Picture for Mark Zhao

Mark Zhao

SlipStream: Adapting Pipelines for Distributed Training of Large DNNs Amid Failures

May 22, 2024
Viaarxiv icon

cedar: Composable and Optimized Machine Learning Input Data Pipelines

Jan 25, 2024
Viaarxiv icon

RecD: Deduplication for End-to-End Deep Learning Recommendation Model Training Infrastructure

Nov 14, 2022
Figure 1 for RecD: Deduplication for End-to-End Deep Learning Recommendation Model Training Infrastructure
Figure 2 for RecD: Deduplication for End-to-End Deep Learning Recommendation Model Training Infrastructure
Figure 3 for RecD: Deduplication for End-to-End Deep Learning Recommendation Model Training Infrastructure
Figure 4 for RecD: Deduplication for End-to-End Deep Learning Recommendation Model Training Infrastructure
Viaarxiv icon

Understanding and Co-designing the Data Ingestion Pipeline for Industry-Scale RecSys Training

Aug 20, 2021
Figure 1 for Understanding and Co-designing the Data Ingestion Pipeline for Industry-Scale RecSys Training
Figure 2 for Understanding and Co-designing the Data Ingestion Pipeline for Industry-Scale RecSys Training
Figure 3 for Understanding and Co-designing the Data Ingestion Pipeline for Industry-Scale RecSys Training
Figure 4 for Understanding and Co-designing the Data Ingestion Pipeline for Industry-Scale RecSys Training
Viaarxiv icon