Kyusu Ahn

Heehoon Kim

Jaehoon Jung

Wookeun Jung

Thanh Tuan Dao

Hongjune Kim

Gangwon Jo

Jungho Park

Junghyun Kim

Sangmin Seo

Jungwon Kim

Choonki Jang

Bernhard Egger

Seorin Kim

Sungeun Hahm

Chanwoo Park

Yujin Jung

Donghun Choi

Junyeol Ryu

Sangik Lee

Minjung Ko

Yonwoo Seong

Woojin Kim

Sangsoo Im

Gwangho Choi

Youngeun Choi

Soram Lee

Jongwon Kim

Hyungdal Kwon

Minhye Park

Minhee Han

Sooyeon Kang

Sunyoo Kim

Jaeki Hong

Hyunjae Lim

Jeesoo Lee

Dongjin Na

Jungwook Kim

Janghyun Son

Hyungmo Kim

PyeongSeok Oh

Youngdong Do

Jiyoung Park

Jaeho Shin

Bojun Seo

Jinyoung Joo

Seonmyeong Bak

Jeongho Nah

Jun Lee

Honggyu Kim

Seungkyun Kim

Joo Hwan Lee

Eunbyung Park

Jinho Pak

Posung Chun

Chihun Kim

Taejun Ha

Kwangseob Kim

Yoonsung Nam

Jongyoung Lee

Seokho Choi

Changhee Jung

Kiwon Kwon

Christophe Dubach

The THUNDER Research Group at Seoul National University is affiliated with the Department of Computer Science and Engineering and the Department of Data Science. The group aims to accelerate the training and inference of large-scale AI models while reducing their size. It focuses on Natural Language Processing using Deep Learning, emphasizing pragmatic approaches.

서울대학교 THUNDER 연구 그룹은 공과대학 컴퓨터공학과와 데이터사이언스 대학원 데이터사이언스학과에 소속되어 있다. 대규모 AI 모델의 훈련 및 추론 가속화와 모델 사이즈 감축을 목표로 하며, 딥러닝 관점에서 자연어 처리에 중점을 두고 실용적인 접근법을 강조한다.

SPipe: Hybrid GPU and CPU Pipeline for Training LLMs under Memory Pressure

Thunder-DeID: Accurate and Efficient De-identification Framework for Korean Court Judgments

Assessing Socio-Cultural Alignment and Technical Safety of Sovereign LLMs

KLAAD: Refining Attention Mechanisms to Reduce Societal Bias in Generative Language Models

UDC-VIT: A Real-World Video Dataset for Under-Display Cameras

Integrating spatial and frequency information for Under-Display Camera image restoration

SnuSOLVER: Optimizing Sparse Direct Solvers for Heterogeneous Systems

Thunder-LLM: Efficiently Adapting LLMs to Korean with Minimal Resources

Thunder-NUBench: A Benchmark for LLMs’ Sentence-Level Negation Understanding

Thunder-Tok: Minimizing Tokens per Word in Tokenizing Korean Texts for Generative Language Models

Enhancing Face Recognition Accuracy for Under-Display Cameras via Image Restoration

Investigating the Effects of Degradation and Restoration on Face Recognition in Under-Display Cameras

ANPMI: Assessing the True Comprehension Capabilities of LLMs for Multiple Choice Questions

FED: Fast and Efficient Dataset Deduplication Framework with GPU Acceleration

Generalizing Clinical De-identification Models by Privacy-safe Data Augmentation using GPT-4

TCCL: Discovering Better Communication Paths for PCIe GPU Clusters

UDC-SIT: A Real-World Dataset for Under-Display Cameras

DeepUM: Tensor Migration and Prefetching in Unified Memory

SnuQS: Scaling Quantum Circuit Simulation using Storage Sevices

SnuHPL: High Performance LINPACK for Heterogeneous GPUs

FARNN: FPGA-GPU Hybrid Acceleration Platform for Recurrent Neural Networks

Turnpike: Lightweight Soft Error Resilience for In-Order Cores

DeepCuts: a deep learning optimization framework for versatile GPU workloads

SnuRHAC: A Runtime for Heterogeneous Accelerator Clusters with CUDA Unified Memory

CyCNN: A Rotation Invariant CNN using Polar Mapping and Cylindrical Convolution Layers

Compiler-Directed Soft Error Resilience for Lightweight GPU Register File Protection

SOFF: An OpenCL High-Level Synthesis Framework for FPGAs

Overlapping Host-to-Device Copy and Computation using Hidden Unified Memory

SNU-NPB 2019: Parallelizing and Optimizing NPB in OpenCL and CUDA for Modern GPUs

FA3C: FPGA-Accelerated Deep Reinforcement Learning

Transparent GPU Memory Management for DNNs

An Auto-tuner for OpenCL Work-Group Size on GPUs

Memory-Access-Pattern Analysis Techniques for OpenCL Kernels

Performance Analysis of CNN Frameworks for GPUs

MAPA: An Automatic Memory Access Pattern Analyzer for GPU Applications

Translating OpenMP Device Constructs to OpenCL Using Unnecessary Data Transfer Elimination

PIPSEA: A Practical IPsec Gateway on Embedded APUs

Efficient Checkpointing of Live Virtual Machines

SnuCL: A unified OpenCL framework for heterogeneous clusters

A Distributed OpenCL Framework using Redundant Computation and Data Replication

Bridging OpenCL and CUDA: A Comparative Analysis and Translation

Accelerating LINPACK with MPI-OpenCL on Clusters of Multi-GPU Nodes

A Performance Model for GPUs with Caches

Versatile and Scalable Parallel Histogram Construction

Lightweight and Block-level Concurrent Sweeping for JavaScript Garbage Collection

OpenCL Framework for ARM Processors with NEON Support

Automatic OpenCL Work-Group Size Selection for Multicore CPUs

Automatic Code Overlay Generation and Partially Redundant Code Fetch Elimination

SnuCL: an OpenCL Framework for Heterogeneous CPU/GPU Clusters

Parallel Programming with Polaris

STeP: The Stanford Temporal Prover

Performance characterization of the NAS Parallel Benchmarks in OpenCL

SnuCL: an OpenCL framework for heterogeneous CPU/GPU clusters

Using a user-level memory thread for correlation prefetching

Achieving a single compute device image in OpenCL for multiple GPUs

Performance analysis of CNN frameworks for GPUs

Using prime numbers for cache indexing to eliminate conflict misses

Concurrent static single assignment form and constant propagation for explicitly parallel programs

THUNDER Research Group

교수님 소개

이재진

연구실 소개

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

SPipe: Hybrid GPU and CPU Pipeline for Training LLMs under Memory Pressure

Thunder-DeID: Accurate and Efficient De-identification Framework for Korean Court Judgments

Assessing Socio-Cultural Alignment and Technical Safety of Sovereign LLMs

연구실 보유 특허

대학별 입시끈

우리 연구실 모집

교수님 소개

이재진

연구실 키워드

알럼나이가 간 회사

연구실 지원 방법

대우 조건

연도별 논문 인용수

저널 & 학회 논문 출판 수

연구실 논문 목록

SPipe: Hybrid GPU and CPU Pipeline for Training LLMs under Memory Pressure

Thunder-DeID: Accurate and Efficient De-identification Framework for Korean Court Judgments

Assessing Socio-Cultural Alignment and Technical Safety of Sovereign LLMs

연구실 보유 특허

최근 3년간 연구실적 높은 연구실

대학별 입시끈

우리 연구실 모집