paper-with-me

Papers

Accelerating Data Loading in Deep Neural Network Training

2019-10-02 · Chih-Chieh Yang, Guojing Cong

Data loading can dominate deep neural network training time on large-scale systems. We present a comprehensive study on accelerating data loading performance in large-scale distributed training. We first identify performance and scalability issues in current data loading implementations. We then propose optimizations that utilize CPU resources to the data loader design. We use an analytical model to characterize the impact of data loading on the overall training time and establish the performance trend as we scale up distributed training. Our model suggests that I/O rate limits the scalability of distributed training, which inspires us to design a locality-aware data loading method. By utilizing software caches, our method can drastically reduce the data loading communication volume in comparison with the original data loading implementation. Finally, we evaluate the proposed optimizations with various experiments. We achieved more than 30x speedup in data loading using 256 nodes with 1,024 learners.

📄 PDF Abstract BibTeX arXiv:1910.01196

Code (0)

등록된 구현이 없습니다.

Tasks

CPU

Similar Papers 제목 키워드 기반

FFCV: Accelerating Training by Removing Data Bottlenecks

2023-06-21 · CVPR 2023 1 · Guillaume Leclerc, Andrew Ilyas, Logan Engstrom, Sung Min Park 외

We present FFCV, a library for easy and fast machine learning model training. FFCV speeds up model training by eliminating (often subtle) data bottlenecks from the training process. In particular, we combine techniques s…

CPUGPU

Progressive Weight Loading: Accelerating Initial Inference and Gradually Boosting Performance on Resource-Constrained Environments

2025-09-26 · Hyunwoo Kim, Junha Lee, Mincheol Choi, Jeonghwan Lee 외 arxiv

Deep learning models have become increasingly large and complex, resulting in higher memory consumption and computational demands. Consequently, model loading times and initial inference latency have increased, posing si…

Knowledge Distillation

DynaComm: Accelerating Distributed CNN Training between Edges and Clouds through Dynamic Communication Scheduling

2021-01-20 · Shangming Cai, Dongsheng Wang, Haixia Wang, Yongqiang Lyu 외

To reduce uploading bandwidth and address privacy concerns, deep learning at the network edge has been an emerging topic. Typically, edge devices collaboratively train a shared model using real-time generated data throug…

Scheduling

Tangram: Accelerating Serverless LLM Loading through GPU Memory Reuse and Affinity

2025-12-01 · Wenbin Zhu, Zhaoyan Shen, Zili Shao, Hongjun Dai 외 arxiv

Serverless Large Language Models (LLMs) have emerged as a cost-effective solution for deploying AI services by enabling a 'pay-as-you-go' pricing model through GPU resource sharing. However, cold-start latency, especiall…

Accelerating Edge Intelligence via Integrated Sensing and Communication

2021-07-20 · Tong Zhang, Shuai Wang, Guoliang Li, Fan Liu 외

Realizing edge intelligence consists of sensing, communication, training, and inference stages. Conventionally, the sensing and communication stages are executed sequentially, which results in excessive amount of dataset…

Dataset GenerationIntegrated sensing and communicationISAC