We address the challenging problem of efficient inference across many devices and resource constraints, especially on edge devices. Conventional approaches either manually design or use neural architecture search (NAS) to find a specialized neural network and train it from scratch for each case, which is computationally prohibitive (causing $CO_2$ emission as much as 5 cars' lifetime) thus unscalable. In this work, we propose to train a once-for-all (OFA) network that supports diverse architectural settings by decoupling training and search, to reduce the cost. We can quickly get a specialized sub-network by selecting from the OFA network without additional training. To efficiently train OFA networks, we also propose a novel progressive shrinking algorithm, a generalized pruning method that reduces the model size across many more dimensions than pruning (depth, width, kernel size, and resolution). It can obtain a surprisingly large number of sub-networks ($> 10^{19}$) that can fit different hardware platforms and latency constraints while maintaining the same level of accuracy as training independently. On diverse edge devices, OFA consistently outperforms state-of-the-art (SOTA) NAS methods (up to 4.0% ImageNet top1 accuracy improvement over MobileNetV3, or same accuracy but 1.5x faster than MobileNetV3, 2.6x faster than EfficientNet w.r.t measured latency) while reducing many orders of magnitude GPU hours and $CO_2$ emission. In particular, OFA achieves a new SOTA 80.0% ImageNet top-1 accuracy under the mobile setting ($<$600M MACs). OFA is the winning solution for the 3rd Low Power Computer Vision Challenge (LPCVC), DSP classification track and the 4th LPCVC, both classification track and detection track. Code and 50 pre-trained models (for many devices & many latency constraints) are released at https://github.com/mit-han-lab/once-for-all.
Monash University（莫纳什大学）；South China University of Technology（华南理工大学）
In many real-world applications, we often need to handle various deployment scenarios, where the resource constraint and the superclass of interest corresponding to a group of classes are dynamically specified. How to efficiently deploy deep models for diverse deployment scenarios is a new challenge. Previous NAS approaches seek to design architectures for all classes simultaneously, which may not be optimal for some individual superclasses. A straightforward solution is to search an architecture from scratch for each deployment scenario, which however is computation-intensive and impractical.…
We achieve very efficient deep learning model deployment that designs neural network architectures to fit different hardware constraints. Given a constraint, most neural architecture search (NAS) methods either sample a set of sub-networks according to a pre-trained accuracy predictor, or adopt the evolutionary algorithm to evolve specialized networks from the supernet. Both approaches are time consuming. Here our key idea for very efficient deployment is, when searching the architecture space, constructing a table that stores the validation accuracy of all candidate blocks at all layers. For …
Neural Architecture Search (NAS) has enabled the possibility of automated machine learning by streamlining the manual development of deep neural network architectures defining a search space, search strategy, and performance estimation strategy. To solve the need for multi-platform deployment of Convolutional Neural Network (CNN) models, Once-For-All (OFA) proposed to decouple Training and Search to deliver a one-shot model of sub-networks that are constrained to various accuracy-latency tradeoffs. We find that the performance estimation strategy for OFA's search severely lacks generalizabilit…
Once-for-All (OFA) is a Neural Architecture Search (NAS) framework designed to address the problem of searching efficient architectures for devices with different resources constraints by decoupling the training and the searching stages. The computationally expensive process of training the OFA neural network is done only once, and then it is possible to perform multiple searches for subnetworks extracted from this trained network according to each deployment scenario. In this work we aim to give one step further in the search for efficiency by explicitly conceiving the search stage as a multi…
When approaching a novel visual recognition problem in a specialized image domain, a common strategy is to start with a pre-trained deep neural network and fine-tune it to the specialized domain. If the target domain covers a smaller visual space than the source domain used for pre-training (e.g. ImageNet), the fine-tuned network is likely to be over-parameterized. However, applying network pruning as a post-processing step to reduce the memory requirements has drawbacks: fine-tuning and pruning are performed independently; pruning parameters are set once and cannot adapt over time; and the hi…
The emergence of CNNs in mainstream deployment has necessitated methods to design and train efficient architectures tailored to maximize the accuracy under diverse hardware & latency constraints. To scale these resource-intensive tasks with an increasing number of deployment targets, Once-For-All (OFA) proposed an approach to jointly train several models at once with a constant training cost. However, this cost remains as high as 40-50 GPU days and also suffers from a combinatorial explosion of sub-optimal model configurations. We seek to reduce this search space -- and hence the training budg…
University of Wisconsin（威斯康星大学）；University of Illinois（伊利诺伊大学）；American Family Insurance（美国家庭保险公司）
Transformer-based models are widely used in natural language processing (NLP). Central to the transformer model is the self-attention mechanism, which captures the interactions of token pairs in the input sequences and depends quadratically on the sequence length. Training such models on longer sequences is expensive. In this paper, we show that a Bernoulli sampling attention mechanism based on Locality Sensitive Hashing (LSH), decreases the quadratic complexity of such models to linear. We bypass the quadratic cost by considering self-attention as a sum of individual tokens associated with Be…
DataS^3: Dataset Subset Selection for Specialization
MIT（麻省理工学院）；University of Haifa（海法大学）；Woods Hole Oceanographic Institution（伍兹霍尔海洋研究所）；UC Berkeley（加州大学伯克利分校）；Princeton University（普林斯顿大学）；Stanford University（斯坦福大学）；McGill University（麦吉尔大学）；Mila - Quebec AI Institute（魁北克AI研究所）；University of Toronto（多伦多大学）；Arizona State University（亚利桑那州立大学）；UCSF（旧金山大学医院）
MatchNAS: Optimizing Edge AI in Sparse-Label Data Contexts via Automating Deep Neural Network Porting for Mobile Deployment
School of Computer Science and Engineering, University of New South Wales（新南威尔士大学计算机科学与工程学院）；Faculty of Engineering & Information Technology, University of Technology Sydney（悉尼科技大学工程与信息技术学院）；CSIRO’s Data61（澳大利亚联邦科学与工业研究组织Data61）；University of New South Wales（新南威尔士大学）
From Bits to Chips: An LLM-based Hardware-Aware Quantization Agent for Streamlined Deployment of LLMs
The University of Arizona（亚利桑那大学）；Clemson University（克莱姆森大学）；University of Georgia（佐治亚大学）；The University of Tulsa（塔尔萨大学）；Northeastern University（东北大学）；Western Digital Corporation（西部数据公司）
