May 2024 - nlplab - mailman.cs.vt.edu

meeting schedule next week
by Huang, Lifu 27 May '24

27 May '24

I will try to schedule at least one meeting with everyone until the EMNLP deadline. All the time are based on EST. Monday: 9-9:30AM: Trevor 9:30-10AM: Mohammad 10-10:30AM: Ying 10:30-11AM: Sijia 11-11:30AM: Zoe, Sijia Tuesday: 9-9:30PM: Barry 9:30-10:15PM: Jingyuan, Zhiyang 10:15-10:45PM: Zhiyang, Minqian 10:45-11:15PM: Minqian, Zhiyang 11:15-11:45PM: Zihao Wednesday: 10-10:30PM: Jeet 10:30-11PM: Tong, Zhiyang --------------------------------- Lifu Huang Assistant Professor Department of Computer Science Virginia Tech

1 0

Fwd: ARC maintenance + announcements
by Huang, Lifu 13 May '24

13 May '24

--------------------------------- Lifu Huang Assistant Professor Department of Computer Science Virginia Tech ---------- Forwarded message ---------- From: arc.users <arc.users(a)groups.vt.edu> Date: May 13, 2024 at 9:51 PM +0800 To: arc.users <arc.users(a)groups.vt.edu> Subject: ARC maintenance + announcements ARC Announcements: Systems Maintenance May 20-24: ARC systems will be offline for maintenance<https://docs.arc.vt.edu/resources/maintenance-cycle.html#tentative-maintena…> Monday – Friday, May 20-24, 2024. The ARC cluster, storage, and associated services will receive routine updates, checks, and maintenance. There will be some reorganization of underlying infrastructure improvements and an update to the cuda-drivers on GPU nodes which will enable support for recent CUDA versions. If you submit a job and find that it is queued with a message like "(ReqNodeNotAvail, Reserved for maintenance)", this means that the job's time limit extends into the maintenance period. Since the job cannot be started, it will be held in queue until the cluster is released from maintenance. Jobs whose time limits allow them to complete prior to 8:00am on Monday will still run, subject to resource availability. Please terminate any running processes on login nodes and disconnect any connections prior to the start of maintenance at 8:00am on Monday, May 20. ARC Start-of-Summer Workshop Series: We are hosting a sequence of workshops in early June. The material in these online-only courses is intended to provide an overview of ARC resources and introduce tools and strategies for getting started and beginning to optimize workflows. Register through the TLOS Professional Development Network (PDN)<https://profdev.tlos.vt.edu/?query=arc> 2024-06-03 10:00am - 12:00pm Introduction to Advanced Research Computing. 2024-06-03 2:00pm - 4:00pm Connect to ARC Systems and Run your first jobs 2024-06-04 10:00am - 12:00pm Running code/software on ARC systems in different ways 2024-06-04 12:00am - 2:00pm Launching Jobs in Parallel on ARC clusters 2024-06-05 10:00am - 12:00pm Monitoring Resource Utilization and Job Efficiency ARC systems status and roadmaps: Tinkercliffs overhaul: The Tinkercliffs cluster was first released in 2020 as ARC flagship CPU-based HPC cluster and dense GPU nodes were added in 2021 and 2022 to extend its capabilities. To keep it available as a computational resource, the current operating system on the CPU nodes requires a major upgrade which we plan to undertake near the end of 2024 calendar year. This will be disruptive and necessitate the reinstallation of all software packages on the CPU nodes but will keep the system in compliance with VT standards and also relieve some software incompatibility issues which have been caused by the outdate OS and the old glibc on those nodes. More details will follow as we finalize the plans and timeline. Owl (coming soon): A new, water-cooled AMD Zen4 Genoa CPU cluster with favorable memory architecture. 84 base nodes with 96 cores and 768GB DDR5 memory + 3 huge-memory nodes with 4TB or 8TB system memory on each node. Falcon (later in 2024): A general purpose GPU cluster for HPC and AI, designed as a replacement for the Infer cluster. 32 nodes with 4xA30 GPUs and 20 nodes with 4xL40S GPUs provides exceptional performance for mid-range AI and HPC workloads. Infer (nearing end of life): The bulk of the nodes in the current Infer cluster were migrated from earlier clusters. They have served us well, but the nodes and other necessary components are aging out. So, after the release of the Falcon cluster, we will begin to shut down Infer. New faces in ARC research collaboration and support team: If you have visited ARC's daily office hours (https://arc.vt.edu/office-hours) or asked for help with using ARC systems via the online ticketing system (https://4help.vt.edu<https://4help.vt.edu/>), then you're probably acquainted with our talented and dedicated team of ARC Helpdesk GRAs. Behind them, ARC has a team of full-time personnel who represent the core of ARCs research collaboration and support efforts. They all have deep and varied experience in computational aspects of research, but four are relatively new in these roles at VT: Ayat Mohammad<https://arc.vt.edu/about/our-team/ayat-mohammed.html> Scalable scientific visualization, Virtual/Augmented reality, Geoscience data analysis, Medical and healthcare data analysis Chris Kuhlman<https://arc.vt.edu/about/our-team/chris-kuhlman.html> Distributed and high performance computing, Social and cognitive modeling and simulation, Network science Sarah Ghazanfari<https://arc.vt.edu/about/our-team/sarah-ghazanfari.html> Molecular dynamics, Materials modeling, Density functional theory, Cheminformatics, and the Finite element method Sofia Lima<http://slima2@vt.edu/> Biology, Computer science, Machine learning Please welcome them if your paths cross and reach out to them if you would like to learn more about their interests or if you think they might be able to help support your research projects! How to get help with using ARC systems: Office Hours: https://arc.vt.edu/office-hours Submit help request via 4Help https://4help.vt.edu<https://4help.vt.edu/> Documentation site: https://docs.arc.vt.edu<https://docs.arc.vt.edu/> Best regards, Advanced Research Computing Virginia Tech

1 0

next week meeting schedule
by Huang, Lifu 05 May '24

05 May '24

Monday: 9:30-10AM: Zhiyang, Minqian 10-10:30AM: Minqian, Zhiyang 10:30-11AM: Jingyuan, Zhiyang 11-11:30AM: Zihao, Jingyuan, Zhiyang 11:30-12PM: Ying 2:30-3:15PM: Sijia, Tong, meeting with Mayfair 3:15-4:15PM: Zhiyang, Ying, meet with Yue 4:30-5PM: Sijia 5-5:30PM: Tong, Sijia 5:30-6PM: Zoe, Sijia Tuesday: 9:30-10AM: Barry 10-10:45AM: Mohammad, Ying, Jingyuan, Zhiyang, Minqian 11-11:30AM: Zihao, meet with Lingwei 3-4PM: Trevor, Jingyuan, Minqian, meet with Amazon 4-5PM: Zhiyang, Minqian, Zihao, Ying, meet with Intuit Wednesday: 10-11AM: Zhiyang, Jingyuan, meet with Yiming 3-4PM: Jingyuan, Barry, meet with STR for LLM meeting 5:30-6PM: Jeet Thursday: 2-2:45PM: Barry, Jingyuan, meet with STR 10-11PM: Zhiyang, Minqian, meet with Youran Friday: 10-11AM: Zihao, WaPost meeting 12-1PM: Minqian, Jingyuan, meet for DARPA FoundSci 9-10pm: Zhiyang, meet with Haibo --------------------------------- Lifu Huang Assistant Professor Department of Computer Science Virginia Tech

1 0

[nlp_reading_group] Reading Discussion on May 3
by Yao, Barry 03 May '24

03 May '24

Hello everyone, In this week's reading discussion, we will talk about: LLM2LLM: Boosting LLMs with Novel Iterative Data Enhancement<https://arxiv.org/abs/2403.15042> Abstract: While many real-world applications still require fine-tuning to reach satisfactory levels of performance, many of them are in the low-data regime, making fine-tuning challenging. To address this, the authors propose LLM2LLM, a targeted and iterative data augmentation strategy that uses a teacher LLM to enhance a small seed dataset by augmenting additional data that can be used for fine-tuning on a specific task. You could attend the meeting in the Data & Decision Building, Room 366, or on Zoom (https://virginiatech.zoom.us/j/4506040955?pwd=TGFtTUVWR2tuYmVlS3JraTQzbjgrU…) at 4-5 pm on Friday. Note that the Zoom link is different from the previous reading discussions. Happy reading! Best regards, Barry Menglong Yao Ph.D. student in Computer Science Virginia Tech Homepage<https://barry-yao.netlify.app/>

1 1