An official website of the United States government
Deep neural networks are increasingly required to operate across diverse hardware platforms, latency constraints, and power budgets, which motivates the need for specialized models for each scenario. However, designing and training a separate model per scenario or serving a large ensemble of models is often impractical. Weight sharing has emerged as a promising paradigm to address this challenge by training a single ''SuperNet'' that subsumes many sub-models (SubNets), and by reusing weights across those SubNets both at training and inference time. This paper provides an abridged survey of our recent advances that leverage weight sharing for efficient AI, covering both training and inference serving. In centralized once-for-all training, Delayed ε-Shrinking (DεS) improves training efficiency by strategically scheduling the introduction of smaller SubNets during training. In a federated fashion, SuperFedNas co-trains a SuperNet across distributed clients and disjoins training and searching, which enables oneshot specialization to many deployment targets at minimal cost. ∇QDARTS integrates quantization into differentiable architecture search, jointly finding neural architectures, weights, and low-precision settings to yield highly efficient models in a single search. For inference serving, SuperServe introduces a weight-shared model with dynamic SubNet routing (SubNetAct) to instantaneously switch among a spectrum of accuracy-latency operating points, coupled with a scheduler (SlackFit) for unpredictable workloads. Finally, SUSHI co-designs model, system, and accelerator to exploit weightshared SuperNets on tinyML devices, caching SubGraphs on FPGA to reduce latency and energy. Together, these works demonstrate that the weight sharing paradigm can dramatically improve the efficiency of both training and inference serving of deep models across a range of scenarios.
more »
« less
DepS: Delayed epsilon-Shrinking for Faster Once-for-All Training
CNNs are increasingly deployed across different hardware, dynamic environments, and low-power embedded devices. This has led to the design and training of CNN architectures with the goal of maximizing accuracy subject to such variable deployment constraints. As the number of deployment scenarios grows, there is a need to find scalable solutions to design and train specialized CNNs. Once-for-all training has emerged as a scalable approach that jointly co-trains many models (subnets) at once with a constant training cost and finds specialized CNNs later. The scalability is achieved by training the full model and simultaneously reducing it to smaller subnets that share model weights (weight-shared shrinking). However, existing once-for-all training approaches incur huge training costs reaching 1200 GPU hours. We argue this is because they either start the process of shrinking the full model too early or too late. Hence, we propose Delayed Epsilon-Shrinking (DepS) that starts the process of shrinking the full model when it is partially trained, which leads to training cost improvement and better in-place knowledge distillation to smaller models. The proposed approach also consists of novel heuristics that dynamically adjust subnet learning rates incrementally, leading to improved weight-shared knowledge distillation from larger to smaller subnets as well. As a result, DepS outperforms state-of-the-art once-for-all training techniques across different datasets including CIFAR10/100, ImageNet-100, and ImageNet-1k on accuracy and cost. It achieves higher ImageNet-1k top1 accuracy or the same accuracy with 1.3x reduction in FLOPs and 2.5x drop in training cost (GPU*hrs).
more »
« less
- Award ID(s):
- 2420977
- PAR ID:
- 10656331
- Publisher / Repository:
- Springer Nature Switzerland
- Date Published:
- ISBN:
- 978-3-031-73023-8
- Page Range / eLocation ID:
- 315 to 331
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Motivated by the transformative capabilities of large language models (LLMs) across various natural language tasks, there has been a growing demand to deploy these models effectively across diverse real-world applications and platforms. However, the challenge of efficiently deploying LLMs has become increasingly pronounced due to the varying application-specific performance requirements and the rapid evolution of computational platforms, which feature diverse resource constraints and deployment flows. These varying requirements necessitate LLMs that can adapt their structures (depth and width) for optimal efficiency across different platforms and application specifications. To address this critical gap, we propose AmoebaLLM, a novel framework designed to enable the instant derivation of LLM subnets of arbitrary shapes, which achieve the accuracyefficiency frontier and can be extracted immediately after a one-time fine-tuning. In this way, AmoebaLLM significantly facilitates rapid deployment tailored to various platforms and applications. Specifically, AmoebaLLM integrates three innovative components: (1) a knowledge-preserving subnet selection strategy that features a dynamic-programming approach for depth shrinking and an importancedriven method for width shrinking; (2) a shape-aware mixture of LoRAs to mitigate gradient conflicts among subnets during fine-tuning; and (3) an in-place distillation scheme with loss-magnitude balancing as the fine-tuning objective. Extensive experiments validate that AmoebaLLM not only sets new standards in LLM adaptability but also successfully delivers subnets that achieve stateof-the-art trade-offs between accuracy and efficiency. Our code is available at https://github.com/GATECH-EIC/AmoebaLLM.more » « less
-
Deep neural network (DNN) is the de facto standard for running a variety of computer vision applications over mobile and embedded systems. Prior to deployment, a DNN is specialized by training to fit the target use scenario (depending on computing power and visual data input). To handle its costly training and meet diverse deployment needs, a “Train Once, Deploy Everywhere” paradigm has been recently proposed by training one super-network and selecting one out of many sub-networks (part of the super-network) for the target scenario; This empowers efficient DNN deployment at low training cost (training once). However, the existing studies tackle some deployment factors like computing power and source data but largely overlook the impact of their runtime dynamics (say, time-varying visual contents and GPU/CPU workloads). In this work, we propose OPA to cover all these deployment factors, particularly those along with runtime dynamics in visual data contents and computing resources. To quickly and accurately learn which sub-network runs “best” in the dynamic deployment scenario, we devise a “One-Predict-All” approach with no need to run all the candidate sub-networks. Instead, we first develop a shallow sub-network to test the water and then use its test results to predict the performance of all other deeper sub-networks. We have implemented and evaluated OPA. Compared to the state-of-the-art, OPA has achieved up to 26% higher Top-1 accuracy for a given latency requirement.more » « less
-
null (Ed.)With the success of Deep Neural Networks (DNN), many recent works have been focusing on developing hardware accelerator for power and resource-limited system via model compression techniques, such as quantization, pruning, low-rank approximation and etc. However, almost all existing compressed DNNs are fixed after deployment, which lacks run-time adaptive structure to adapt to its dynamic hardware resource allocation, power budget, throughput requirement, as well as dynamic workload. As the countermeasure, to construct a novel run-time dynamic DNN structure, we propose a novel DNN sub-network sampling method via non-uniform channel selection for subnets generation. Thus, user can trade off between power, speed, computing load and accuracy on-the-fly after the deployment, depending on the dynamic requirements or specifications of the given system. We verify the proposed model on both CIFAR-10 and ImageNet dataset using ResNets, which outperforms the same sub-nets trained individually and other related works. It shows that, our method can achieve latency trade-off among 13.4, 24.6, 41.3, 62.1(ms) and 30.5, 38.7, 51, 65.4(ms) for GPU with 128 batch-size and CPU respectively on ImageNet using ResNet18.more » « less
-
In the past years, Deep convolution neural network has achieved great success in many artificial intelligence applications. However, its enormous model size and massive computation cost have become the main obstacle for deployment of such powerful algorithm in the low power and resource-limited mobile systems. As the countermeasure to this problem, deep neural networks with ternarized weights (i.e. -1, 0, +1) have been widely explored to greatly reduce model size and computational cost, with limited accuracy degradation. In this work, we propose a novel ternarized neural network training method which simultaneously optimizes both weights and quantizer during training, differentiating from prior works. Instead of fixed and uniform weight ternarization, we are the first to incorporate the thresholds of weight ternarization into a closed-form representation using truncated Gaussian approximation, enabling simultaneous optimization of weights and quantizer through back-propagation training. With both of the first and last layer ternarized, the experiments on the ImageNet classification task show that our ternarized ResNet-18/34/50 only has 3.9/2.52/2.16% accuracy degradation in comparison to the full-precision counterparts.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Book Chapter:
- https://doi.org/10.1007/978-3-031-73024-5_19
