Picture for Ranggi Hwang

Ranggi Hwang

LazyDP: Co-Designing Algorithm-Software for Scalable Training of Differentially Private Recommendation Models

Apr 12, 2024
Figure 1 for LazyDP: Co-Designing Algorithm-Software for Scalable Training of Differentially Private Recommendation Models
Figure 2 for LazyDP: Co-Designing Algorithm-Software for Scalable Training of Differentially Private Recommendation Models
Figure 3 for LazyDP: Co-Designing Algorithm-Software for Scalable Training of Differentially Private Recommendation Models
Figure 4 for LazyDP: Co-Designing Algorithm-Software for Scalable Training of Differentially Private Recommendation Models
Viaarxiv icon

Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference

Aug 23, 2023
Figure 1 for Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference
Figure 2 for Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference
Figure 3 for Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference
Figure 4 for Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference
Viaarxiv icon

DiVa: An Accelerator for Differentially Private Machine Learning

Aug 26, 2022
Figure 1 for DiVa: An Accelerator for Differentially Private Machine Learning
Figure 2 for DiVa: An Accelerator for Differentially Private Machine Learning
Figure 3 for DiVa: An Accelerator for Differentially Private Machine Learning
Figure 4 for DiVa: An Accelerator for Differentially Private Machine Learning
Viaarxiv icon

GROW: A Row-Stationary Sparse-Dense GEMM Accelerator for Memory-Efficient Graph Convolutional Neural Networks

Mar 02, 2022
Figure 1 for GROW: A Row-Stationary Sparse-Dense GEMM Accelerator for Memory-Efficient Graph Convolutional Neural Networks
Figure 2 for GROW: A Row-Stationary Sparse-Dense GEMM Accelerator for Memory-Efficient Graph Convolutional Neural Networks
Figure 3 for GROW: A Row-Stationary Sparse-Dense GEMM Accelerator for Memory-Efficient Graph Convolutional Neural Networks
Figure 4 for GROW: A Row-Stationary Sparse-Dense GEMM Accelerator for Memory-Efficient Graph Convolutional Neural Networks
Viaarxiv icon

Centaur: A Chiplet-based, Hybrid Sparse-Dense Accelerator for Personalized Recommendations

May 12, 2020
Figure 1 for Centaur: A Chiplet-based, Hybrid Sparse-Dense Accelerator for Personalized Recommendations
Figure 2 for Centaur: A Chiplet-based, Hybrid Sparse-Dense Accelerator for Personalized Recommendations
Figure 3 for Centaur: A Chiplet-based, Hybrid Sparse-Dense Accelerator for Personalized Recommendations
Figure 4 for Centaur: A Chiplet-based, Hybrid Sparse-Dense Accelerator for Personalized Recommendations
Viaarxiv icon