PT-Logo
Forgot your password?
BenchmarkXPRT Blog banner

Tag Archives: machine learning

Understanding AIXPRT batch size

Last week, we wrote about the basics of understanding AIXPRT results. This week, we’re discussing one of the benchmark’s key test configuration variables: batch size. Talking about batch size can be confusing, because the phrase can refer to different concepts depending on the machine learning (ML) context in which it’s used. AIXPRT tests inference, so we’ll focus on how we use batch sizes in that context. For those who are interested, we provide more information about training batch size at the bottom of this post.

Batch size in inference
In the context of ML inference, the concept of batch size is straightforward. It simply refers to the number of combined input samples (e.g., images) that the tester wants the algorithm to process simultaneously. The purpose of adjusting batch size when testing inference performance is to achieve an optimal balance between latency (speed) and throughput (the total amount processed over time).

Because of the lighter load of processing one image at a time, Batch 1 often produces the fastest latency times, and can be a good indicator of how a system handles near-real-time inference demands from client devices. Larger batch sizes (8, 16, 32, 64, or 128) can result in higher throughput on test hardware that is capable of completing more inference work in parallel. However, this increased throughput can come at the expense of latency. Running concurrent inferences via larger batch sizes is a good way to test for maximum throughput on servers.

Configuring inference batch size in AIXPRT
A good practice when trying to figure out where to start with batch size is to match the batch size to the number of cores under test (e.g., Batch 8 for eight cores). To adjust batch size in AIXPRT, testers must edit the configuration files located in AIXPRT/Config. To represent a spectrum of common tunings, AIXPRT CP2 tests Batches 1, 2, 4, 8, 16, and 32 by default.

The screenshot below shows part of a sample config file. The numbers in the lines immediately below “batch_sizes” indicate the batch size. This test configuration would run tests using both Batch 1 and Batch 2. To change batch size, simply replace those numbers and save the changes.

Config_snip
Batch size in training
As we noted above, training batch size is different than inference batch size. For training, a batch is the group of samples used to train a model during one iteration and batch size is number of samples in a batch. (Note that in this context, an iteration is a single update of the algorithm’s parameters, not a complete test run.) With a batch size of one, the algorithm applies a single training sample to an image it is processing before updating its parameters. With a batch size of two, it would apply two training examples to an image before updating its parameters, and so on. Because neural network algorithms are iterative, a larger batch size setting during training increases the total number of iterations that occur during each pass through the data set. In combination with other variables, training batch size may ultimately affect metrics such as model accuracy and convergence (the point where additional training does not improve accuracy).

In the coming weeks, we’ll discuss other test configuration variables such as precision and the number of concurrent instances. We hope this series of blog entries will answer some of the common questions people have when first running the benchmark and help to make the AIXPRT testing process more approachable for testers who are just starting to explore machine learning. If you have any questions or comments, please feel free to contact us.

Justin

Understanding AIXPRT results

Last week, we discussed the changes we made to the AIXPRT Community Preview 2 (CP2) download page as part of our ongoing effort to make AIXPRT easier to use. This week, we want to discuss the basics of understanding AIXPRT results by talking about the numbers that really matter and how to access and read the actual results files.

To understand AIXPRT results at a high level, it’s important to revisit the core purpose of the benchmark. AIXPRT’s bundled toolkits measure inference latency (the speed of image processing) and throughput (the number of images processed in a given time period) for image recognition (ResNet-50) and object detection (SSD-MobileNet v1) tasks. Testers have the option of adjusting variables such as batch size (the number of input samples to process simultaneously) to try and achieve higher levels of throughput, but higher throughput can come at the expense of increased latency per task. In real-time or near real-time use cases such as performing image recognition on individual photos being captured by a camera, lower latency is important because it improves the user experience. In other cases, such as performing image recognition on a large library of photos, achieving higher throughput might be preferable; designating larger batch sizes or running concurrent instances might allow the overall workload to complete more quickly.

The dynamics of these performance tradeoffs ensure that there is no single good score for all machine learning scenarios. Some testers might prefer lower latency, while others would sacrifice latency to achieve the higher level of throughput that their use case demands.

Testers can find latency and throughput numbers for each completed run in a JSON results file in the AIXPRT/Results folder. The test also generates CSV results files that are in the same folder. The raw results files report values for each AI task configuration (e.g., ResNet-50, Batch1, on CPU). Parsing and consolidating the raw data can take some time, so we’re developing a results file parsing tool to make the job much easier.

The results parsing tool is currently available in the AIXPRT CP2 OpenVINO – Windows package, and we hope to make it available for more packages soon. Using the tool is as simple as running a single command, and detailed instructions for how to do so are in the AIXPRT OpenVINO on Windows user guide. The tool produces a summary (example below) that makes it easier to quickly identify relevant comparison points such as maximum throughput and minimum latency.

AIXPRT results summary

In addition to the summary, the tool displays the throughput and latency results for each AI task configuration tested by the benchmark. AIXPRT runs each AI task multiple times and reports the average inference throughput and corresponding latency percentiles.

AIXPRT results details

We hope that this information helps to make it easier to understand AIXPRT results. If you have any questions or comments, please feel free to contact us.

Justin

Navigating the AIXPRT Community Preview download page just got easier

AIXPRT Community Preview 2 (CP2) has been generating quite a bit of interest among the BenchmarkXPRT Development Community and members of the tech press. We’re excited that the tool has piqued curiosity and that folks are recognizing its value for technical analysis. When talking with folks about test setup and configuration, we keep hearing the same questions:

  • How do I find the exact toolkit or package that I need?
  • How do I find the instructions for a specific toolkit?
  • What test configuration variables are most important for producing consistent, relevant results?
  • How do I know which values to choose when configuring options such as iterations, concurrent instances, and batch size?


In the coming weeks, we’ll be working to provide detailed answers to questions about test configuration. In response to the confusion about finding specific packages and instructions, we’ve redesigned the CP2 download page to make it easier for you to find what you need. Below, we show a snapshot from the new CP2 download table. Instead of having to download the entire CP2 package that includes the OpenVINO, TensorFlow, and TensorRT in TensorFlow test packages, you can now download one package at a time. In the Documentation column, we’ve posted package-specific instructions, so you won’t have to wade through the entire installation guide to find the instructions you need.

AIXPRT Community Preview download table

We hope these changes make it easier for people to experiment with AIXPRT. As always, please feel free to contact us with any questions or comments you may have.

Justin

Making AIXPRT easier to use

We’re glad to see so much interest in the AIXPRT CP2 build. Over the past few days, we’ve received two questions about the setup process: 1) where to find instructions for setting up AIXPRT on Windows, and 2) whether we could make it easier to install Intel OpenVINO on test systems.

In response to the first question, testers can find the relevant instructions for each framework in the readme files included in the AIXPRT install package. Instructions for Windows installation are in section 3 of the OpenVINO and TensorFlow readmes. Please note that whether you’re running AIXPRT on Ubuntu or Windows, be sure to read the “Known Issues” section in the readme, as there may be issues relevant to your specific configuration.

The readme files for each respective framework in the CP2 package are located here:

  • AIXPRT_0.5_CP2\AIXPRT_OpenVINO_0.5_CP2.zip\AIXPRT\Modules\Deep-Learning
  • AIXPRT_0.5_CP2\AIXPRT_TensorFLow_0.5_CP2.zip\AIXPRT\Modules\Deep-Learning
  • AIXPRT_0.5_CP2\AIXPRT_TensorFlow_TensorRT_0.5_CP2.zip\AIXPRT\Modules\Deep-Learning


We’re also working on consolidating the instructions into a central document that will make it easier for everyone to find the instructions they need.

In response to the question about OpenVINO installation, we’re working on an AIXPRT CP2 package that includes a precompiled version of OpenVINO R5.0.1 for easy installation on Windows via a few quick commands, and a script that installs the necessary OpenVINO dependencies. We’re currently testing the build, and we’ll make it available to testers as soon as possible.

The tests themselves will not change, so the new build will not influence existing results from Ubuntu or Windows. We hope it will simply facilitate the setup and testing process for many users.

We appreciate each bit of feedback that we receive, so if you have any suggestions for AIXPRT, please let us know!

Justin

News on AIXPRT development

(more…)

Answering questions about the AIXPRT Community Preview

Over the last two weeks, we’ve received a few questions about the AIXPRT Community Preview. Specifically, community members have asked about the project’s focus, possible future steps, and the results table. We decided to answer each of these here in the blog, since others are likely to have the same questions. We encourage folks to submit any new questions they may have.

PT previously stated that AIXPRT would be focused on edge devices. The current published results are from desktops and laptops. Is the focus of AIXPRT changing?

In the past, we did say that the focus of AIXPRT would be edge inference devices. After much feedback, we’ve come to understand that focus is probably too restrictive. PCs and laptops are using inference machine learning, and a decent amount of inference is taking place on servers in the cloud until phones are capable enough to handle the workloads. We now see all of these devices as potential targets for AIXPRT.

How did you choose the current results in your database?

We ran the AIXPRT CP on some of the systems we used during development and testing. We will continue to publish additional results as we test available systems in our lab. We’d love to get results from the community that cover a wider base of devices.

Will you be publishing results from servers?

We welcome server results submissions from the community, and will review them for publication on our site.

Will AIXPRT ever be available for Windows systems?

This is a possibility we’re actively exploring, and we hope to be able to share more about it soon.

What’s the best way to navigate the results table?

AIXPRT can run three toolkits, utilize two networks, and target CPU or GPU hardware. Together, these configuration options produce a lot of data points. To make it easier to handle all these variables, we’re working to improve the navigation, sorting, and filtering capabilities of the results table. In the meantime, a few tips:

  • There are two tabs at the top of the table, one for the ResNet-50 network and one for the SSD-MobileNet network. You can click the tabs to move between results for these networks.
  • Clicking any of the column headers will sort the data in that column A-Z (with the first click) or Z-A (with a second click).
  • To see if an individual test targeted a system’s CPU or GPU, read the description in the Summary column, e.g. Intel Core i7-7600U GPU / OpenVINO.
  • Clicking the entry in the Source column will take you to a more detailed page listing additional test configuration and system hardware information.

 

We’ll continue to share more information about AIXPRT in the coming weeks. Do you have additional questions or comments about AIXPRT? Let us know.

Justin

Check out the other XPRTs: