index.html

{% extends "!layout.html" %}

{% set title = "Welcome To Neural Network Intelligence !!!"%}

{% block document %}

<div>
  <div class="chinese"><a href="https://nni.readthedocs.io/zh/latest/">简体中文</a></div>
  <b>NNI (Neural Network Intelligence)</b> is a lightweight but powerful toolkit to
  help users <b>automate</b>
  <a href="{{ pathto('FeatureEngineering/Overview') }}">Feature Engineering</a>,
  <a href="{{ pathto('NAS/Overview') }}">Neural Architecture Search</a>,
  <a href="{{ pathto('Tuner/BuiltinTuner') }}">Hyperparameter Tuning</a> and
  <a href="{{ pathto('Compressor/Overview') }}">Model Compression</a>.
</div>
<p class="topMargin">
  The tool manages automated machine learning (AutoML) experiments,
  <b>dispatches and runs</b>
  experiments' trial jobs generated by tuning algorithms to search the best neural
  architecture and/or hyper-parameters in
  <b>different training environments</b> like
  <a href="{{ pathto('TrainingService/LocalMode') }}">Local Machine</a>,
  <a href="{{ pathto('TrainingService/RemoteMachineMode') }}">Remote Servers</a>,
  <a href="{{ pathto('TrainingService/PaiMode') }}">OpenPAI</a>,
  <a href="{{ pathto('TrainingService/KubeflowMode') }}">Kubeflow</a>,
  <a href="{{ pathto('TrainingService/FrameworkControllerMode') }}">FrameworkController on K8S (AKS etc.)</a>
  <a href="{{ pathto('TrainingService/DLTSMode') }}">DLWorkspace (aka. DLTS)</a>
  and other cloud options.
</p>
<!-- Who should consider using NNI -->
<div>
  <h1 class="title">Who should consider using NNI</h1>
  <ul>
    <li>Those who want to <b>try different AutoML algorithms</b> in their training code/model.</li>
    <li>Those who want to run AutoML trial jobs <b>in different environments</b> to speed up search.</li>
    <li>Researchers and data scientists who want to easily <b>implement and experiement new AutoML
        algorithms</b>
      , may it be: hyperparameter tuning algorithm,
      neural architect search algorithm or model compression algorithm.
    </li>
    <li>ML Platform owners who want to <b>support AutoML in their platform</b></li>
  </ul>
</div>
<!-- nni release to version -->
<div class="inline">
  <h3><a href="https://github.com/microsoft/nni/releases">NNI {{ release }} has been released!</a></h3>
  <img width="48" src="_static/img/release_icon.png">
</div>
<!-- NNI capabilities in a glance -->
<div>
  <h1 class="title">NNI capabilities in a glance</h1>
  <p>
    NNI provides CommandLine Tool as well as an user friendly WebUI to manage training experiements.
    With the extensible API, you can customize your own AutoML algorithms and training services.
    To make it easy for new users, NNI also provides a set of build-in stat-of-the-art
    AutoML algorithms and out of box support for popular training platforms.
  </p>
  <p>
    Within the following table, we summarized the current NNI capabilities,
    we are gradually adding new capabilities and we'd love to have your contribution.
  </p>
</div>

<p align="center">
  <a href="#overview"><img src="_static/img/overview.svg" /></a>
</p>

<table class="list">
  <tbody>
    <tr align="center" valign="bottom" class="column">
      <td></td>
      <td class="framework">
        <b>Frameworks & Libraries</b>
      </td>
      <td>
        <b>Algorithms</b>
      </td>
      <td>
        <b>Training Services</b>
      </td>
    </tr>
    </tr>
    <tr>
      <td class="verticalMiddle"><b>Built-in</b></td>
      <td>
        <ul class="firstUl">
          <li><b>Supported Frameworks</b></li>
          <ul class="circle">
            <li>PyTorch</li>
            <li>Keras</li>
            <li>TensorFlow</li>
            <li>MXNet</li>
            <li>Caffe2</li>
            <a href="{{ pathto('SupportedFramework_Library') }}">More...</a><br />
          </ul>
        </ul>
        <ul class="firstUl">
          <li><b>Supported Libraries</b></li>
          <ul class="circle">
            <li>Scikit-learn</li>
            <li>XGBoost</li>
            <li>LightGBM</li>
            <a href="{{ pathto('SupportedFramework_Library') }}">More...</a><br />
          </ul>
        </ul>
        <ul class="firstUl">
          <li><b>Examples</b></li>
          <ul class="circle">
            <li><a href="https://github.com/microsoft/nni/tree/master/examples/trials/mnist-pytorch">MNIST-pytorch</li>
            </a>
            <li><a href="https://github.com/microsoft/nni/tree/master/examples/trials/mnist-tfv1">MNIST-tensorflow</li>
            </a>
            <li><a href="https://github.com/microsoft/nni/tree/master/examples/trials/mnist-keras">MNIST-keras</li></a>
            <li><a href="{{ pathto('TrialExample/GbdtExample') }}">Auto-gbdt</a></li>
            <li><a href="{{ pathto('TrialExample/Cifar10Examples') }}">Cifar10-pytorch</li></a>
            <li><a href="{{ pathto('TrialExample/SklearnExamples') }}">Scikit-learn</a></li>
            <li><a href="{{ pathto('TrialExample/EfficientNet') }}">EfficientNet</a></li>
            <a href="{{ pathto('SupportedFramework_Library') }}">More...</a><br />
          </ul>
        </ul>
      </td>
      <td align="left">
        <a href="{{ pathto('Tuner/BuiltinTuner') }}">Hyperparameter Tuning</a>
        <ul class="firstUl">
          <div><b>Exhaustive search</b></div>
          <ul class="circle">
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">Random Search</a></li>
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">Grid Search</a></li>
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">Batch</a></li>
          </ul>
          <div><b>Heuristic search</b></div>
          <ul class="circle">
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">Naïve Evolution</a></li>
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">Anneal</a></li>
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">Hyperband</a></li>
          </ul>
          <div><b>Bayesian optimization</b></div>
          <ul class="circle">
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">BOHB</a></li>
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">TPE</a></li>
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">SMAC</a></li>
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">Metis Tuner</a></li>
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">GP Tuner</a> </li>
          </ul>
          <div><b>RL Based</b></div>
          <ul class="circle">
            <li><a href="{{ pathto('Tuner/BuiltinTuner') }}">PPO Tuner</a> </li>
          </ul>
        </ul>
        <a href="{{ pathto('NAS/Overview') }}">Neural Architecture Search</a>
        <ul class="firstUl">
          <ul class="circle">
            <li><a href="{{ pathto('NAS/ENAS') }}">ENAS</a></li>
            <li><a href="{{ pathto('NAS/DARTS') }}">DARTS</a></li>
            <li><a href="{{ pathto('NAS/PDARTS') }}">P-DARTS</a></li>
            <li><a href="{{ pathto('NAS/CDARTS') }}">CDARTS</a></li>
            <li><a href="{{ pathto('NAS/SPOS') }}">SPOS</a></li>
            <li><a href="{{ pathto('NAS/Proxylessnas') }}">ProxylessNAS</a></li>
            <li><a href="{{ pathto('Tuner/NetworkmorphismTuner') }}">Network Morphism</a> </li>
            <li><a href="{{ pathto('NAS/TextNAS') }}">TextNAS</a> </li>
          </ul>
        </ul>
        <a href="{{ pathto('Compressor/Overview') }}">Model Compression</a>
        <ul class="firstUl">
          <div><b>Pruning</b></div>
          <ul class="circle">
            <li><a href="{{ pathto('Compressor/Pruner') }}">AGP Pruner</a></li>
            <li><a href="{{ pathto('Compressor/Pruner') }}">Slim Pruner</a></li>
            <li><a href="{{ pathto('Compressor/Pruner') }}">FPGM Pruner</a></li>
          </ul>
          <div><b>Quantization</b></div>
          <ul class="circle">
            <li><a href="{{ pathto('Compressor/Quantizer') }}">QAT Quantizer</a></li>
            <li><a href="{{ pathto('Compressor/Quantizer') }}">DoReFa Quantizer</a></li>
          </ul>
        </ul>
        <a href="{{ pathto('FeatureEngineering/Overview') }}">Feature Engineering (Beta)</a>
        <ul class="circle">
          <li><a href="{{ pathto('FeatureEngineering/GradientFeatureSelector') }}">GradientFeatureSelector</a></li>
          <li><a href="{{ pathto('FeatureEngineering/GBDTSelector') }}">GBDTSelector</a></li>
        </ul>
        <a href="{{ pathto('Assessor/BuiltinAssessor') }}">Early Stop Algorithms</a>
        <ul class="circle">
          <li><a href="{{ pathto('Assessor/BuiltinAssessor') }}">Median Stop</a></li>
          <li><a href="{{ pathto('Assessor/BuiltinAssessor') }}">Curve Fitting</a></li>
        </ul>
      </td>
      <td>
        <ul class="firstUl">
          <li><a href="{{ pathto('TrainingService/LocalMode') }}">Local Machine</a></li>
          <li><a href="{{ pathto('TrainingService/RemoteMachineMode') }}">Remote Servers</a></li>
          <li><b>Kubernetes based services</b></li>
          <ul class="circle">
            <li><a href="{{ pathto('TrainingService/PaiMode') }}">OpenPAI</a></li>
            <li><a href="{{ pathto('TrainingService/KubeflowMode') }}">Kubeflow</a></li>
            <li><a href="{{ pathto('TrainingService/FrameworkControllerMode') }}">FrameworkController on K8S
                (AKSetc.)</a>
            </li>
            <li><a href="{{ pathto('TrainingService/DLTSMode') }}">DLWorkspace (aka. DLTS)</a></li>
          </ul>
        </ul>
      </td>
    </tr>
    <tr valign="top">
      <td class="verticalMiddle"><b>References</b></td>
      <td>
        <ul class="firstUl">
          <li><a href="https://nni.readthedocs.io/en/latest/autotune_ref.html#trial">Python API</a></li>
          <li><a href="{{ pathto('Tutorial/AnnotationSpec') }}">NNI Annotation</a></li>
          <li><a href="https://nni.readthedocs.io/en/latest/installation.html">Supported OS</a></li>
        </ul>
      </td>
      <td>
        <ul class="firstUl">
          <li><a href="{{ pathto('Tuner/CustomizeTuner') }}">CustomizeTuner</a></li>
          <li><a href="{{ pathto('Assessor/CustomizeAssessor') }}">CustomizeAssessor</a></li>
        </ul>
      </td>
      <td>
        <ul class="firstUl">
          <li><a href="{{ pathto('TrainingService/SupportTrainingService') }}">Support TrainingService</a></li>
          <li><a href="{{ pathto('TrainingService/HowToImplementTrainingService') }}">Implement TrainingService</a></li>
        </ul>
      </td>
    </tr>
  </tbody>
</table>

<!-- Installation -->
<div>
  <h1 class="title">Installation</h1>
  <div>
    <h2 class="second-title">Install</h2>
    <p>
      NNI supports and is tested on Ubuntu >= 16.04, macOS >= 10.14.1,
      and Windows 10 >= 1809. Simply run the following `pip install`
      in an environment that has `python 64-bit >= 3.5`.
    </p>
    <div class="command-intro">Linux or macOS</div>
    <div class="command">python3 -m pip install --upgrade nni</div>
    <div class="command-intro">Windows</div>
    <div class="command">python -m pip install --upgrade nni</div>
    <p class="topMargin">If you want to try latest code, please <a href="{{ pathto('Installation') }}">install
        NNI</a> from source code.
    </p>
    <p>For detail system requirements of NNI, please refer to <a href="{{ pathto('Tutorial/InstallationLinux') }}">here</a>
      for Linux & macOS, and <a href="{{ pathto('Tutorial/InstallationWin') }}">here</a> for Windows.</p>
  </div>
  <div>
    <p>Note:</p>
    <ul>
      <li>If there is any privilege issue, add --user to install NNI in the user directory.</li>
      <li>Currently NNI on Windows supports local, remote and pai mode. Anaconda or Miniconda is highly
        recommended to install <a href="">NNI on Windows</a>.</li>
      <li>If there is any error like Segmentation fault, please refer to <a
          href="{{ pathto('Tutorial/Installation') }}">FAQ</a>. For FAQ on Windows, please refer
        to NNI on Windows.</li>
    </ul>
  </div>
  <div>
    <h2 class="second-title">Verify installation</h2>
    <p>
      The following example is built on TensorFlow 1.x. Make sure <b>TensorFlow 1.x is used</b> when running
      it.
    </p>
    <ul>
      <li>
        <p>Download the examples via clone the source code.</p>
        <div class="command">git clone -b v1.5 https://github.com/Microsoft/nni.git</div>
      </li>
      <li>
        <p>Run the MNIST example.</p>
        <div class="command-intro">Linux or macOS</div>
        <div class="command">nnictl create --config nni/examples/trials/mnist-tfv1/config.yml</div>
        <div class="command-intro">Windows</div>
        <div class="command">nnictl create --config nni\examples\trials\mnist-tfv1\config_windows.yml</div>
      </li>
      <li>
        <p>
          Wait for the message INFO: Successfully started experiment! in the command line.
          This message indicates that your experiment has been successfully started.
          You can explore the experiment using the Web UI url.
        </p>
        <!-- Indentation affects style！ -->
        <pre class="code">
INFO: Starting restful server...
INFO: Successfully started Restful server!
INFO: Setting local config...
INFO: Successfully set local config!
INFO: Starting experiment...
INFO: Successfully started experiment!
-----------------------------------------------------------------------
The experiment id is egchD4qy
The Web UI urls are: http://223.255.255.1:8080   http://127.0.0.1:8080
-----------------------------------------------------------------------

You can use these commands to get more information about the experiment
-----------------------------------------------------------------------
  commands                       description
1. nnictl experiment show        show the information of experiments
2. nnictl trial ls               list all of trial jobs
3. nnictl top                    monitor the status of running experiments
4. nnictl log stderr             show stderr log content
5. nnictl log stdout             show stdout log content
6. nnictl stop                   stop an experiment
7. nnictl trial kill             kill a trial job by id
8. nnictl --help                 get help information about nnictl
-----------------------------------------------------------------------
</pre>
      </li>
      <li>
        Open the Web UI url in your browser, you can view detail information of the experiment and
        all the submitted trial jobs as shown below. <a href="{{ pathto('Tutorial/WebUI') }}">Here</a> are more Web UI
        pages.
        <div class="ui-img">
          <img src="_images/webui_overview_page.png">
          <img src="_images/webui_trialdetail_page.png">
        </div>
  </div>
  </li>
  </ul>
</div>

<!-- Documentation -->
<div>
  <h1 class="title">Documentation</h1>
  <ul>
    <li>To learn about what's NNI, read the <a href="{{ pathto('Overview') }}">NNI Overview</a>.</li>
    <li>To get yourself familiar with how to use NNI, read the <a href="{{ pathto('index') }}">documentation</a>.</li>
    <li>To get started and install NNI on your system, please refer to <a href="{{ pathto('installation') }}">Install NNI</a>.</li>
  </ul>
</div>

<!-- Contributing -->
<div>
  <h1 class="title">Contributing</h1>
  <p>
    This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor
    License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your
    contribution.
    For details, visit <a href="https://cla.microsoft.com">https://cla.microsoft.com</a>.
  </p>
  <p>
    When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and
    decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You
    will only need to do this once across all repos using our CLA.
  </p>
  <p>
    This project has adopted the <a href="https://opensource.microsoft.com/codeofconduct/">Microsoft Open Source Code of
      Conduct</a>. For more information see the Code of
    <a href="https://opensource.microsoft.com/codeofconduct/faq/">Conduct FAQ</a> or contact <a
      href="mailto:opencode@microsoft.com">opencode@microsoft.com</a> with any additional questions or comments.
  </p>
  <p>
    After getting familiar with contribution agreements, you are ready to create your first PR =), follow the NNI
    developer tutorials to get start:
  </p>
  <ul>
    <li>We recommend new contributors to start with simple issues: '<a
        href="https://github.com/Microsoft/nni/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22">good first
        issue</a>' or '<a
        href="https://github.com/microsoft/nni/issues?q=is%3Aopen+is%3Aissue+label%3A%22help+wanted%22">help-wanted</a>'.
    </li>
    <li><a href="{{ pathto('Tutorial/SetupNniDeveloperEnvironment') }}">NNI developer environment installation
        tutorial</a></li>
    <li><a href="{{ pathto('Tutorial/HowToDebug') }}">How to debug</a></li>
    <li>
      If you have any questions on usage, review <a href="{{ pathto('Tutorial/FAQ') }}">FAQ</a> first, if there are no
      relevant issues and answers to your
      question, try contact NNI dev team and users in
      <a
        href="https://gitter.im/Microsoft/nni?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge">Gitter</a>
      or
      <a href="https://github.com/microsoft/nni/issues/new/choose">File an issue</a> on GitHub.
    </li>
    <li><a href="{{ pathto('Tuner/CustomizeTuner') }}">Customize your own Tuner</a></li>
    <li><a href="{{ pathto('TrainingService/HowToImplementTrainingService') }}">Implement customized TrainingService</a>
    </li>
    <li><a href="{{ pathto('NAS/Advanced') }}">Implement a new NAS trainer on NNI</a></li>
    <li><a href="{{ pathto('Tuner/CustomizeAdvisor') }}">Customize your own Advisor</a></li>
  </ul>
</div>

<!-- External Repositories and References -->
<div>
  <h1 class="title">External Repositories and References</h1>
  <p>With authors' permission, we listed a set of NNI usage examples and relevant articles.</p>
  <ul>
    <h2>External Repositories</h2>
    <li>Run <a href="{{ pathto('NAS/ENAS') }}">ENAS</a> with NNI</li>
    <li>
      <a
        href="https://github.com/microsoft/nni/blob/master/examples/feature_engineering/auto-feature-engineering/README.md">Automatic
        Feature Engineering</a> with NNI
    </li>
    <li><a
        href="https://github.com/microsoft/recommenders/blob/master/notebooks/04_model_select_and_optimize/nni_surprise_svd.ipynb">Hyperparameter
        Tuning for Matrix Factorization</a> with NNI</li>
    <li><a href="https://github.com/ksachdeva/scikit-nni">scikit-nni</a> Hyper-parameter search for scikit-learn
      pipelines using NNI</li>
  </ul>

  <!-- Relevant Articles -->
  <ul>
    <h2>Relevant Articles</h2>
    <li><a href="{{ pathto('CommunitySharings/HpoComparision') }}">Hyper Parameter Optimization Comparison</a></li>
    <li><a href="{{ pathto('CommunitySharings/NasComparision') }}">Neural Architecture Search Comparison</a></li>
    <li><a href="{{ pathto('CommunitySharings/ParallelizingTpeSearch') }}">Parallelizing a Sequential Algorithm TPE</a>
    </li>
    <li><a href="{{ pathto('CommunitySharings/RecommendersSvd') }}">Automatically tuning SVD with NNI</a></li>
    <li><a href="{{ pathto('CommunitySharings/SptagAutoTune') }}">Automatically tuning SPTAG with NNI</a></li>
    <li><a
        href="https://towardsdatascience.com/find-thy-hyper-parameters-for-scikit-learn-pipelines-using-microsoft-nni-f1015b1224c1">
        Find thy hyper-parameters for scikit-learn pipelines using Microsoft NNI
      </a></li>
    <li>
      Blog (in Chinese) -
      <a
        href="http://gaocegege.com/Blog/%E6%9C%BA%E5%99%A8%E5%AD%A6%E4%B9%A0/katib-new#%E6%80%BB%E7%BB%93%E4%B8%8E%E5%88%86%E6%9E%90">
        AutoML tools (Advisor, NNI and Google Vizier) comparison
      </a>
      by @gaocegege - 总结与分析 section of design and implementation of kubeflow/katib
    </li>
    <li>
      Blog (in Chinese) -
      <a href="https://mp.weixin.qq.com/s/7_KRT-rRojQbNuJzkjFMuA">A summary of NNI new capabilities in 2019</a> by @squirrelsc
    </li>
  </ul>
</div>

<!-- feedback -->
<div>
  <h1 class="title">Feedback</h1>
  <ul>
    <li><a href="https://github.com/microsoft/nni/issues/new/choose">File an issue</a> on GitHub.</li>
    <li>Ask a question with NNI tags on <a
        href="https://stackoverflow.com/questions/tagged/nni?sort=Newest&edited=true">Stack Overflow</a>.
    </li>
    <li>Discuss on the NNI <a
        href="https://gitter.im/Microsoft/nni?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge">Gitter</a>
      in NNI.</li>
  </ul>
  <div>
    <div>Join IM discussion groups:</div>
    <table border=1 style="border-collapse: collapse;">
      <tbody>
        <tr style="line-height: 30px;">
          <th>Gitter</th>
          <td></td>
          <th>WeChat</th>
        </tr>
        <tr>
          <td class="QR">
            <img src="https://user-images.githubusercontent.com/39592018/80665738-e0574a80-8acc-11ea-91bc-0836dc4cbf89.png" alt="Gitter" />
          </td>
          <td width="80" align="center" class="or">OR</td>
          <td class="QR">
            <img src="https://github.com/scarlett2018/nniutil/raw/master/wechat.png" alt="NNI Wechat" />
          </td>
        </tr>
      </tbody>
    </table>
  </div>
</div>

<!-- Related Projects -->
<div>
  <h1 class="title">Related Projects</h1>
  <p>
    Targeting at openness and advancing state-of-art technology,
    <a href="https://www.microsoft.com/en-us/research/group/systems-research-group-asia/">Microsoft Research (MSR)</a>
    had also released few
    other open source projects.</p>
  <ul id="relatedProject">
    <li>
      <a href="https://github.com/Microsoft/pai">OpenPAI</a> : an open source platform that provides complete AI model
      training and resource management
      capabilities, it is easy to extend and supports on-premise,
      cloud and hybrid environments in various scale.
    </li>
    <li>
      <a href="https://github.com/Microsoft/frameworkcontroller">FrameworkController</a> : an open source
      general-purpose Kubernetes Pod Controller that orchestrate
      all kinds of applications on Kubernetes by a single controller.
    </li>
    <li>
      <a href="https://github.com/Microsoft/MMdnn">MMdnn</a> : A comprehensive, cross-framework solution to convert,
      visualize and diagnose deep neural network
      models. The "MM" in MMdnn stands for model management
      and "dnn" is an acronym for deep neural network.
    </li>
    <li>
      <a href="https://github.com/Microsoft/SPTAG">SPTAG</a> : Space Partition Tree And Graph (SPTAG) is an open
      source library
      for large scale vector approximate nearest neighbor search scenario.
    </li>
  </ul>
  <p>We encourage researchers and students leverage these projects to accelerate the AI development and research.</p>
</div>

<!-- License -->
<div>
  <h1 class="title">License</h1>
  <p>The entire codebase is under <a href="https://github.com/microsoft/nni/blob/master/LICENSE">MIT license</a></p>
</div>
</div>
{% endblock %}