Simple Dynamic Batching Inference

Related tags

Deep LearningSDBI
Overview

Simple Dynamic Batching Inference

解决了什么问题?

众所周知,Batch对于GPU上深度学习模型的运行效率影响很大。。。

是在Inference时。搜索、推荐等场景自带比较大的batch,问题不大。但更多场景面临的往往是稀碎的请求(比如图片服务里一次一张图)。

如果想提高服务的吞吐,把稀碎的请求动态攒成Batch再送GPU处理就是刚需。

NV的Triton包含了Dynamic Batching功能。我也用cpp写过一版。但是发现在部署、特别是给别人用python来调用的时候,始终是比较麻烦的。比如要各种配置环境或用NGC的镜像、走个本地rpc等。。

反过来想,只要程序瓶颈还卡在计算上,就有机会用python写一版至少吞吐上可以打平cpp的Dynamic Batching。好处是使用会方便很多。

出于个人需要和兴趣,之前基于multiprocess.Queue写过一版Dynamic Batching。但是Queue本身对于延迟的影响非常大,数字比较难看。

最近发现Python 3.8支持了共享内存,用python写了个基于SharedMemory的Dynamic Batching。

跟大家分享一下效果。

测试环境

模型Resnet50,输入(N,3,224,224)。使用某云的V100。

测试结果

我们先测一下Torch性能上限,好对数据有个基本了解。

然后一步步看不同功能的影响。

对应测试命令:

# 生成一个假模型
python fake_resnet50.py
# 测试
python benchmark.py  --no_dynamic_batch --worker_num=N --worker_batch=M

MPS

多进程Torch + MPS。

进程数量 Batch Latency Throughput
1 1 4.54 ms 220.10 pic/s
4 1 8.05 ms 496.52 pic/s
8 1 13.97 ms 572.57 pic/s
16 1 28.15 ms 526.42 pic/s

可以看出MPS是很有效的,没有MPS时,多进程轮占时间片,多个进程吞吐基本也就卡在200多。

加了多进程后,多进程的kernel在同一context下调度。在8的时候达到最高。

Batching

基于以上数据,再看下Batching的影响。

进程数量 Batch Latency Throughput
4 1 8.05 ms 496.52 pic/s
1 4 6.43 ms 622.07 pic/s
进程数量 Batch Latency Throughput
8 1 13.97 ms 572.57 pic/s
1 8 10.43 ms 766.93 pic/s
进程数量 Batch Latency Throughput
16 1 28.15 ms 526.42 pic/s
1 16 18.03 ms 887.20 pic/s

可以看到MPS虽然对吞吐有帮助,但是有条件的话,Batching依旧是更好的选择。

MPS+Batching测Torch上限

在测一下Batch=32(或者其他比较高的数字都可),看一下torch框架的上限。

进程数量 Batch Latency Throughput
1 32 33.54 ms 953.60 pic/s
2 32 56.98 ms 1123.20 pic/s
3 32 78.96 ms 1215.47 pic/s
4 32 109.89 ms 1164.80 pic/s

即便batch比较大了,但MPS依旧有提升。

Dynamic Batching

实际应用中,琐碎请求会带来的性能下降。如果对于延迟的要求没有非常苛刻,那么是可以通过牺牲一部分延迟(用来打Batch),换取更高的吞吐(省钱)。

所以这轮测试的场景是,有N个数据(业务)进程,每个进程数据batch=1,达到MPS+Batching的上限吞吐。

先试一下对上述最大吞吐的case。128个数据(业务)进程,每个进程灌一张图,后台通过共享内存传输数据并打batch。

测试命令:

python benchmark.py --worker_num=128 --worker_batch=1 --max_batch_size=32 --model_num=3 --wait_time=0.01
数据(业务)进程 GPU模型进程 Latency Throughput
128 3 103.45 ms 1237.33 pic/s

能够达到极限延迟,但比最理想的情况增加了20%+的延迟。

找个小的场景试一下:

python benchmark.py --worker_num=8 --worker_batch=1 --max_batch_size=4 --model_num=2 --wait_time=0.003
数据(业务)进程 GPU模型进程 Latency Throughput
8 2 13.04 ms 613.40 pic/s

跟前面Torch测试的数字对比,可以理解成这case下8个请求进程被分成两组,总体基本能够达到batch=4的吞吐。

时间都去哪了?

针对1200+的最大吞吐场景分析了一下:

延迟由 batch + MPS 的 79 ms 增加至 Dynamic Batching 的 103ms.其中,

  • 19ms 左右是拼batch的时间,其中10ms是命令中的等待时间,还有8.3ms的np.concat时间。
  • 分割输出回各数据进程大概用了1ms。
  • 各种队列的等待时间。

总的来说没有不太合理的地方,在benchmark里我也把各部分时间收集和打出来了。

施工图

施工图

虽然源码不长(<1000行),结构也简单。但各种进程和通信还是有点多的。

程序启动时创建context进程,每个数据进程创建模型实例时:

  • context 进程会查看是否已存在对应的模型backend进程
    • 存在 -> 通过shared memory 建立连接
    • 不存在 -> 创建backend进程 -> 创建模型进程
  • 多个模型进程是为了充分利用MPS
  • 当用户进程中有多段模型时,会创建相应多个backend进程,比如识别+检测等等
  • 进程间不传输数据,仅传输shared memory地址和tensor元信息。

代码 & 相关说明

原理大概就是这个 shared_memory sample

测试代码:benchmark.py

使用样例:sample.py

  • 基本跟用pytorch差不多,load+forward。但是:
    • 要指定数据最大尺寸,用来分配shared memory
    • 最后要用一个Run函数启动,因为要提前初始化一些进程变量
    • 需要为模型指定name。当程序涉及到多个模型的时候,数据进程通过name连接到特定的模型进程。

Konwn issues

multiprocess.shared_memory在回收时,在一些系统下会报leak或已经释放的error/warning,一些系统正常。

错的系统我跑官方示例也有错。所以还不好判断是什么原因。如果觉得可以忍又不想烦可以用下面的命令禁掉。

export PYTHONWARNINGS=ignore

最后

If 有人感兴趣 and 我有时间

  • 支持一下TensorRT/TensorCore FP16,以及某个特定版本的TF。
  • 输出还没有全用shared memory(主要是我懒),所以大输出模型的 吞吐/延迟 会受到数据拷贝的影响。可以改进。。。
Employee-Managment - Company employee registration software in the face recognition system

Employee-Managment Company employee registration software in the face recognitio

Alireza Kiaeipour 7 Jul 10, 2022
PyTorch implementation of MICCAI 2018 paper "Liver Lesion Detection from Weakly-labeled Multi-phase CT Volumes with a Grouped Single Shot MultiBox Detector"

Grouped SSD (GSSD) for liver lesion detection from multi-phase CT Note: the MICCAI 2018 paper only covers the multi-phase lesion detection part of thi

Sang-gil Lee 36 Oct 12, 2022
Pytorch implementation of PTNet for high-resolution and longitudinal infant MRI synthesis

Pyramid Transformer Net (PTNet) Project | Paper Pytorch implementation of PTNet for high-resolution and longitudinal infant MRI synthesis. PTNet: A Hi

Xuzhe Johnny Zhang 6 Jun 08, 2022
Recurrent Variational Autoencoder that generates sequential data implemented with pytorch

Pytorch Recurrent Variational Autoencoder Model: This is the implementation of Samuel Bowman's Generating Sentences from a Continuous Space with Kim's

Daniil Gavrilov 347 Nov 14, 2022
Multimodal commodity image retrieval 多模态商品图像检索

Multimodal commodity image retrieval 多模态商品图像检索 Not finished yet... introduce explain:The specific description of the project and the product image dat

hongjie 8 Nov 25, 2022
Model Zoo for MindSpore

Welcome to the Model Zoo for MindSpore In order to facilitate developers to enjoy the benefits of MindSpore framework, we will continue to add typical

MindSpore 226 Jan 07, 2023
The personal repository of the work: *DanceNet3D: Music Based Dance Generation with Parametric Motion Transformer*.

DanceNet3D The personal repository of the work: DanceNet3D: Music Based Dance Generation with Parametric Motion Transformer. Dataset and Results Pleas

南嘉Nanga 36 Dec 21, 2022
Example scripts for the detection of lanes using the ultra fast lane detection model in ONNX.

Example scripts for the detection of lanes using the ultra fast lane detection model in ONNX.

Ibai Gorordo 35 Sep 07, 2022
TensorFlow Implementation of Unsupervised Cross-Domain Image Generation

Domain Transfer Network (DTN) TensorFlow implementation of Unsupervised Cross-Domain Image Generation. Requirements Python 2.7 TensorFlow 0.12 Pickle

Yunjey Choi 865 Nov 17, 2022
BasicVSR++: Improving Video Super-Resolution with Enhanced Propagation and Alignment

BasicVSR++: Improving Video Super-Resolution with Enhanced Propagation and Alignment

Holy Wu 35 Jan 01, 2023
COVID-VIT: Classification of Covid-19 from CT chest images based on vision transformer models

COVID-ViT COVID-VIT: Classification of Covid-19 from CT chest images based on vision transformer models This code is to response to te MIA-COV19 compe

17 Dec 30, 2022
Change Detection in SAR Images Based on Multiscale Capsule Network

SAR_CD_MS_CapsNet Code for the paper "Change Detection in SAR Images Based on Multiscale Capsule Network" , IEEE Geoscience and Remote Sensing Letters

Feng Gao 21 Nov 29, 2022
An e-commerce company wants to segment its customers and determine marketing strategies according to these segments.

customer_segmentation_with_rfm Business Problem : An e-commerce company wants to

Buse Yıldırım 3 Jan 06, 2022
LexGLUE: A Benchmark Dataset for Legal Language Understanding in English

LexGLUE: A Benchmark Dataset for Legal Language Understanding in English ⚖️ 🏆 🧑‍🎓 👩‍⚖️ Dataset Summary Inspired by the recent widespread use of th

95 Dec 08, 2022
Behind the Curtain: Learning Occluded Shapes for 3D Object Detection

Behind the Curtain: Learning Occluded Shapes for 3D Object Detection Acknowledgement We implement our model, BtcDet, based on [OpenPcdet 0.3.0]. Insta

Qiangeng Xu 163 Dec 19, 2022
Nightmare-Writeup - Writeup for the Nightmare CTF Challenge from 2022 DiceCTF

Nightmare: One Byte to ROP // Alternate Solution TLDR: One byte write, no leak.

1 Feb 17, 2022
Material related to the Principles of Cloud Computing course.

CloudComputingCourse Material related to the Principles of Cloud Computing course. This repository comprises material that I use to teach my Principle

Aniruddha Gokhale 15 Dec 02, 2022
MXNet implementation for: Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octave Convolution

Octave Convolution MXNet implementation for: Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octave Convolution Imag

Meta Research 549 Dec 28, 2022
Doge-Prediction - Coding Club prediction ig

Doge-Prediction Coding Club prediction ig Basically: Create an application that

1 Jan 10, 2022
Suite of 500 procedurally-generated NLP tasks to study language model adaptability

TaskBench500 The TaskBench500 dataset and code for generating tasks. Data The TaskBench dataset is available under wget http://web.mit.edu/bzl/www/Tas

Belinda Li 20 May 17, 2022