Picture for Hyoukjun Kwon

Hyoukjun Kwon

Characterizing the Accuracy - Efficiency Trade-off of Low-rank Decomposition in Language Models

May 10, 2024
Viaarxiv icon

SCAR: Scheduling Multi-Model AI Workloads on Heterogeneous Multi-Chiplet Module Accelerators

May 01, 2024
Viaarxiv icon

NonGEMM Bench: Understanding the Performance Horizon of the Latest ML Workloads with NonGEMM Workloads

Apr 17, 2024
Viaarxiv icon

Inter-Layer Scheduling Space Exploration for Multi-model Inference on Heterogeneous Chiplets

Dec 14, 2023
Viaarxiv icon

SDRM3: A Dynamic Scheduler for Dynamic Real-time Multi-model ML Workloads

Dec 07, 2022
Figure 1 for SDRM3: A Dynamic Scheduler for Dynamic Real-time Multi-model ML Workloads
Figure 2 for SDRM3: A Dynamic Scheduler for Dynamic Real-time Multi-model ML Workloads
Figure 3 for SDRM3: A Dynamic Scheduler for Dynamic Real-time Multi-model ML Workloads
Figure 4 for SDRM3: A Dynamic Scheduler for Dynamic Real-time Multi-model ML Workloads
Viaarxiv icon

XRBench: An Extended Reality (XR) Machine Learning Benchmark Suite for the Metaverse

Add code
Nov 16, 2022
Figure 1 for XRBench: An Extended Reality (XR) Machine Learning Benchmark Suite for the Metaverse
Figure 2 for XRBench: An Extended Reality (XR) Machine Learning Benchmark Suite for the Metaverse
Figure 3 for XRBench: An Extended Reality (XR) Machine Learning Benchmark Suite for the Metaverse
Figure 4 for XRBench: An Extended Reality (XR) Machine Learning Benchmark Suite for the Metaverse
Viaarxiv icon

Multi-Scale High-Resolution Vision Transformer for Semantic Segmentation

Nov 23, 2021
Figure 1 for Multi-Scale High-Resolution Vision Transformer for Semantic Segmentation
Figure 2 for Multi-Scale High-Resolution Vision Transformer for Semantic Segmentation
Figure 3 for Multi-Scale High-Resolution Vision Transformer for Semantic Segmentation
Figure 4 for Multi-Scale High-Resolution Vision Transformer for Semantic Segmentation
Viaarxiv icon

Evaluating Spatial Accelerator Architectures with Tiled Matrix-Matrix Multiplication

Jun 19, 2021
Figure 1 for Evaluating Spatial Accelerator Architectures with Tiled Matrix-Matrix Multiplication
Figure 2 for Evaluating Spatial Accelerator Architectures with Tiled Matrix-Matrix Multiplication
Figure 3 for Evaluating Spatial Accelerator Architectures with Tiled Matrix-Matrix Multiplication
Figure 4 for Evaluating Spatial Accelerator Architectures with Tiled Matrix-Matrix Multiplication
Viaarxiv icon

MARVEL: A Decoupled Model-driven Approach for Efficiently Mapping Convolutions on Spatial DNN Accelerators

Feb 18, 2020
Figure 1 for MARVEL: A Decoupled Model-driven Approach for Efficiently Mapping Convolutions on Spatial DNN Accelerators
Figure 2 for MARVEL: A Decoupled Model-driven Approach for Efficiently Mapping Convolutions on Spatial DNN Accelerators
Figure 3 for MARVEL: A Decoupled Model-driven Approach for Efficiently Mapping Convolutions on Spatial DNN Accelerators
Figure 4 for MARVEL: A Decoupled Model-driven Approach for Efficiently Mapping Convolutions on Spatial DNN Accelerators
Viaarxiv icon

Co-Exploration of Neural Architectures and Heterogeneous ASIC Accelerator Designs Targeting Multiple Tasks

Feb 10, 2020
Figure 1 for Co-Exploration of Neural Architectures and Heterogeneous ASIC Accelerator Designs Targeting Multiple Tasks
Figure 2 for Co-Exploration of Neural Architectures and Heterogeneous ASIC Accelerator Designs Targeting Multiple Tasks
Figure 3 for Co-Exploration of Neural Architectures and Heterogeneous ASIC Accelerator Designs Targeting Multiple Tasks
Figure 4 for Co-Exploration of Neural Architectures and Heterogeneous ASIC Accelerator Designs Targeting Multiple Tasks
Viaarxiv icon