Digging into MLPerf Benchmark Suite to Inform AI Infrastructure Decisions

By John Russell

April 9, 2019

With machine learning and deep learning storming into the datacenter, the new challenge is optimizing infrastructure choices to support diverse ML and DL workflows. How many accelerators do you (really) need? Which interconnect scheme works best? How do various frameworks compare on different compute architectures? When can you choose less expensive CPUs and rely on the GPUs to do the magic? How different are the needs of training versus inferencing?

“There is no one server that does the job perfectly well,” said Ramesh Radhakrishnan, distinguished engineer, Dell EMC, to a packed session[I] at GTC last month. “You see a variety of servers used to execute these kinds of workloads.” Precisely to this point, a flurry of benchmarking tools is emerging to help make sense of ML/DL performance requirements and optimizations. There’s the Deep500 with grand ambitions but still very nascent and aimed mostly at very large scale systems. There are early movers – DeepBench, TF_CNN_Bench, and DAWNBench, for example – with typically narrower strengths and notable shortfalls. More recently, MLPerf has started emerging as popular tool that borrows from those coming before it.

MLPerf, a broader ML/DL benchmarking effort supported by industry and academia, is gaining a foothold at least for assessing training workloads. In his GTC session, Demystifying Deep Learning Infrastructure Choices Using MLPerf Benchmark Suite, Radhakrishnan took the audience on a test drive through MLPerf by presenting performance data from testing on four different systems (1-to-8 GPUs) with differing topologies He also offered comments about the other benchmarks and their contribution to progress so far.

Yes the GPUs profiled are all from Nvidia and the systems from Dell EMC – the occasion was GTC after all – but Radhakrishnan’s even-handed approach made for a solid primer. At HPCwire’s request, Nvidia agreed to make the link to Radhakrishnan’s streamed session public ahead of schedule and in time for inclusion in this article. It’s 44 minutes well-spent for those seeking a MLPerf overview. A few of the session highlights are covered here along with key slides from Radhakrishnan’s session.

Given the sudden rise of ML and DL it is perhaps not surprising that benchmarking tools have started sprouting. Each has value. “[Take] TF_CNN_Bench (TensorFlow convolutional neural network benchmark), for example,” said Radhakrishnan. “It’s very commonly used and people use it for different types of GPUs and it works well but is focused on a single domain, convolutional networks. If you are deploying a translation-based network that’s using RNN (recurrent neural networks), can you take the same observations that you made using a CNN (convolutional neural network) and assume the same thing for your other model? [No], there’s going to be differences. So you want to have a wide coverage on domains.”

DeepBench from Baidu does cover different domains. “It is primarily used for measuring the performance of the core operations that happen in your neural networks. [It has] CNN and RNN coverage [but] if you want to look at system level performance, this is not a tool for that. It doesn’t account for software frameworks, it doesn’t account for distributed training – you need the right set of tools for that,” noted Radhakrishnan.

Enter MLPerf. Early last year, “the folks that introduced DAWNBench from Stanford University worked with Google, Baidu, AMD, Intel and others to come up with MLPerf. This addressed a lot of shortcomings in previous benchmarks. It has coverage of different domains. In terms of metrics it borrows from DAWNBench, which uses execution time to [attain accuracy]. You are using Docker containers and producing all the information needed along with providing data sets so anyone can reproduce these [MLPerf] results,” said Radhakrishnan.

Importantly, emphasized Radhakrishnan, MLPerf has extensive support from industry and universities. At this writing more than 30 companies and roughly ten academic institutions are listed on the MLPerf website. This is key to culling out bias and driving MLPerf’s ability to make fair comparisons between systems or between two different ML/DL elements, according to Radhakrishnan. It also gives the organization “freedom to enable innovation to happen.”

So what is MLPerf? The organization describes it as, “A broad ML benchmark suite for measuring performance of ML software frameworks, ML hardware accelerators, and ML cloud platforms.” The current release, version 0.5, has five different domains: image classification, object detection, language translation, reinforcement learning, and recommendation. The MLPerf metric of choice – speedup – is the amount to time that it takes train a neural net model in these domains to state of the art target accuracy.

You can see the models used for each domain on the slide. Generally they represent state of the art models said Radhakrishnan – ResNet-50 used in image identification, is a deep (50 layer) residual neural network introduced in 2015. RNN GNMT (Google Neural Machine Translation, introduced in 2016) and Transformer (an NLP approach introduced in 2017) are likewise leading edge. One of MLPerf’s goals, said Radhakrishnan, is stay very current.

“It’s a very agile development model. There were actually seven or eight domains initially considered; these five made it [into the spec], but you can definitely see more domains being added and those models being refreshed based on what’s current and state of the art,” he said. “MLPerf is definitely a needed tool today and [we’re] really excited about looking to see how it changes over times and expands and adds domains as. It covers training now but I think they are really close to releasing the inference workload as well.”

The first MLPerf published results were in December. Nvidia, no surprise, trumpeted its strong performance. (See HPCwire article, Nvidia Leads Alpha MLPerf Benchmarking Round.) Radhakrishnan suggested those results didn’t have enough data points to examine some of the questions Dell EMC and colleagues at the University of Texas (College of Engineering) were interested in. His slide deck, which will be available soon, is good way to get a fast overview of the exercise.

“What we did in our work is actually run MLPerf against a wide range of platforms, GPU platforms, starting from a workstation with a single CPU and two GPUs, all the way to [a server with] eight GPUs. We decided to do a whole suite of tests ranging across scaling, comparing different GPU models, and looking at interconnect technologies whether it is PCIe or NVLink to help answer some of those questions,” he said.

Shown below is a kind of summary slide showing system performance based on number of GPUs used. The MLPerf score is how much faster the system executes compared to the MLPerf reference “which in this case is a single P100 instance and that P100 is not very well optimized either,” said Radhakrishnan. The smallest machine, based on Quadro GV100, scored 14.

“The score of 14 means this two-GPU platform ran 14 times faster than one P100.  However you should not make the assumption you are going to see that speed difference between P100s and GV100s every time. What this lets you do is you can actually calculate, because now that your baseline is the same, you can make the assumption that an eight-GPU platform performs four time more efficiently than my two-GPU platform. So you can make assumptions on scale. We are seeing some of the benchmarks scale linearly, so something like SSD (single shot multi-box detector) and other image classifications, get really good speed up going from two-to-four-to-eight GPUs. Some of the other benchmarks, recommendation and translation, don’t see that same efficiency and scale,” he said.

The tasks/features outlined in orange in the slide below are the areas benchmarked by Dell EMC.

“One question which always comes up is which framework should I use. Since there is so much information on Tensorflow from Google and Mxnet from Nvidia, we took those two submissions (from the initial MLPerf 0.20 run) and recreated those runs. We did this on a four-GPU NVLink platform, and plotted accuracy against time. Mxnet actually completed the target accuracy in about 245 min whereas TensorFlow took about 265 minutes. One of the reasons they are so close to each other is they take advantage of the Nvidia cuDNN library. In this case we actually looked at the GPU kernels that were being implemented and found that here were eight kernels called by both of these frameworks, they were exactly the same kernels, and they account for about 40 percent on average, of the execution time,” said Radhakrishnan.

For the purposes of his presentation, Radhakrishnan introduced two hypothetical characters Dave, a data scientist concerned only with performance, and Sally, a datacenter infrastructure manager concerned with cost, efficiency, and satisfying diverse constituencies. He discussed how their needs differ and how MLPerf can inform both.

“I showed Dave these results and he said “are you sure your Mxnet is optimized and running properly because I thought tensor flow was slower?” He was right in one way. Because this part of the submission that was done used TensorFlow version 1.12 [which] introduced XLA compiling. What XLA compiling does is basically does tensor fusion where it takes the graph and optimizes some of the optimized kernels that for your particular infrastructure, right, or your GPUs. So when we ran the same run without XLA compiler enabled, it actually took almost 200 minutes longer. One key takeaway is to make sure you are using the right optimizations, so no matter what you have underneath. If you aren’t taking advantage a compiling or using the right options you are going to leave a lot of performance on the table,” said Radhakrishnan.

One interesting result was the dramatic impact being able to use mixed precision during training. Radhakrishnan’s team looked Tesla V100-PCIe performance when used in precision and mixed precision modes.

“With Volta you get the capability of using mixed precision training. What this means is it uses a technique where it combines half-precision floating point and single-precision in order to improve performance and keep the same accuracy you get in your single precision frame. In addition to memory savings, you are also using dedicated hardware on V100s to do 4×4 matrix multiplies. However, there are some coding changes involved to take advantage of this,” noted Radhakrishnan.

As is shown in the slide below, it is worth the effort. Here improvements of roughly 150 percent to 330 percent were achieved. “A Resnet-50 job on an eight-GPU system that runs in about ten minutes if you take advantage of mixed precision; if you don’t it take advantage of it, [it takes] close to 500 minutes. After I prepared this slide yesterday, Jensen Huang (Nvidia CEO) announced you are getting this with practically just one or two lines change in your code. There’s an API exposed and once you enable that, it does different things you are supposed to do to make sure you still maintain the same accuracy,” said Radhakrishnan.

Radhakrishnan presented results from four different interconnect schemes including use of NVLink with GPUs using SXM2 form factor (permits high clock rate), use of PCIe switches, and PCIe standalone. Radhakrishnan noted the last arrangement (scheme on far right of slide) allows users to take a container “pin it to a CPU and then use a GPU” and that this provides the best CPU-to-GPU bandwidth and that’s useful for some applications though he didn’t specify examples.

“If you look at how machine learning performs across all of these designs. If you look at translation benchmarks you see that improves performance by anywhere from 40-to-50 percent and when you look at it in terms of minutes, that’s 20 minutes and 30 minutes. It is a sizeable improvement when you actually look at the percentage. There’s [also] improvement for MaskRCNN. When it comes image classification, you are seeing that the peer-to-peer capability provided by the PCIe switch makes a difference; you actually get difference between that platform and the two where you cannot do peer-to-peer.

“So system design does matter and as an infrastructure provider you definitely want to make sure to assign the tasks that are doing a lot of GPU-to-GPU communication, to the right platform. If you had a mix of platforms in your datacenter, if you have workloads that are actually doing a lot of peer-to-peer like translation benchmark, you want to make sure they get queued up on your NVLink platforms. For some of the other benchmarks, the differences that you get are not that significant,” said Radhakrishnan 

The bottom line, according to Radhakrishnan, is MLPerf provides a needed tool for assessing infrastructure choices as well as ML choices. “Use the tools to monitor and actually figure out how many GPUs you need [because] throwing as many GPUs as you can at the problem does not result in a linear increase in your performance. There are some workloads in which you can definitely see improvement and there are others where you are better off running at one or two GPUs,” he said.

The same is true for interconnect. “We looked at what NVLink brings to the table. For a two GPU system there are a couple of benchmarks for which NVLink did not make a difference,” he said but noted on larger systems NVLink can be critical when performance is your top priority although even there performance can vary by the model used.

There’s quite a bit more in Radhakrishnan’s presentation, including for example deeper dives into CPU utilization – it turns out TensorFlow is a CPU hog – and GPU utilization. It will be interesting to watch MLPerf’s development and how widely adopted it becomes.

Slide source: GTC2019

Link to Radhakrishnan’s GTC streamed (slides and audio) session: https://on-demand.gputechconf.com/gtc/2019/video/_/s9553-demystifying-deep-learning-infrastructure-choices-using-mlperf-benchmark-suite/

[i]Demystifying Deep Learning Infrastructure Choices Using MLPerf Benchmark Suite

Abstract: We’ll describe a new benchmark suite proposed by the Deep Learning community for machine learning workloads. We’ll present a quantitative analysis of an early version (0.5) of benchmark known as MLPerf and explain performance impact of NVIDIA GPU architecture across a range of DL applications. This work includes evaluating MLPerf performance on Turing, Volta, and Pascal to demonstrate the performance impact of NVIDIA GPU architecture across a range of DL applications. We’ll evaluate the impact of system-level technologies — Nvlink vs. PCIe topology — using server- and workstation-class platforms to show how system architecture impacts DL training workloads. We also plan to discuss our work to characterize MLPerf benchmark performance using profiling tools (GPU, CPU, memory & I/O), our hyperparameter tuning study (batch size, learning rate, SGD optimizer) on MLPerf performance, and map real world application use cases to MLPerf suite and how to quantify results for specific DL practitioner use cases. https://gputechconf2019.smarteventscloud.com/connect/sessionDetail.ww?SESSION_ID=263091

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industy updates delivered to you every week!

Army Seeks AI Ground Truth

April 3, 2020

Deep neural networks are being mustered by U.S. military researchers to marshal new technology forces on the Internet of Battlefield Things. U.S. Army and industry researchers said this week they have developed a “c Read more…

By George Leopold

Piz Daint Tackles Marsquakes

April 3, 2020

Even as researchers use supercomputers to probe the mysteries of earthquakes here on Earth, others are setting their sights on quakes just a little farther away. Researchers at ETH Zürich in Switzerland have applied sup Read more…

By Oliver Peckham

HPC Career Notes: April 2020 Edition

April 2, 2020

In this monthly feature, we’ll keep you up-to-date on the latest career developments for individuals in the high-performance computing community. Whether it’s a promotion, new company hire, or even an accolade, we’ Read more…

By Mariana Iriarte

AMD Epyc CPUs Now on Bare Metal IBM Cloud Servers

April 1, 2020

AMD’s expanding presence in the datacenter and cloud computing markets took a step forward with today’s announcement that its 7nm 2nd Gen Epyc 7642 CPUs are now available on IBM Cloud bare metal servers. AMD, whose Read more…

By Doug Black

Supercomputer Testing Probes Viral Transmission in Airplanes

April 1, 2020

It might be a long time before the general public is flying again, but the question remains: how high-risk is air travel in terms of viral infection? In an article for the Texas Advanced Computing Center (TACC), Faith Si Read more…

By Staff report

AWS Solution Channel

Amazon FSx for Lustre Update: Persistent Storage for Long-Term, High-Performance Workloads

Last year I wrote about Amazon FSx for Lustre and told you how our customers can use it to create pebibyte-scale, highly parallel POSIX-compliant file systems that serve thousands of simultaneous clients driving millions of IOPS (Input/Output Operations per Second) with sub-millisecond latency. Read more…

ECP Milestone Report Details Progress and Directions

April 1, 2020

The Exascale Computing Project (ECP) milestone report issued last week presents a good snapshot of progress in preparing applications for exascale computing. There are roughly 30 ECP application development (AD) subproj Read more…

By John Russell

ECP Milestone Report Details Progress and Directions

April 1, 2020

The Exascale Computing Project (ECP) milestone report issued last week presents a good snapshot of progress in preparing applications for exascale computing. Th Read more…

By John Russell

Pandemic ‘Wipes Out’ 2020 HPC Market Growth, Flat to 12% Drop Expected

March 31, 2020

As the world battles the still accelerating novel coronavirus, the HPC community has mounted a forceful response to the pandemic on many fronts. But these efforts won't inoculate the HPC industry from the economic effects of COVID-19. Market watcher Intersect360 Research has revised its 2020 forecast for HPC products and services, projecting... Read more…

By Tiffany Trader

LLNL Leverages Supercomputing to Identify COVID-19 Antibody Candidates

March 30, 2020

As COVID-19 sweeps the globe to devastating effect, supercomputers around the world are spinning up to fight back by working on diagnosis, epidemiology, treatme Read more…

By Staff report

Weather at Exascale: Load Balancing for Heterogeneous Systems

March 30, 2020

The first months of 2020 were dominated by weather and climate supercomputing news, with major announcements coming from the UK, the European Centre for Medium- Read more…

By Oliver Peckham

Q&A Part Two: ORNL’s Pooser on Progress in Quantum Communication

March 30, 2020

Quantum computing seems to get more than its fair share of attention compared to quantum communication. That’s despite the fact that quantum networking may be Read more…

By John Russell

DoE Expands on Role of COVID-19 Supercomputing Consortium

March 25, 2020

After announcing the launch of the COVID-19 High Performance Computing Consortium on Sunday, the Department of Energy yesterday provided more details on its sco Read more…

By John Russell

[email protected] Rallies a Legion of Computers Against the Coronavirus

March 24, 2020

Last week, we highlighted [email protected], a massive, crowdsourced computer network that has turned its resources against the coronavirus pandemic sweeping the globe – but [email protected] isn’t the only game in town. The internet is buzzing with crowdsourced computing... Read more…

By Oliver Peckham

Conversation: ANL’s Rick Stevens on DoE’s AI for Science Project

March 23, 2020

With release of the Department of Energy’s AI for Science report in late February, the effort to build a national AI program, modeled loosely on the U.S. Exascale Initiative, enters a new phase. Project leaders have already had early discussions with Congress... Read more…

By John Russell

[email protected] Turns Its Massive Crowdsourced Computer Network Against COVID-19

March 16, 2020

For gamers, fighting against a global crisis is usually pure fantasy – but now, it’s looking more like a reality. As supercomputers around the world spin up Read more…

By Oliver Peckham

Julia Programming’s Dramatic Rise in HPC and Elsewhere

January 14, 2020

Back in 2012 a paper by four computer scientists including Alan Edelman of MIT introduced Julia, A Fast Dynamic Language for Technical Computing. At the time, t Read more…

By John Russell

Global Supercomputing Is Mobilizing Against COVID-19

March 12, 2020

Tech has been taking some heavy losses from the coronavirus pandemic. Global supply chains have been disrupted, virtually every major tech conference taking place over the next few months has been canceled... Read more…

By Oliver Peckham

[email protected] Rallies a Legion of Computers Against the Coronavirus

March 24, 2020

Last week, we highlighted [email protected], a massive, crowdsourced computer network that has turned its resources against the coronavirus pandemic sweeping the globe – but [email protected] isn’t the only game in town. The internet is buzzing with crowdsourced computing... Read more…

By Oliver Peckham

DoE Expands on Role of COVID-19 Supercomputing Consortium

March 25, 2020

After announcing the launch of the COVID-19 High Performance Computing Consortium on Sunday, the Department of Energy yesterday provided more details on its sco Read more…

By John Russell

Steve Scott Lays Out HPE-Cray Blended Product Roadmap

March 11, 2020

Last week, the day before the El Capitan processor disclosures were made at HPE's new headquarters in San Jose, Steve Scott (CTO for HPC & AI at HPE, and former Cray CTO) was on-hand at the Rice Oil & Gas HPC conference in Houston. He was there to discuss the HPE-Cray transition and blended roadmap, as well as his favorite topic, Cray's eighth-gen networking technology, Slingshot. Read more…

By Tiffany Trader

Fujitsu A64FX Supercomputer to Be Deployed at Nagoya University This Summer

February 3, 2020

Japanese tech giant Fujitsu announced today that it will supply Nagoya University Information Technology Center with the first commercial supercomputer powered Read more…

By Tiffany Trader

Tech Conferences Are Being Canceled Due to Coronavirus

March 3, 2020

Several conferences scheduled to take place in the coming weeks, including Nvidia’s GPU Technology Conference (GTC) and the Strata Data + AI conference, have Read more…

By Alex Woodie

Leading Solution Providers

SC 2019 Virtual Booth Video Tour

AMD
AMD
ASROCK RACK
ASROCK RACK
AWS
AWS
CEJN
CJEN
CRAY
CRAY
DDN
DDN
DELL EMC
DELL EMC
IBM
IBM
MELLANOX
MELLANOX
ONE STOP SYSTEMS
ONE STOP SYSTEMS
PANASAS
PANASAS
SIX NINES IT
SIX NINES IT
VERNE GLOBAL
VERNE GLOBAL
WEKAIO
WEKAIO

Cray to Provide NOAA with Two AMD-Powered Supercomputers

February 24, 2020

The United States’ National Oceanic and Atmospheric Administration (NOAA) last week announced plans for a major refresh of its operational weather forecasting supercomputers, part of a 10-year, $505.2 million program, which will secure two HPE-Cray systems for NOAA’s National Weather Service to be fielded later this year and put into production in early 2022. Read more…

By Tiffany Trader

Exascale Watch: El Capitan Will Use AMD CPUs & GPUs to Reach 2 Exaflops

March 4, 2020

HPE and its collaborators reported today that El Capitan, the forthcoming exascale supercomputer to be sited at Lawrence Livermore National Laboratory and serve Read more…

By John Russell

Summit Supercomputer is Already Making its Mark on Science

September 20, 2018

Summit, now the fastest supercomputer in the world, is quickly making its mark in science – five of the six finalists just announced for the prestigious 2018 Read more…

By John Russell

IBM Unveils Latest Achievements in AI Hardware

December 13, 2019

“The increased capabilities of contemporary AI models provide unprecedented recognition accuracy, but often at the expense of larger computational and energet Read more…

By Oliver Peckham

TACC Supercomputers Run Simulations Illuminating COVID-19, DNA Replication

March 19, 2020

As supercomputers around the world spin up to combat the coronavirus, the Texas Advanced Computing Center (TACC) is announcing results that may help to illumina Read more…

By Staff report

IBM Debuts IC922 Power Server for AI Inferencing and Data Management

January 28, 2020

IBM today launched a Power9-based inference server – the IC922 – that features up to six Nvidia T4 GPUs, PCIe Gen 4 and OpenCAPI connectivity, and can accom Read more…

By John Russell

Summit Joins the Fight Against the Coronavirus

March 6, 2020

With the coronavirus sweeping the globe, tech conferences and supply chains are being hit hard – but now, tech is hitting back. Oak Ridge National Laboratory Read more…

By Staff report

University of Stuttgart Inaugurates ‘Hawk’ Supercomputer

February 20, 2020

This week, the new “Hawk” supercomputer was inaugurated in a ceremony at the High-Performance Computing Center of the University of Stuttgart (HLRS). Offici Read more…

By Staff report

  • arrow
  • Click Here for More Headlines
  • arrow
Do NOT follow this link or you will be banned from the site!
Share This