Alert button
Picture for Arvind Krishnamurthy

Arvind Krishnamurthy

Alert button

University of Washington

ForestColl: Efficient Collective Communications on Heterogeneous Network Fabrics

Feb 09, 2024
Liangyu Zhao, Saeed Maleki, Ziyue Yang, Hossein Pourreza, Aashaka Shah, Changho Hwang, Arvind Krishnamurthy

Viaarxiv icon

Atom: Low-bit Quantization for Efficient and Accurate LLM Serving

Nov 07, 2023
Yilong Zhao, Chien-Yu Lin, Kan Zhu, Zihao Ye, Lequn Chen, Size Zheng, Luis Ceze, Arvind Krishnamurthy, Tianqi Chen, Baris Kasikci

Figure 1 for Atom: Low-bit Quantization for Efficient and Accurate LLM Serving
Figure 2 for Atom: Low-bit Quantization for Efficient and Accurate LLM Serving
Figure 3 for Atom: Low-bit Quantization for Efficient and Accurate LLM Serving
Figure 4 for Atom: Low-bit Quantization for Efficient and Accurate LLM Serving
Viaarxiv icon

Punica: Multi-Tenant LoRA Serving

Oct 28, 2023
Lequn Chen, Zihao Ye, Yongji Wu, Danyang Zhuo, Luis Ceze, Arvind Krishnamurthy

Figure 1 for Punica: Multi-Tenant LoRA Serving
Figure 2 for Punica: Multi-Tenant LoRA Serving
Figure 3 for Punica: Multi-Tenant LoRA Serving
Figure 4 for Punica: Multi-Tenant LoRA Serving
Viaarxiv icon

Symphony: Optimized Model Serving using Centralized Orchestration

Aug 14, 2023
Lequn Chen, Weixin Deng, Anirudh Canumalla, Yu Xin, Matthai Philipose, Arvind Krishnamurthy

Figure 1 for Symphony: Optimized Model Serving using Centralized Orchestration
Figure 2 for Symphony: Optimized Model Serving using Centralized Orchestration
Figure 3 for Symphony: Optimized Model Serving using Centralized Orchestration
Figure 4 for Symphony: Optimized Model Serving using Centralized Orchestration
Viaarxiv icon

Bandwidth Optimal Pipeline Schedule for Collective Communication

May 31, 2023
Liangyu Zhao, Arvind Krishnamurthy

Figure 1 for Bandwidth Optimal Pipeline Schedule for Collective Communication
Figure 2 for Bandwidth Optimal Pipeline Schedule for Collective Communication
Viaarxiv icon

Optimal Direct-Connect Topologies for Collective Communications

Feb 07, 2022
Liangyu Zhao, Siddharth Pal, Tapan Chugh, Weiyang Wang, Prithwish Basu, Joud Khoury, Arvind Krishnamurthy

Figure 1 for Optimal Direct-Connect Topologies for Collective Communications
Figure 2 for Optimal Direct-Connect Topologies for Collective Communications
Figure 3 for Optimal Direct-Connect Topologies for Collective Communications
Figure 4 for Optimal Direct-Connect Topologies for Collective Communications
Viaarxiv icon

Cloud Collectives: Towards Cloud-aware Collectives forML Workloads with Rank Reordering

May 28, 2021
Liang Luo, Jacob Nelson, Arvind Krishnamurthy, Luis Ceze

Figure 1 for Cloud Collectives: Towards Cloud-aware Collectives forML Workloads with Rank Reordering
Figure 2 for Cloud Collectives: Towards Cloud-aware Collectives forML Workloads with Rank Reordering
Figure 3 for Cloud Collectives: Towards Cloud-aware Collectives forML Workloads with Rank Reordering
Figure 4 for Cloud Collectives: Towards Cloud-aware Collectives forML Workloads with Rank Reordering
Viaarxiv icon

AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly

May 22, 2021
Yuchen Jin, Tianyi Zhou, Liangyu Zhao, Yibo Zhu, Chuanxiong Guo, Marco Canini, Arvind Krishnamurthy

Figure 1 for AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly
Figure 2 for AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly
Figure 3 for AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly
Figure 4 for AutoLRS: Automatic Learning-Rate Schedule by Bayesian Optimization on the Fly
Viaarxiv icon

Scaling Distributed Machine Learning with In-Network Aggregation

Feb 22, 2019
Amedeo Sapio, Marco Canini, Chen-Yu Ho, Jacob Nelson, Panos Kalnis, Changhoon Kim, Arvind Krishnamurthy, Masoud Moshref, Dan R. K. Ports, Peter Richtárik

Figure 1 for Scaling Distributed Machine Learning with In-Network Aggregation
Figure 2 for Scaling Distributed Machine Learning with In-Network Aggregation
Figure 3 for Scaling Distributed Machine Learning with In-Network Aggregation
Figure 4 for Scaling Distributed Machine Learning with In-Network Aggregation
Viaarxiv icon