.
More information available on the Cloud Technologies Laboratory page
MEE
The Model Execution Environment (MEE) is a technology stack that enables the execution of task streams on high-performance computing infrastructures, including resources provided by ACK Cyfronet AGH. The platform is designed to facilitate easy computation for domain scientists—individuals with basic knowledge of interacting with computing clusters and other types of HPC resources.
MEE is accessible through a user-friendly web-based UI. The platform also provides programming interfaces that allow integration with external tools.
The MEE platform is integrated with the authentication and security mechanisms of the PLGrid infrastructure. Users with PLGrid accounts can access MEE through them and submit computations within their assigned quotas. Each computation is executed under a specific research grant, which can be configured within the platform.
The Model Execution Environment has been used in a number of research projects, including the European projects EurValve, PRIMAGE, and InSilicoWorld, the POLVAS consortium, and the Sano Centre for Computational Medicine. MEE has also been used as a prototype implementation of a simulation platform for a virtual twin developed as part of the European roadmap prepared by the EDITH project.
More information: https://mee.cyfronet.pl
Contact
For assistance or if you would like to cooperate, please contact us via the Helpdesk system.