Picture for Jaeyong Song

Jaeyong Song

Pipette: Automatic Fine-grained Large Language Model Training Configurator for Real-World Clusters

Add code
May 28, 2024
Viaarxiv icon

Smart-Infinity: Fast Large Language Model Training using Near-Storage Processing on a Real System

Add code
Mar 11, 2024
Viaarxiv icon

PeerAiD: Improving Adversarial Distillation from a Specialized Peer Tutor

Add code
Mar 11, 2024
Viaarxiv icon

GraNNDis: Efficient Unified Distributed Training Framework for Deep GNNs on Large Clusters

Add code
Nov 12, 2023
Viaarxiv icon

Pipe-BD: Pipelined Parallel Blockwise Distillation

Add code
Jan 29, 2023
Viaarxiv icon

SGCN: Exploiting Compressed-Sparse Features in Deep Graph Convolutional Network Accelerators

Add code
Jan 25, 2023
Figure 1 for SGCN: Exploiting Compressed-Sparse Features in Deep Graph Convolutional Network Accelerators
Figure 2 for SGCN: Exploiting Compressed-Sparse Features in Deep Graph Convolutional Network Accelerators
Figure 3 for SGCN: Exploiting Compressed-Sparse Features in Deep Graph Convolutional Network Accelerators
Figure 4 for SGCN: Exploiting Compressed-Sparse Features in Deep Graph Convolutional Network Accelerators
Viaarxiv icon

Optimus-CC: Efficient Large NLP Model Training with 3D Parallelism Aware Communication Compression

Add code
Jan 24, 2023
Viaarxiv icon

Slice-and-Forge: Making Better Use of Caches for Graph Convolutional Network Accelerators

Add code
Jan 24, 2023
Figure 1 for Slice-and-Forge: Making Better Use of Caches for Graph Convolutional Network Accelerators
Figure 2 for Slice-and-Forge: Making Better Use of Caches for Graph Convolutional Network Accelerators
Figure 3 for Slice-and-Forge: Making Better Use of Caches for Graph Convolutional Network Accelerators
Figure 4 for Slice-and-Forge: Making Better Use of Caches for Graph Convolutional Network Accelerators
Viaarxiv icon