**Getting Started with NNI** === NNI (Nerual Network Intelligance) is a toolkit to help users running automated machine learning experiment. The tool dispatchs and runs trail jobs that generated by tunning algorithms to search the best neural architecture and/or hyper-parameters at different enviroments (e.g. local, remote servers, Cloud). ``` AutoML experiment Training Services ┌────────┐ ┌────────────────────────┐ ┌────────────────┐ │ nnictl │ ─────> │ nni_manager │ │ Local Machine │ └────────┘ │ sdk/tuner │ └────────────────┘ │ hyperopt_tuner │ │ evlution_tuner │ trail jobs ┌────────────────┐ │ ... │ ────────> │ Remote Servers │ ├────────────────────────┤ └────────────────┘ │ trail job source code │ │ sdk/annotation │ ┌────────────────┐ ├────────────────────────┤ │ Yarn,K8s, │ │ nni_board │ │ ... │ └────────────────────────┘ └────────────────┘ ``` ## **Who should consider using NNI** * You want to try different AutoML algorithms for your training code (model) at local * You want to run AutoML trail jobs in different enviroments to speed up search (e.g. remote servers, Cloud) * As a reseacher and data scientist, you want to implement your own AutoML algorithms and compare with other algorithms * As a ML platform owner, you want to support AutoML in your platform ## **Setup** * __Dependencies__ nni requires: ``` python >= 3.5 node >= 10.9.0 yarn >= 1.9.4 ``` Before install nni, please make sure you have installed python environment correctly. * __User installation__ * clone nni repository git clone https://github.com/Microsoft/NeuralNetworkIntelligence * run install.sh cd NeuralNetworkIntelligence sh ./install.sh For more details about installation, please refer to [Installation instructions](Installation.md). ## **Quick start: run an experiment at local** Requirements: * local enviroment setup [TODO] Run the following command to create an experiemnt for [mnist] ```bash nnictl create --config /usr/share/nni/examples/trials/mnist-annotation/config.yml ``` This command will start the experiment and WebUI. The WebUI endpoint will be shown in the output of this command (for example, `http://localhost:8080`). Open this URL using your browsers. You can analyze your experiment through WebUI, or open trials' tensorboard. ## **Quick start: run a customized experiment** An experiment is to run multiple trial jobs, each trial job tries a configuration which includes a specific neural architecture (or model) and hyper-parameter values. To run an experiment through NNI, you should: * Provide a runnable trial * Provide or choose a tuner * Provide a yaml experiment configure file * (optional) Provide or choose an assessor **Prepare trial**: Let's use a simple trial example, e.g. mnist, provided by NNI. After you installed NNI, NNI examples have been put in /usr/share/nni/examples, run `ls /usr/share/nni/examples/trials` to see all the trial examples. You can simply execute the following command to run the NNI mnist example: python /usr/share/nni/examples/trials/mnist-annotation/mnist.py This command will be filled in the yaml configure file below. Please refer to [here]() for how to write your own trial. **Prepare tuner**: NNI supports several popular automl algorithms, including Random Search, Tree of Parzen Estimators (TPE), Bayesian Optimization etc. Users can write their own tuner (refer to [here]()), but for simplicity, here we can choose a tuner provided by NNI as below: tunerName: TPE optimizationMode: maximize *tunerName* is used to specify a tuner in NNI, *optimizationMode* is to indicate whether you want to maximize or minimize your trial's result. **Prepare configure file**: Since you have already known which trial code you are going to run and which tuner you are going to use, it is time to prepare the yaml configure file. NNI provides a demo configure file for each trial example, `cat /usr/share/nni/examples/trials/mnist-annotation/config.yml` to see it. Its content is basically shown below: ``` authorName: your_name experimentName: auto_mnist # how many trials could be concurrently running trialConcurrency: 2 # maximum experiment running duration maxExecDuration: 3h # empty means never stop maxTrialNum: 100 # choice: local, remote trainingServicePlatform: local # choice: true, false useAnnotation: true tuner: tunerName: TPE optimizationMode: Maximize trial: trialCommand: python mnist.py trialCodeDir: /usr/share/nni/examples/trials/mnist-annotation trialGpuNum: 0 ``` Here *useAnnotation* is true because this trial example uses our python annotation (refer to [here]() for details). For trial, we should provide *trialCommand* which is the command to run the trial, provide *trialCodeDir* where the trial code is. The command will be executed in this directory. We should also provide how many GPUs a trial requires. With all these steps done, we can run the experiment with the following command: nnictl create --config /usr/share/nni/examples/trials/mnist-annotation/config.yml You can refer to [here](NNICTLDOC.md) for more usage guide of *nnictl* command line tool. ## View experiment results The experiment has been running now, NNI provides WebUI for you to view experiment progress, to control your experiment, and some other appealing features. The WebUI is opened by default by `nnictl create`. ## Further reading * [How to write a trial running on NNI (Mnist as an example)?](WriteYourTrial.md) * [Tutorial of NNI python annotation.](../tools/annotation/README.md) * [Tuners supported by NNI.](../src/sdk/pynni/nni/README.md) * [How to enable early stop (i.e. assessor) in an experiment?](EnableAssessor.md) * [How to run an experiment on multiple machines?](RemoteMachineMode.md) * [How to write a customized tuner?](../examples/tuners/README.md) * [How to write a customized assessor?](../examples/assessors/README.md) * [How to resume an experiment?]() * [Tutorial of the command tool *nnictl*.](NNICTLDOC.md) * [How to use *nnictl* to control multiple experiments?]() ## How to contribute TBD