Transformer models have emerged as the state-of-the-art in many natural language processing and computer vision applications due to their capability of attending to longer sequences of tokens and supporting parallel processing more efficiently. Nevertheless, the training and inference of transformer models are computationally expensive and memory intensive. Meanwhile, utilizing the sparsity in deep learning models has proven to be an effective approach to alleviate the computation challenge as well as help to fit large models in edge devices. As high-performance CPUs and GPUs are generally not flexible enough to explore low-level sparsity, a number of specialized hardware accelerators have been proposed for transformer models. This paper provides a comprehensive review of hardware transformer accelerators that have been proposed to explore sparsity for computation and memory optimizations. We classify existing works based on the strategies of utilizing sparsity and identify their pros and cons in those strategies. Based on our analysis, we point out promising directions and recommendations for future works on improving the effective sparse execution of transformer hardware accelerators.
more »
« less
22.9 A 12nm 18.1TFLOPs/W Sparse Transformer Processor with Entropy-Based Early Exit, Mixed-Precision Predication and Fine-Grained Power Management
Large language models have substantially advanced nuance and context understanding in natural language processing (NLP), further fueling the growth of intelligent conversational interfaces and virtual assistants. However, their hefty computational and memory demands make them potentially expensive to deploy on cloudless edge platforms with strict latency and energy requirements. For example, an inference pass using the state-of-the-art BERT-base model must serially traverse through 12 computationally intensive transformer layers, each layer containing 12 parallel attention heads whose outputs concatenate to drive a large feed-forward network. To reduce computation latency, several algorithmic optimizations have been proposed, e.g., a recent algorithm dynamically matches linguistic complexity with model sizes via entropy-based early exit. Deploying such transformer models on edge platforms requires careful co-design and optimizations from algorithms to circuits, where energy consumption is a key design consideration.
more »
« less
- PAR ID:
- 10417643
- Date Published:
- Journal Name:
- 2023 IEEE International Solid- State Circuits Conference (ISSCC)
- Page Range / eLocation ID:
- 342 to 344
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Convolutional neural networks (CNNs) play an important role in today's mobile and edge computing systems for vision-based tasks like object classification and detection. However, state-of-the-art methods on CNN acceleration are trapped in either limited practical latency speed-up on general computing platforms or latency speed-up with severe accuracy loss. In this paper, we propose a spatial-based dynamic CNN acceleration framework, NeuLens, for mobile and edge platforms. Specially, we design a novel dynamic inference mechanism, assemble region-aware convolution (ARAC) supernet, that peels off redundant operations inside CNN models as many as possible based on spatial redundancy and channel slicing. In ARAC supernet, the CNN inference flow is split into multiple independent micro-flows, and the computational cost of each can be autonomously adjusted based on its tiled-input content and application requirements. These micro-flows can be loaded into hardware like GPUs as single models. Consequently, its operation reduction can be well translated into latency speed-up and is compatible with hardware-level accelerations. Moreover, the inference accuracy can be well preserved by identifying critical regions on images and processing them in the original resolution with large micro-flow. Based on our evaluation, NeuLens outperforms baseline methods by up to 58% latency reduction with the same accuracy and by up to 67.9% accuracy improvement under the same latency/memory constraints.more » « less
-
This paper explores the problem of deploying machine learning (ML)-based object detection and segmentation models on edge platforms to enable realtime caveline detection for Autonomous Underwater Vehicles (AUVs) used for under-water cave exploration and mapping. We specifically investigate three ML models, i.e., U-Net, Vision Transformer (ViT), and YOLOv8, deployed on three edge platforms: Raspberry Pi-4, Intel Neural Compute Stick 2 (NCS2), and NVIDIA Jetson Nano. The experimental results unveil clear tradeoffs between model accuracy, processing speed, and energy consumption. The most accurate model has shown to be U-Net with an 85.53 F1-score and 85.38 Intersection Over Union (IoU) value. Meanwhile, the highest inference speed and lowest energy consumption are achieved by the YOLOv8 model deployed on Jetson Nano operating in the high-power and low-power modes, respectively. The comprehensive quantitative analyses and comparative results provided in the paper highlight important nuances that can guide the deployment of caveline detection systems on underwater robots for ensuring safe and reliable AUV navigation during underwater cave exploration and mapping missions.more » « less
-
Model parallelism has become a necessity for training modern large-scale deep language models. In this work, we identify a new and orthogonal dimension from existing model parallel approaches: it is possible to perform pipeline parallelism within a single training sequence for Transformer-based language models thanks to its autoregressive property. This enables a more fine-grained pipeline compared with previous work. With this key idea, we design TeraPipe, a high-performance token-level pipeline parallel algorithm for synchronous model-parallel training of Transformer-based language models. We develop a novel dynamic programming-based algorithm to calculate the optimal pipelining execution scheme given a specific model and cluster configuration. We show that TeraPipe can speed up the training by 5.0x for the largest GPT-3 model with 175 billion parameters on an AWS cluster with 48 p3.16xlarge instances compared with state-of-the-art model-parallel methods. The code for reproduction can be found at https://github.com/zhuohan123/terapipemore » « less
-
The rapidly increasing size of deep-learning models has renewed interest in alternatives to digital-electronic computers as a means to dramatically reduce the energy cost of running state-of-the-art neural networks. Optical matrix-vector multipliers are best suited to performing computations with very large operands, which suggests that large Transformer models could be a good target for them. In this paper, we investigate---through a combination of simulations and experiments on prototype optical hardware---the feasibility and potential energy benefits of running Transformer models on future optical accelerators that perform matrix-vector multiplication. We use simulations, with noise models validated by small-scale optical experiments, to show that optical accelerators for matrix-vector multiplication should be able to accurately run a typical Transformer architecture model for language processing. We demonstrate that optical accelerators can achieve the same (or better) perplexity as digital-electronic processors at 8-bit precision, provided that the optical hardware uses sufficiently many photons per inference, which translates directly to a requirement on optical energy per inference. We studied numerically how the requirement on optical energy per inference changes as a function of the Transformer width $d$ and found that the optical energy per multiply--accumulate (MAC) scales approximately as $\frac{1}{d}$, giving an asymptotic advantage over digital systems. We also analyze the total system energy costs for optical accelerators running Transformers, including both optical and electronic costs, as a function of model size. We predict that well-engineered, large-scale optical hardware should be able to achieve a $100 \times$ energy-efficiency advantage over current digital-electronic processors in running some of the largest current Transformer models, and if both the models and the optical hardware are scaled to the quadrillion-parameter regime, optical accelerators could have a $>8,000\times$ energy-efficiency advantage. Under plausible assumptions about future improvements to electronics and Transformer quantization techniques (5× cheaper memory access, double the digital--analog conversion efficiency, and 4-bit precision), we estimate that the energy advantage for optical processors versus electronic processors operating at 300~fJ/MAC could grow to $>100,000\times$.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/ISSCC42615.2023.10067817
