Deep learning workload scheduling in gpu datacenters: Taxonomy, challenges and vision

W Gao, Q Hu, Z Ye, P Sun, X Wang, Y Luo… - arXiv preprint arXiv …, 2022 - arxiv.org
Deep learning (DL) shows its prosperity in a wide variety of fields. The development of a DL
model is a time-consuming and resource-intensive procedure. Hence, dedicated GPU …

Retiarii: A deep learning {Exploratory-Training} framework

Q Zhang, Z Han, F Yang, Y Zhang, Z Liu… - … USENIX Symposium on …, 2020 - usenix.org
Traditional deep learning frameworks such as TensorFlow and PyTorch support training on
a single deep neural network (DNN) model, which involves computing the weights iteratively …

Nautilus: An optimized system for deep transfer learning over evolving training datasets

S Nakandala, A Kumar - … of the 2022 International Conference on …, 2022 - dl.acm.org
Deep learning (DL) has revolutionized unstructured data analytics. But in most cases, DL
needs massive labeled datasets and large compute clusters, which hinders its adoption …

Horizontally fused training array: An effective hardware utilization squeezer for training novel deep learning models

S Wang, P Yang, Y Zheng, X Li… - … of Machine Learning …, 2021 - proceedings.mlsys.org
Driven by the tremendous effort in researching novel deep learning (DL) algorithms, the
training cost of developing new models increases staggeringly in recent years. We analyze …

Saturn: An Optimized Data System for Large Model Deep Learning Workloads

K Nagrecha, A Kumar - arXiv preprint arXiv:2309.01226, 2023 - arxiv.org
Large language models such as GPT-3 & ChatGPT have transformed deep learning (DL),
powering applications that have captured the public's imagination. These models are rapidly …

Efficient supernet training using path parallelism

Y Xu, L Cheng, X Cai, X Ma, W Chen… - … Symposium on High …, 2023 - ieeexplore.ieee.org
Compared to conventional neural networks, training a supernet for Neural Architecture
Search (NAS) is very time consuming. Although current works have demonstrated that …

Declarative data serving: the future of machine learning inference on the edge

T Shaowang, N Jain, DD Matthews… - Proceedings of the VLDB …, 2021 - dl.acm.org
Recent advances in computer architecture and networking have ushered in a new age of
edge computing, where computation is placed close to the point of data collection to …

EdgeNN: Efficient Neural Network Inference for CPU-GPU Integrated Edge Devices

C Zhang, F Zhang, K Chen, M Chen… - 2023 IEEE 39th …, 2023 - ieeexplore.ieee.org
With the development of the architectures and the growth of AIoT application requirements,
data processing on edge has become popular. Neural network inference is widely employed …

Rotary: A Resource Arbitration Framework for Progressive Iterative Analytics

R Liu, AJ Elmore, MJ Franklin… - 2023 IEEE 39th …, 2023 - ieeexplore.ieee.org
Increasingly modern computing applications employ progressive iterative analytics, as best
exemplified by two prevalent cases, approximate query processing (AQP) and deep learning …

Riveter: Adaptive Query Suspension and Resumption Framework for Cloud Native Databases

R Liu, AJ Elmore, MJ Franklin… - 2024 IEEE 40th …, 2024 - ieeexplore.ieee.org
In modern cloud environments, ephemeral resources with intermittent availability and
fluctuating monetary costs are becoming common. This dynamic nature presents a new …