SE-MoE: A Scalable and Efficient Mixture-of-Experts Distributed Training and Inference System

With the increasing diversity of ML infrastructures nowadays, distributed training over heterogeneous computing systems is desired to facilitate the production of big models. Mixture-of-Experts (MoE) models have been proposed to lower the cost of training subject to the overall size of models/data through gating and parallelism in a divide-and-conquer fashion. While DeepSpeed has made efforts in carrying out large-scale MoE training over heterogeneous infrastructures, the efficiency of training and inference could be further improved from several system aspects, including load balancing, communication/computation efficiency, and memory footprint limits. In this work, we present SE-MoE that proposes Elastic MoE training with 2D prefetch and Fusion communication over Hierarchical storage, so as to enjoy efficient parallelisms in various types. For scalable inference in a single node, especially when the model size is larger than GPU memory, SE-MoE forms the CPU-GPU memory jointly into a ring of sections to load the model, and executes the computation tasks across the memory sections in a round-robin manner for efficient inference. We carried out extensive experiments to evaluate SE-MoE, where SE-MoE successfully trains a Unified Feature Optimization (UFO) model with a Sparsely-Gated Mixture-of-Experts model of 12B parameters in 8 days on 48 A100 GPU cards. The comparison against the state-of-the-art shows that SE-MoE outperformed DeepSpeed with 33% higher throughput (tokens per second) in training and 13% higher throughput in inference in general. Particularly, under unbalanced MoE Tasks, e.g., UFO, SE-MoE achieved 64% higher throughput with 18% lower memory footprints. The code of the framework will be released on: https://github.com/PaddlePaddle/Paddle.

An Empirical Study of Low Precision Quantization for TinyML

Tiny machine learning (tinyML) has emerged during the past few years aiming to deploy machine learning models to embedded AI processors with highly constrained memory and computation capacity. Low precision quantization is an important model compression technique that can greatly reduce both memory consumption and computation cost of model inference. In this study, we focus on post-training quantization (PTQ) algorithms that quantize a model to low-bit (less than 8-bit) precision with only a small set of calibration data and benchmark them on different tinyML use cases. To achieve a fair comparison, we build a simulated quantization framework to investigate recent PTQ algorithms. Furthermore, we break down those algorithms into essential components and re-assembled a generic PTQ pipeline. With ablation study on different alternatives of components in the pipeline, we reveal key design choices when performing low precision quantization. We hope this work could provide useful data points and shed lights on the future research of low precision quantization.

End-to-end Adaptive Distributed Training on PaddlePaddle

Distributed training has become a pervasive and effective approach for training a large neural network (NN) model with processing massive data. However, it is very challenging to satisfy requirements from various NN models, diverse computing resources, and their dynamic changes during a training job. In this study, we design our distributed training framework in a systematic end-to-end view to provide the built-in adaptive ability for different scenarios, especially for industrial applications and production environments, by fully considering resource allocation, model partition, task placement, and distributed execution. Based on the unified distributed graph and the unified cluster object, our adaptive framework is equipped with a global cost model and a global planner, which can enable arbitrary parallelism, resource-aware placement, multi-mode execution, fault-tolerant, and elastic distributed training. The experiments demonstrate that our framework can satisfy various requirements from the diversity of applications and the heterogeneity of resources with highly competitive performance. The ERNIE language model with 260 billion parameters is efficiently trained on thousands of AI processors with 91.7% weak scalability. The throughput of the model from the recommender system by employing the heterogeneous pipeline asynchronous execution can be increased up to 2.1 times and 3.3 times that of the GPU-only and CPU-only training respectively. Moreover, the fault-tolerant and elastic distributed training have been successfully applied to the online industrial applications, which give a reduction of 34.49% in the number of failed long-term training jobs and an increase of 33.91% for the global scheduling efficiency in the production environment.

Multilevel Image Thresholding Using a Fully Informed Cuckoo Search Algorithm

Though effective in the segmentation, conventional multilevel thresholding methods are computationally expensive as exhaustive search are used for optimal thresholds to optimize the objective functions. To overcome this problem, population-based metaheuristic algorithms are widely used to improve the searching capacity. In this paper, we improve a popular metaheuristic called cuckoo search using a ring topology based fully informed strategy. In this strategy, each individual in the population learns from its neighborhoods to improve the cooperation of the population and the learning efficiency. Best solution or best fitness value can be obtained from the initial random threshold values, whose quality is evaluated by the correlation function. Experimental results have been examined on various numbers of thresholds. The results demonstrate that the proposed algorithm is more accurate and efficient than other four popular methods.

Novel Co-variant Feature Point Matching Based on Gaussian Mixture Model

The feature frame is a key idea of feature matching problem between two images. However, most of the traditional matching methods only simply employ the spatial location information (the coordinates), which ignores the shape and orientation information of the local feature. Such additional information can be obtained along with coordinates using general co-variant detectors such as DOG, Hessian, Harris-Affine and MSER. In this paper, we develop a novel method considering all the feature center position coordinates, the local feature shape and orientation information based on Gaussian Mixture Model for co-variant feature matching. We proposed three sub-versions in our method for solving the matching problem in different conditions: rigid, affine and non-rigid, respectively, which all optimized by expectation maximization algorithm. Due to the effective utilization of the additional shape and orientation information, the proposed model can significantly improve the performance in terms of convergence speed and recall. Besides, it is more robust to the outliers.

Low-Power Computer Vision: Status, Challenges, Opportunities

Computer vision has achieved impressive progress in recent years. Meanwhile, mobile phones have become the primary computing platforms for millions of people. In addition to mobile phones, many autonomous systems rely on visual data for making decisions and some of these systems have limited energy (such as unmanned aerial vehicles also called drones and mobile robots). These systems rely on batteries and energy efficiency is critical. This article serves two main purposes: (1) Examine the state-of-the-art for low-power solutions to detect objects in images. Since 2015, the IEEE Annual International Low-Power Image Recognition Challenge (LPIRC) has been held to identify the most energy-efficient computer vision solutions. This article summarizes 2018 winners' solutions. (2) Suggest directions for research as well as opportunities for low-power computer vision.

Low Power Inference for On-Device Visual Recognition with a Quantization-Friendly Solution

The IEEE Low-Power Image Recognition Challenge (LPIRC) is an annual competition started in 2015 that encourages joint hardware and software solutions for computer vision systems with low latency and power. Track 1 of the competition in 2018 focused on the innovation of software solutions with fixed inference engine and hardware. This decision allows participants to submit models online and not worry about building and bringing custom hardware on-site, which attracted a historically large number of submissions. Among the diverse solutions, the winning solution proposed a quantization-friendly framework for MobileNets that achieves an accuracy of 72.67% on the holdout dataset with an average latency of 27ms on a single CPU core of Google Pixel2 phone, which is superior to the best real-time MobileNet models at the time.

2018 Low-Power Image Recognition Challenge

The Low-Power Image Recognition Challenge (LPIRC, https://rebootingcomputing.ieee.org/lpirc) is an annual competition started in 2015. The competition identifies the best technologies that can classify and detect objects in images efficiently (short execution time and low energy consumption) and accurately (high precision). Over the four years, the winners' scores have improved more than 24 times. As computer vision is widely used in many battery-powered systems (such as drones and mobile phones), the need for low-power computer vision will become increasingly important. This paper summarizes LPIRC 2018 by describing the three different tracks and the winners' solutions.

Radiative Transport Based Flame Volume Reconstruction from Videos

We introduce a novel approach for flame volume reconstruction from videos using inexpensive charge-coupled device (CCD) consumer cameras. The approach includes an economical data capture technique using inexpensive CCD cameras. Leveraging the smear feature of the CCD chip, we present a technique for synchronizing CCD cameras while capturing flame videos from different views. Our reconstruction is based on the radiative transport equation which enables complex phenomena such as emission, extinction, and scattering to be used in the rendering process. Both the color intensity and temperature reconstructions are implemented using the CUDA parallel computing framework, which provides real-time performance and allows visualization of reconstruction results after every iteration. We present the results of our approach using real captured data and physically-based simulated data. Finally, we also compare our approach against the other state-of-the-art flame volume reconstruction methods and demonstrate the efficacy and efficiency of our approach in four different applications: (1) rendering of reconstructed flames in virtual environments, (2) rendering of reconstructed flames in augmented reality, (3) flame stylization, and (4) reconstruction of other semitransparent phenomena.

A Quantization-Friendly Separable Convolution for MobileNets

As deep learning (DL) is being rapidly pushed to edge computing, researchers invented various ways to make inference computation more efficient on mobile/IoT devices, such as network pruning, parameter compression, and etc. Quantization, as one of the key approaches, can effectively offload GPU, and make it possible to deploy DL on fixed-point pipeline. Unfortunately, not all existing networks design are friendly to quantization. For example, the popular lightweight MobileNetV1, while it successfully reduces parameter size and computation latency with separable convolution, our experiment shows its quantized models have large accuracy gap against its float point models. To resolve this, we analyzed the root cause of quantization loss and proposed a quantization-friendly separable convolution architecture. By evaluating the image classification task on ImageNet2012 dataset, our modified MobileNetV1 model can archive 8-bit inference top-1 accuracy in 68.03%, almost closed the gap to the float pipeline.