Merge pull request #222 from microsoft/master

merge master

azure-pipelines.yml

@@ -22,8 +22,8 @@ jobs:
       cd src/nni_manager
       yarn eslint
       # uncomment following 2 lines to enable webui eslint
-      # cd ../webui
-      # yarn eslint
+      cd ../webui
+      yarn eslint
     displayName: 'Run eslint'
   - script: |
       python3 -m pip install torch==0.4.1 --user

docs/en_US/Assessor/CustomizeAssessor.md

@@ -54,6 +54,8 @@ assessor:
 
 Please noted in **2**. The object `trial_history` are exact the object that Trial send to Assessor by using SDK `report_intermediate_result` function.
 
+The working directory of your assessor is `<home>/nni/experiments/<experiment_id>/log`, which can be retrieved with environment variable `NNI_LOG_DIRECTORY`,
+
 More detail example you could see:
 > * [medianstop-assessor](https://github.com/Microsoft/nni/tree/master/src/sdk/pynni/nni/medianstop_assessor)
 > * [curvefitting-assessor](https://github.com/Microsoft/nni/tree/master/src/sdk/pynni/nni/curvefitting_assessor)
\ No newline at end of file

docs/en_US/TrainingService/PaiMode.md

@@ -34,10 +34,13 @@ trial:
   cpuNum: 1
   memoryMB: 8196
   image: msranni/nni:latest
+  virtualCluster: default
+  nniManagerNFSMountPath: /home/user/mnt
+  containerNFSMountPath: /mnt/data/user
 # Configuration to access OpenPAI Cluster
 paiConfig:
   userName: your_pai_nni_user
-  passWord: your_pai_password
+  token: your_pai_token
   host: 10.1.1.1
 ```
 
@@ -53,52 +56,13 @@ Compared with [LocalMode](LocalMode.md) and [RemoteMachineMode](RemoteMachineMod
     * We already build a docker image [nnimsra/nni](https://hub.docker.com/r/msranni/nni/) on [Docker Hub](https://hub.docker.com/). It contains NNI python packages, Node modules and javascript artifact files required to start experiment, and all of NNI dependencies. The docker file used to build this image can be found at [here](https://github.com/Microsoft/nni/tree/master/deployment/docker/Dockerfile). You can either use this image directly in your config file, or build your own image based on it.
 * virtualCluster
     * Optional key. Set the virtualCluster of OpenPAI. If omitted, the job will run on default virtual cluster.
-* shmMB
-    * Optional key. Set the shmMB configuration of OpenPAI, it set the shared memory for one task in the task role.
-* authFile
-    * Optional key, Set the auth file path for private registry while using PAI mode, [Refer](https://github.com/microsoft/pai/blob/2ea69b45faa018662bc164ed7733f6fdbb4c42b3/docs/faq.md#q-how-to-use-private-docker-registry-job-image-when-submitting-an-openpai-job), you can prepare the authFile and simply provide the local path of this file, NNI will upload this file to HDFS for you.
-* portList  
-    * Optional key. Set the portList configuration of OpenPAI, it specifies a list of port used in container, [Refer](https://github.com/microsoft/pai/blob/b2324866d0280a2d22958717ea6025740f71b9f0/docs/job_tutorial.md#specification).  
-    The config schema in NNI is shown below:
-    ```
-    portList:
-      - label: test
-        beginAt: 8080
-        portNumber: 2
-    ``` 
-    Let's say you want to launch a tensorboard in the mnist example using the port. So the first step is to write a wrapper script `launch_pai.sh` of `mnist.py`.
-
-    ```bash
-    export TENSORBOARD_PORT=PAI_PORT_LIST_${PAI_CURRENT_TASK_ROLE_NAME}_0_tensorboard
-    tensorboard --logdir . --port ${!TENSORBOARD_PORT} &
-    python3 mnist.py
-    ```
-    The config file of portList should be filled as following:
+* nniManagerNFSMountPath
+    * Required key. Set the mount path in your nniManager machine.
+* containerNFSMountPath
+    * Required key. Set the mount path in your container used in PAI.
+* paiStoragePlugin
+    * Required key. Set the storage plugin name used in PAI.
 
-    ```yaml
-  trial:
-    command: bash launch_pai.sh
-    portList:
-      - label: tensorboard
-        beginAt: 0
-        portNumber: 1
-    ```
-
-NNI support two kind of authorization method in PAI, including password and PAI token, [refer](https://github.com/microsoft/pai/blob/b6bd2ab1c8890f91b7ac5859743274d2aa923c22/docs/rest-server/API.md#2-authentication). The authorization is configured in `paiConfig` field.  
-For password authorization, the `paiConfig` schema is:
-```
-paiConfig:
-  userName: your_pai_nni_user
-  passWord: your_pai_password
-  host: 10.1.1.1
-```  
-For pai token authorization, the `paiConfig` schema is:
-```
-paiConfig:
-  userName: your_pai_nni_user
-  token: your_pai_token
-  host: 10.1.1.1
-```
 
 Once complete to fill NNI experiment config file and save (for example, save as exp_pai.yml), then run the following command
 ```
@@ -121,9 +85,7 @@ And you will be redirected to HDFS web portal to browse the output files of that
 You can see there're three fils in output folder: stderr, stdout, and trial.log
 
 ## data management
-If your training data is not too large, it could be put into codeDir, and nni will upload the data to hdfs, or you could build your own docker image with the data. If you have large dataset, it's not appropriate to put the data in codeDir, and you could follow the [guidance](https://github.com/microsoft/pai/blob/master/docs/user/storage.md) to mount the data folder in container.
-
-If you also want to save trial's other output into HDFS, like model files, you can use environment variable `NNI_OUTPUT_DIR` in your trial code to save your own output files, and NNI SDK will copy all the files in `NNI_OUTPUT_DIR` from trial's container to HDFS, the target path is `hdfs://host:port/{username}/nni/{experiments}/{experimentId}/trials/{trialId}/nnioutput`
+Befour using NNI to start your experiment, users should set the corresponding mount data path in your nniManager machine. PAI has their own storage(NFS, AzureBlob ...), and the storage will used in PAI will be mounted to the container when it start a job. Users should set the PAI storage type by `paiStoragePlugin` field to choose a storage in PAI. Then users should mount the storage to their nniManager machine, and set the `nniManagerNFSMountPath` field in configuration file, NNI will generate bash files and copy data in `codeDir` to the `nniManagerNFSMountPath` folder, then NNI will start a trial job. The data in `nniManagerNFSMountPath` will be sync to PAI storage, and will be mounted to PAI's container. The data path in container is set in `containerNFSMountPath`, NNI will enter this folder first, and then run scripts to start a trial job. 
 
 ## version check
 NNI support version check feature in since version 0.6. It is a policy to insure the version of NNIManager is consistent with trialKeeper, and avoid errors caused by version incompatibility.

docs/en_US/TrainingService/PaiYarnMode.md

+**Run an Experiment on OpenpaiYarn**
+===
+The original `pai` mode is modificated to `paiYarn` mode, which is a distributed training platform based on Yarn.
+
+## Setup environment
+Install NNI, follow the install guide [here](../Tutorial/QuickStart.md).
+
+## Run an experiment
+Use `examples/trials/mnist-annotation` as an example. The NNI config YAML file's content is like:
+
+```yaml
+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, pai, paiYarn
+trainingServicePlatform: paiYarn
+# search space file
+searchSpacePath: search_space.json
+# choice: true, false
+useAnnotation: true
+tuner:
+  builtinTunerName: TPE
+  classArgs:
+    optimize_mode: maximize
+trial:
+  command: python3 mnist.py
+  codeDir: ~/nni/examples/trials/mnist-annotation
+  gpuNum: 0
+  cpuNum: 1
+  memoryMB: 8196
+  image: msranni/nni:latest
+# Configuration to access OpenpaiYarn Cluster
+paiYarnConfig:
+  userName: your_paiYarn_nni_user
+  passWord: your_paiYarn_password
+  host: 10.1.1.1
+```
+
+Note: You should set `trainingServicePlatform: paiYarn` in NNI config YAML file if you want to start experiment in paiYarn mode.
+
+Compared with [LocalMode](LocalMode.md) and [RemoteMachineMode](RemoteMachineMode.md), trial configuration in paiYarn mode have these additional keys:
+* cpuNum
+    * Required key. Should be positive number based on your trial program's CPU  requirement
+* memoryMB
+    * Required key. Should be positive number based on your trial program's memory requirement
+* image
+    * Required key. In paiYarn mode, your trial program will be scheduled by OpenpaiYarn to run in [Docker container](https://www.docker.com/). This key is used to specify the Docker image used to create the container in which your trial will run.
+    * We already build a docker image [nnimsra/nni](https://hub.docker.com/r/msranni/nni/) on [Docker Hub](https://hub.docker.com/). It contains NNI python packages, Node modules and javascript artifact files required to start experiment, and all of NNI dependencies. The docker file used to build this image can be found at [here](https://github.com/Microsoft/nni/tree/master/deployment/docker/Dockerfile). You can either use this image directly in your config file, or build your own image based on it.
+* virtualCluster
+    * Optional key. Set the virtualCluster of OpenpaiYarn. If omitted, the job will run on default virtual cluster.
+* shmMB
+    * Optional key. Set the shmMB configuration of OpenpaiYarn, it set the shared memory for one task in the task role.
+* authFile
+    * Optional key, Set the auth file path for private registry while using paiYarn mode, [Refer](https://github.com/microsoft/paiYarn/blob/2ea69b45faa018662bc164ed7733f6fdbb4c42b3/docs/faq.md#q-how-to-use-private-docker-registry-job-image-when-submitting-an-openpaiYarn-job), you can prepare the authFile and simply provide the local path of this file, NNI will upload this file to HDFS for you.
+* portList  
+    * Optional key. Set the portList configuration of OpenpaiYarn, it specifies a list of port used in container, [Refer](https://github.com/microsoft/paiYarn/blob/b2324866d0280a2d22958717ea6025740f71b9f0/docs/job_tutorial.md#specification).  
+    The config schema in NNI is shown below:
+    ```
+    portList:
+      - label: test
+        beginAt: 8080
+        portNumber: 2
+    ``` 
+    Let's say you want to launch a tensorboard in the mnist example using the port. So the first step is to write a wrapper script `launch_paiYarn.sh` of `mnist.py`.
+
+    ```bash
+    export TENSORBOARD_PORT=paiYarn_PORT_LIST_${paiYarn_CURRENT_TASK_ROLE_NAME}_0_tensorboard
+    tensorboard --logdir . --port ${!TENSORBOARD_PORT} &
+    python3 mnist.py
+    ```
+    The config file of portList should be filled as following:
+
+    ```yaml
+  trial:
+    command: bash launch_paiYarn.sh
+    portList:
+      - label: tensorboard
+        beginAt: 0
+        portNumber: 1
+    ```
+
+NNI support two kind of authorization method in paiYarn, including password and paiYarn token, [refer](https://github.com/microsoft/paiYarn/blob/b6bd2ab1c8890f91b7ac5859743274d2aa923c22/docs/rest-server/API.md#2-authentication). The authorization is configured in `paiYarnConfig` field.  
+For password authorization, the `paiYarnConfig` schema is:
+```
+paiYarnConfig:
+  userName: your_paiYarn_nni_user
+  passWord: your_paiYarn_password
+  host: 10.1.1.1
+```  
+For paiYarn token authorization, the `paiYarnConfig` schema is:
+```
+paiYarnConfig:
+  userName: your_paiYarn_nni_user
+  token: your_paiYarn_token
+  host: 10.1.1.1
+```
+
+Once complete to fill NNI experiment config file and save (for example, save as exp_paiYarn.yml), then run the following command
+```
+nnictl create --config exp_paiYarn.yml
+```
+to start the experiment in paiYarn mode. NNI will create OpenpaiYarn job for each trial, and the job name format is something like `nni_exp_{experiment_id}_trial_{trial_id}`.
+You can see jobs created by NNI in the OpenpaiYarn cluster's web portal, like:
+![](../../img/nni_paiYarn_joblist.jpg)
+
+Notice: In paiYarn mode, NNIManager will start a rest server and listen on a port which is your NNI WebUI's port plus 1. For example, if your WebUI port is `8080`, the rest server will listen on `8081`, to receive metrics from trial job running in Kubernetes. So you should `enable 8081` TCP port in your firewall rule to allow incoming traffic.
+
+Once a trial job is completed, you can goto NNI WebUI's overview page (like http://localhost:8080/oview) to check trial's information.
+
+Expand a trial information in trial list view, click the logPath link like:
+![](../../img/nni_webui_joblist.jpg)
+
+And you will be redirected to HDFS web portal to browse the output files of that trial in HDFS:
+![](../../img/nni_trial_hdfs_output.jpg)
+
+You can see there're three fils in output folder: stderr, stdout, and trial.log
+
+## data management
+If your training data is not too large, it could be put into codeDir, and nni will upload the data to hdfs, or you could build your own docker image with the data. If you have large dataset, it's not appropriate to put the data in codeDir, and you could follow the [guidance](https://github.com/microsoft/paiYarn/blob/master/docs/user/storage.md) to mount the data folder in container.
+
+If you also want to save trial's other output into HDFS, like model files, you can use environment variable `NNI_OUTPUT_DIR` in your trial code to save your own output files, and NNI SDK will copy all the files in `NNI_OUTPUT_DIR` from trial's container to HDFS, the target path is `hdfs://host:port/{username}/nni/{experiments}/{experimentId}/trials/{trialId}/nnioutput`
+
+## version check
+NNI support version check feature in since version 0.6. It is a policy to insure the version of NNIManager is consistent with trialKeeper, and avoid errors caused by version incompatibility.
+Check policy:
+1. NNIManager before v0.6 could run any version of trialKeeper, trialKeeper support backward compatibility.
+2. Since version 0.6, NNIManager version should keep same with triakKeeper version. For example, if NNIManager version is 0.6, trialKeeper version should be 0.6 too.
+3. Note that the version check feature only check first two digits of version.For example, NNIManager v0.6.1 could use trialKeeper v0.6 or trialKeeper v0.6.2, but could not use trialKeeper v0.5.1 or trialKeeper v0.7.
+
+If you could not run your experiment and want to know if it is caused by version check, you could check your webUI, and there will be an error message about version check.
+![](../../img/version_check.png)

docs/en_US/Tuner/BuiltinTuner.md

@@ -160,6 +160,7 @@ Similar to TPE, SMAC is also a black-box tuner which can be tried in various sce
 **Requirement of classArgs**
 
 * **optimize_mode** (*maximize or minimize, optional, default = maximize*) - If 'maximize', the tuner will target to maximize metrics. If 'minimize', the tuner will target to minimize metrics.
+* **config_dedup** (*True or False, optional, default = False*) - If True, the tuner will not generate a configuration that has been already generated. If False, a configuration may be generated twice, but it is rare for relatively large search space.
 
 **Usage example**
 

docs/en_US/Tuner/CustomizeAdvisor.md

@@ -33,6 +33,8 @@ advisor:
     arg1: value1
 ```
 
+**Note that** The working directory of your advisor is `<home>/nni/experiments/<experiment_id>/log`, which can be retrieved with environment variable `NNI_LOG_DIRECTORY`.
+
 ## Example
 
 Here we provide an [example](https://github.com/microsoft/nni/tree/master/examples/tuners/mnist_keras_customized_advisor).

docs/en_US/Tuner/CustomizeTuner.md

@@ -77,7 +77,7 @@ parameters = {"dropout": 0.3, "learning_rate": 0.4}
 value = 0.93
 ```
 
-**Note that** if you want to access a file (e.g., `data.txt`) in the directory of your own tuner, you cannot use `open('data.txt', 'r')`. Instead, you should use the following:
+**Note that** The working directory of your tuner is `<home>/nni/experiments/<experiment_id>/log`, which can be retrieved with environment variable `NNI_LOG_DIRECTORY`, therefore, if you want to access a file (e.g., `data.txt`) in the directory of your own tuner, you cannot use `open('data.txt', 'r')`. Instead, you should use the following:
 
 ```python
 _pwd = os.path.dirname(__file__)

examples/model_compress/QAT_torch_quantizer.py

@@ -35,7 +35,6 @@ def train(model, quantizer, device, train_loader, optimizer):
         loss = F.nll_loss(output, target)
         loss.backward()
         optimizer.step()
-        quantizer.step()
         if batch_idx % 100 == 0:
             print('{:2.0f}%  Loss {}'.format(100 * batch_idx / len(train_loader), loss.item()))
 

src/nni_manager/common/manager.ts

@@ -27,7 +27,7 @@ interface ExperimentParams {
     versionCheck?: boolean;
     logCollection?: string;
     tuner?: {
-        className: string;
+        className?: string;
         builtinTunerName?: string;
         codeDir?: string;
         classArgs?: any;
@@ -37,7 +37,7 @@ interface ExperimentParams {
         gpuIndices?: string;
     };
     assessor?: {
-        className: string;
+        className?: string;
         builtinAssessorName?: string;
         codeDir?: string;
         classArgs?: any;
@@ -45,7 +45,7 @@ interface ExperimentParams {
         checkpointDir: string;
     };
     advisor?: {
-        className: string;
+        className?: string;
         builtinAdvisorName?: string;
         codeDir?: string;
         classArgs?: any;

src/nni_manager/common/utils.ts

@@ -17,7 +17,7 @@ import * as util from 'util';
 
 import { Database, DataStore } from './datastore';
 import { ExperimentStartupInfo, getExperimentStartupInfo, setExperimentStartupInfo } from './experimentStartupInfo';
-import { Manager } from './manager';
+import { ExperimentParams, Manager } from './manager';
 import { HyperParameters, TrainingService, TrialJobStatus } from './trainingService';
 
 function getExperimentRootDir(): string {
@@ -130,15 +130,6 @@ function parseArg(names: string[]): string {
     return '';
 }
 
-function encodeCmdLineArgs(args: any): any {
-    if(process.platform === 'win32'){
-        return JSON.stringify(args);
-    }
-    else{
-        return JSON.stringify(JSON.stringify(args));
-    }
-}
-
 function getCmdPy(): string {
     let cmd = 'python3';
     if(process.platform === 'win32'){
@@ -150,83 +141,14 @@ function getCmdPy(): string {
 /**
  * Generate command line to start automl algorithm(s),
  * either start advisor or start a process which runs tuner and assessor
- * @param tuner : For builtin tuner:
- *     {
- *         className: 'EvolutionTuner'
- *         classArgs: {
- *             optimize_mode: 'maximize',
- *             population_size: 3
- *         }
- *     }
- * customized:
- *     {
- *         codeDir: '/tmp/mytuner'
- *         classFile: 'best_tuner.py'
- *         className: 'BestTuner'
- *         classArgs: {
- *             optimize_mode: 'maximize',
- *             population_size: 3
- *         }
- *     }
  *
- * @param assessor: similiar as tuner
- * @param advisor: similar as tuner
+ * @param expParams: experiment startup parameters
  *
  */
-function getMsgDispatcherCommand(tuner: any, assessor: any, advisor: any, multiPhase: boolean = false, multiThread: boolean = false): string {
-    if ((tuner || assessor) && advisor) {
-        throw new Error('Error: specify both tuner/assessor and advisor is not allowed');
-    }
-    if (!tuner && !advisor) {
-        throw new Error('Error: specify neither tuner nor advisor is not allowed');
-    }
-    let command: string = `${getCmdPy()} -m nni`;
-    if (multiPhase) {
-        command += ' --multi_phase';
-    }
-
-    if (multiThread) {
-        command += ' --multi_thread';
-    }
-
-    if (advisor) {
-        command += ` --advisor_class_name ${advisor.className}`;
-        if (advisor.classArgs !== undefined) {
-            command += ` --advisor_args ${encodeCmdLineArgs(advisor.classArgs)}`;
-        }
-        if (advisor.codeDir !== undefined && advisor.codeDir.length > 1) {
-            command += ` --advisor_directory ${advisor.codeDir}`;
-        }
-        if (advisor.classFileName !== undefined && advisor.classFileName.length > 1) {
-            command += ` --advisor_class_filename ${advisor.classFileName}`;
-        }
-    } else {
-        command += ` --tuner_class_name ${tuner.className}`;
-        if (tuner.classArgs !== undefined) {
-            command += ` --tuner_args ${encodeCmdLineArgs(tuner.classArgs)}`;
-        }
-        if (tuner.codeDir !== undefined && tuner.codeDir.length > 1) {
-            command += ` --tuner_directory ${tuner.codeDir}`;
-        }
-        if (tuner.classFileName !== undefined && tuner.classFileName.length > 1) {
-            command += ` --tuner_class_filename ${tuner.classFileName}`;
-        }
-
-        if (assessor !== undefined && assessor.className !== undefined) {
-            command += ` --assessor_class_name ${assessor.className}`;
-            if (assessor.classArgs !== undefined) {
-                command += ` --assessor_args ${encodeCmdLineArgs(assessor.classArgs)}`;
-            }
-            if (assessor.codeDir !== undefined && assessor.codeDir.length > 1) {
-                command += ` --assessor_directory ${assessor.codeDir}`;
-            }
-            if (assessor.classFileName !== undefined && assessor.classFileName.length > 1) {
-                command += ` --assessor_class_filename ${assessor.classFileName}`;
-            }
-        }
-    }
-
-    return command;
+function getMsgDispatcherCommand(expParams: ExperimentParams): string {
+    const clonedParams = Object.assign({}, expParams);
+    delete clonedParams.searchSpace;
+    return `${getCmdPy()} -m nni --exp_params ${Buffer.from(JSON.stringify(clonedParams)).toString('base64')}`;
 }
 
 /**

src/nni_manager/config/kubeflow/tfjob-crd-v1.json

+{
+  "kind": "CustomResourceDefinition", 
+  "spec": {
+    "scope": "Namespaced", 
+    "version": "v1", 
+    "group": "kubeflow.org", 
+    "names": {
+      "kind": "TFJob", 
+      "plural": "tfjobs", 
+      "singular": "tfjob"
+    }
+  }, 
+  "apiVersion": "apiextensions.k8s.io/v1beta1",
+  "metadata": {
+    "name": "tfjobs.kubeflow.org"
+  }
+}

src/nni_manager/core/nnimanager.ts

@@ -170,8 +170,7 @@ class NNIManager implements Manager {
             this.trainingService.setClusterMetadata('log_collection', expParams.logCollection.toString());
         }
 
-        const dispatcherCommand: string = getMsgDispatcherCommand(expParams.tuner, expParams.assessor, expParams.advisor,
-            expParams.multiPhase, expParams.multiThread);
+        const dispatcherCommand: string = getMsgDispatcherCommand(expParams);
         this.log.debug(`dispatcher command: ${dispatcherCommand}`);
         const checkpointDir: string = await this.createCheckpointDir();
         this.setupTuner(
@@ -211,8 +210,7 @@ class NNIManager implements Manager {
             this.trainingService.setClusterMetadata('version_check', expParams.versionCheck.toString());
         }
 
-        const dispatcherCommand: string = getMsgDispatcherCommand(expParams.tuner, expParams.assessor, expParams.advisor,
-            expParams.multiPhase, expParams.multiThread);
+        const dispatcherCommand: string = getMsgDispatcherCommand(expParams);
         this.log.debug(`dispatcher command: ${dispatcherCommand}`);
         const checkpointDir: string = await this.createCheckpointDir();
         this.setupTuner(
@@ -361,6 +359,7 @@ class NNIManager implements Manager {
         }
 
         const nniEnv = {
+            SDK_PROCESS: 'dispatcher',
             NNI_MODE: mode,
             NNI_CHECKPOINT_DIRECTORY: dataDirectory,
             NNI_LOG_DIRECTORY: getLogDir(),

src/nni_manager/core/test/ipcInterfaceTerminate.test.ts

@@ -21,18 +21,26 @@ function startProcess(): void {
     const dispatcherCmd: string = getMsgDispatcherCommand(
         // Mock tuner config
         {
-            className: 'DummyTuner',
-            codeDir: './',
-            classFileName: 'dummy_tuner.py'
-        },
-        // Mock assessor config
-        {
-            className: 'DummyAssessor',
-            codeDir: './',
-            classFileName: 'dummy_assessor.py'
-        },
-        // advisor
-        undefined
+            experimentName: 'exp1',
+            maxExecDuration: 3600,
+            searchSpace: '',
+            trainingServicePlatform: 'local',
+            authorName: '',
+            trialConcurrency: 1,
+            maxTrialNum: 5,
+            tuner: {
+                className: 'DummyTuner',
+                codeDir: './',
+                classFileName: 'dummy_tuner.py',
+                checkpointDir: './'
+            },
+            assessor: {
+                className: 'DummyAssessor',
+                codeDir: './',
+                classFileName: 'dummy_assessor.py',
+                checkpointDir: './'
+            }
+        }
     );
     const proc: ChildProcess = getTunerProc(dispatcherCmd, stdio,  'core/test', process.env);
     proc.on('error', (error: Error): void => {

src/nni_manager/core/test/nnimanager.test.ts

@@ -42,9 +42,9 @@ describe('Unit test for nnimanager', function () {
         maxExecDuration: 5,
         maxTrialNum: 3,
         trainingServicePlatform: 'local',
-        searchSpace: '{"x":1}',
+        searchSpace: '{"lr": {"_type": "choice", "_value": [0.01,0.001]}}',
         tuner: {
-            className: 'TPE',
+            builtinTunerName: 'TPE',
             classArgs: {
                 optimize_mode: 'maximize'
             },
@@ -52,7 +52,7 @@ describe('Unit test for nnimanager', function () {
             gpuNum: 0
         },
         assessor: {
-            className: 'Medianstop',
+            builtinAssessorName: 'Medianstop',
             checkpointDir: '',
             gpuNum: 1
         }
@@ -65,9 +65,9 @@ describe('Unit test for nnimanager', function () {
         maxExecDuration: 6,
         maxTrialNum: 2,
         trainingServicePlatform: 'local',
-        searchSpace: '{"y":2}',
+        searchSpace: '{"lr": {"_type": "choice", "_value": [0.01,0.001]}}',
         tuner: {
-            className: 'TPE',
+            builtinTunerName: 'TPE',
             classArgs: {
                 optimize_mode: 'maximize'
             },
@@ -75,7 +75,7 @@ describe('Unit test for nnimanager', function () {
             gpuNum: 0
         },
         assessor: {
-            className: 'Medianstop',
+            builtinAssessorName: 'Medianstop',
             checkpointDir: '',
             gpuNum: 1
         }
@@ -198,7 +198,7 @@ describe('Unit test for nnimanager', function () {
     it('test updateExperimentProfile SEARCH_SPACE',  () => {
         return nniManager.updateExperimentProfile(experimentProfile, 'SEARCH_SPACE').then(() => {
             nniManager.getExperimentProfile().then((updateProfile) => {
-                expect(updateProfile.params.searchSpace).to.be.equal('{"y":2}');
+                expect(updateProfile.params.searchSpace).to.be.equal('{"lr": {"_type": "choice", "_value": [0.01,0.001]}}');
             });
         }).catch((error) => {
             assert.fail(error);

src/nni_manager/main.ts

@@ -20,7 +20,8 @@ import { NNIRestServer } from './rest_server/nniRestServer';
 import { FrameworkControllerTrainingService } from './training_service/kubernetes/frameworkcontroller/frameworkcontrollerTrainingService';
 import { KubeflowTrainingService } from './training_service/kubernetes/kubeflow/kubeflowTrainingService';
 import { LocalTrainingService } from './training_service/local/localTrainingService';
-import { PAITrainingService } from './training_service/pai/paiTrainingService';
+import { PAIK8STrainingService } from './training_service/pai/paiK8S/paiK8STrainingService';
+import { PAIYarnTrainingService } from './training_service/pai/paiYarn/paiYarnTrainingService';
 import {
     RemoteMachineTrainingService
 } from './training_service/remote_machine/remoteMachineTrainingService';
@@ -44,7 +45,11 @@ async function initContainer(platformMode: string, logFileName?: string): Promis
             .scope(Scope.Singleton);
     } else if (platformMode === 'pai') {
         Container.bind(TrainingService)
-            .to(PAITrainingService)
+            .to(PAIK8STrainingService)
+            .scope(Scope.Singleton);
+    } else if (platformMode === 'paiYarn') {
+            Container.bind(TrainingService)
+            .to(PAIYarnTrainingService)
             .scope(Scope.Singleton);
     } else if (platformMode === 'kubeflow') {
         Container.bind(TrainingService)
@@ -76,7 +81,7 @@ async function initContainer(platformMode: string, logFileName?: string): Promis
 
 function usage(): void {
     console.info('usage: node main.js --port <port> --mode \
-    <local/remote/pai/kubeflow/frameworkcontroller> --start_mode <new/resume> --experiment_id <id>');
+    <local/remote/pai/kubeflow/frameworkcontroller/paiYarn> --start_mode <new/resume> --experiment_id <id>');
 }
 
 const strPort: string = parseArg(['--port', '-p']);
@@ -88,7 +93,7 @@ if (!strPort || strPort.length === 0) {
 const port: number = parseInt(strPort, 10);
 
 const mode: string = parseArg(['--mode', '-m']);
-if (!['local', 'remote', 'pai', 'kubeflow', 'frameworkcontroller'].includes(mode)) {
+if (!['local', 'remote', 'pai', 'kubeflow', 'frameworkcontroller', 'paiYarn'].includes(mode)) {
     console.log(`FATAL: unknown mode: ${mode}`);
     usage();
     process.exit(1);

src/nni_manager/rest_server/restValidationSchemas.ts

@@ -36,6 +36,9 @@ export namespace ValidationSchemas {
                 virtualCluster: joi.string(),
                 shmMB: joi.number(),
                 authFile: joi.string(),
+                nniManagerNFSMountPath: joi.string().min(1),
+                containerNFSMountPath: joi.string().min(1),
+                paiStoragePlugin: joi.string().min(1),
                 nasMode: joi.string().valid('classic_mode', 'enas_mode', 'oneshot_mode', 'darts_mode'),
                 portList: joi.array().items(joi.object({
                     label: joi.string().required(),
@@ -89,12 +92,18 @@ export namespace ValidationSchemas {
                     })
                 })
             }),
-            pai_config: joi.object({ // eslint-disable-line @typescript-eslint/camelcase
+            pai_yarn_config: joi.object({ // eslint-disable-line @typescript-eslint/camelcase
                 userName: joi.string().min(1).required(),
                 passWord: joi.string().min(1),
                 token: joi.string().min(1),
                 host: joi.string().min(1).required()
             }),
+            pai_config: joi.object({ // eslint-disable-line @typescript-eslint/camelcase
+                userName: joi.string().min(1).required(),
+                passWord: joi.string().min(1),
+                token: joi.string().min(1),
+                host: joi.string().min(1).required(),
+            }),
             kubeflow_config: joi.object({ // eslint-disable-line @typescript-eslint/camelcase
                 operator: joi.string().min(1).required(),
                 storage: joi.string().min(1),

src/nni_manager/training_service/common/trialConfigMetadataKey.ts

@@ -13,6 +13,7 @@ export enum TrialConfigMetadataKey {
     EXPERIMENT_ID = 'experimentId',
     MULTI_PHASE = 'multiPhase',
     RANDOM_SCHEDULER = 'random_scheduler',
+    PAI_YARN_CLUSTER_CONFIG = 'pai_yarn_config',
     PAI_CLUSTER_CONFIG = 'pai_config',
     KUBEFLOW_CLUSTER_CONFIG = 'kubeflow_config',
     NNI_MANAGER_IP = 'nni_manager_ip',

src/nni_manager/training_service/kubernetes/kubeflow/kubeflowApiClient.ts

@@ -65,6 +65,25 @@ class TFOperatorClientV1Beta2 extends KubernetesCRDClient {
     }
 }
 
+class TFOperatorClientV1 extends KubernetesCRDClient {
+    /**
+     * constructor, to initialize tfjob CRD definition
+     */
+    public constructor() {
+        super();
+        this.crdSchema = JSON.parse(fs.readFileSync('./config/kubeflow/tfjob-crd-v1.json', 'utf8'));
+        this.client.addCustomResourceDefinition(this.crdSchema);
+    }
+
+    protected get operator(): any {
+        return this.client.apis['kubeflow.org'].v1.namespaces('default').tfjobs;
+    }
+
+    public get containerName(): string {
+        return 'tensorflow';
+    }
+}
+
 class PyTorchOperatorClientV1Alpha2 extends KubernetesCRDClient {
     /**
      * constructor, to initialize tfjob CRD definition
@@ -142,6 +161,9 @@ class KubeflowOperatorClientFactory {
                     case 'v1beta2': {
                         return new TFOperatorClientV1Beta2();
                     }
+                    case 'v1': {
+                        return new TFOperatorClientV1();
+                    }
                     default:
                         throw new Error(`Invalid tf-operator apiVersion ${operatorApiVersion}`);
                 }

src/nni_manager/training_service/pai/paiConfig.ts

@@ -4,91 +4,8 @@
 'use strict';
 
 import {TrialConfig} from '../common/trialConfig';
+import { TrialJobApplicationForm, TrialJobDetail, TrialJobStatus  } from '../../common/trainingService';
 
-/**
- * Task role for PAI
- */
-export class PAITaskRole {
-    // Name for the task role
-    public readonly name: string;
-    // Number of tasks for the task role, no less than 1
-    public readonly taskNumber: number;
-    // CPU number for one task in the task role, no less than 1
-    public readonly cpuNumber: number;
-    // Memory for one task in the task role, no less than 100
-    public readonly memoryMB: number;
-    // GPU number for one task in the task role, no less than 0
-    public readonly gpuNumber: number;
-    // Executable command for tasks in the task role, can not be empty
-    public readonly command: string;
-    //Shared memory for one task in the task role
-    public readonly shmMB?: number;
-    //portList to specify the port used in container
-    public portList?: PortListMetaData[];
-
-    /**
-     * Constructor
-     * @param name Name for the task role
-     * @param taskNumber Number of tasks for the task role, no less than 1
-     * @param cpuNumber CPU number for one task in the task role, no less than 1
-     * @param memoryMB Memory for one task in the task role, no less than 100
-     * @param gpuNumber GPU number for one task in the task role, no less than 0
-     * @param command Executable command for tasks in the task role, can not be empty
-     */
-    constructor(name: string, taskNumber: number, cpuNumber: number, memoryMB: number, gpuNumber: number,
-                command: string, shmMB?: number, portList?: PortListMetaData[]) {
-        this.name = name;
-        this.taskNumber = taskNumber;
-        this.cpuNumber = cpuNumber;
-        this.memoryMB = memoryMB;
-        this.gpuNumber = gpuNumber;
-        this.command = command;
-        this.shmMB = shmMB;
-        this.portList = portList;
-    }
-}
-
-/**
- * Trial job configuration submitted to PAI
- */
-export class PAIJobConfig {
-    // Name for the job, need to be unique
-    public readonly jobName: string;
-    // URL pointing to the Docker image for all tasks in the job
-    public readonly image: string;
-    // Code directory on HDFS
-    public readonly codeDir: string;
-    //authentication file used for private Docker registry 
-    public readonly authFile?: string;
-
-    // List of taskRole, one task role at least
-    public taskRoles: PAITaskRole[];
-
-    //The virtual cluster job runs on.
-    public readonly virtualCluster: string;
-
-    /**
-     * Constructor
-     * @param jobName Name for the job, need to be unique
-     * @param image URL pointing to the Docker image for all tasks in the job
-     * @param dataDir Data directory existing on HDFS
-     * @param outputDir Output directory on HDFS
-     * @param taskRoles List of taskRole, one task role at least
-     */
-    constructor(jobName: string, image: string, codeDir: string,
-                taskRoles: PAITaskRole[], virtualCluster: string, authFile?: string) {
-        this.jobName = jobName;
-        this.image = image;
-        this.codeDir = codeDir;
-        this.taskRoles = taskRoles;
-        this.virtualCluster = virtualCluster;
-        this.authFile = authFile;
-    }
-}
-
-/**
- * PAI cluster configuration
- */
 export class PAIClusterConfig {
     public readonly userName: string;
     public readonly passWord?: string;
@@ -111,41 +28,31 @@ export class PAIClusterConfig {
 }
 
 /**
- * portList data structure used in PAI taskRole
+ * PAI trial job detail
  */
-export class PortListMetaData {
-    public readonly label: string = '';
-    public readonly beginAt: number = 0;
-    public readonly portNumber: number = 0;
-}
-  
-
-/**
- * PAI trial configuration
- */
-export class NNIPAITrialConfig extends TrialConfig {
-    public readonly cpuNum: number;
-    public readonly memoryMB: number;
-    public readonly image: string;
-
-    //The virtual cluster job runs on. If omitted, the job will run on default virtual cluster
-    public virtualCluster?: string;
-    //Shared memory for one task in the task role
-    public shmMB?: number;
-    //authentication file used for private Docker registry 
-    public authFile?: string;
-    //portList to specify the port used in container
-    public portList?: PortListMetaData[];
-
-    constructor(command: string, codeDir: string, gpuNum: number, cpuNum: number, memoryMB: number,
-                image: string, virtualCluster?: string, shmMB?: number, authFile?: string, portList?: PortListMetaData[]) {
-        super(command, codeDir, gpuNum);
-        this.cpuNum = cpuNum;
-        this.memoryMB = memoryMB;
-        this.image = image;
-        this.virtualCluster = virtualCluster;
-        this.shmMB = shmMB;
-        this.authFile = authFile;
-        this.portList = portList;
+export class PAITrialJobDetail implements TrialJobDetail {
+    public id: string;
+    public status: TrialJobStatus;
+    public paiJobName: string;
+    public submitTime: number;
+    public startTime?: number;
+    public endTime?: number;
+    public tags?: string[];
+    public url?: string;
+    public workingDirectory: string;
+    public form: TrialJobApplicationForm;
+    public logPath: string;
+    public isEarlyStopped?: boolean;
+
+    constructor(id: string, status: TrialJobStatus, paiJobName: string,
+                submitTime: number, workingDirectory: string, form: TrialJobApplicationForm, logPath: string) {
+        this.id = id;
+        this.status = status;
+        this.paiJobName = paiJobName;
+        this.submitTime = submitTime;
+        this.workingDirectory = workingDirectory;
+        this.form = form;
+        this.tags = [];
+        this.logPath = logPath;
     }
 }

src/nni_manager/training_service/pai/paiJobInfoCollector.ts

@@ -8,8 +8,7 @@ import { Deferred } from 'ts-deferred';
 import { NNIError, NNIErrorNames } from '../../common/errors';
 import { getLogger, Logger } from '../../common/log';
 import { TrialJobStatus } from '../../common/trainingService';
-import { PAIClusterConfig } from './paiConfig';
-import { PAITrialJobDetail } from './paiData';
+import { PAIClusterConfig, PAITrialJobDetail } from './paiConfig';
 
 /**
  * Collector PAI jobs info from PAI cluster, and update pai job status locally
@@ -26,8 +25,8 @@ export class PAIJobInfoCollector {
         this.finalStatuses = ['SUCCEEDED', 'FAILED', 'USER_CANCELED', 'SYS_CANCELED', 'EARLY_STOPPED'];
     }
 
-    public async retrieveTrialStatus(paiToken? : string, paiClusterConfig?: PAIClusterConfig): Promise<void> {
-        if (paiClusterConfig === undefined || paiToken === undefined) {
+    public async retrieveTrialStatus(token? : string, paiBaseClusterConfig?: PAIClusterConfig): Promise<void> {
+        if (paiBaseClusterConfig === undefined || token === undefined) {
             return Promise.resolve();
         }
 
@@ -36,7 +35,7 @@ export class PAIJobInfoCollector {
             if (paiTrialJob === undefined) {
                 throw new NNIError(NNIErrorNames.NOT_FOUND, `trial job id ${trialJobId} not found`);
             }
-            updatePaiTrialJobs.push(this.getSinglePAITrialJobInfo(paiTrialJob, paiToken, paiClusterConfig));
+            updatePaiTrialJobs.push(this.getSinglePAITrialJobInfo(paiTrialJob, token, paiBaseClusterConfig));
         }
 
         await Promise.all(updatePaiTrialJobs);
@@ -56,7 +55,7 @@ export class PAIJobInfoCollector {
             uri: `http://${paiClusterConfig.host}/rest-server/api/v1/user/${paiClusterConfig.userName}/jobs/${paiTrialJob.paiJobName}`,
             method: 'GET',
             json: true,
-            headers: {
+               headers: {
                 'Content-Type': 'application/json',
                 Authorization: `Bearer ${paiToken}`
             }
@@ -114,8 +113,12 @@ export class PAIJobInfoCollector {
                             paiTrialJob.endTime = response.body.jobStatus.completedTime;
                         }
                         // Set pai trial job's url to WebHDFS output path
-                        if (paiTrialJob.hdfsLogPath !== undefined) {
-                            paiTrialJob.url += `,${paiTrialJob.hdfsLogPath}`;
+                        if (paiTrialJob.logPath !== undefined) {
+                            if (paiTrialJob.url) {
+                                paiTrialJob.url += `,${paiTrialJob.logPath}`;
+                            } else {
+                                paiTrialJob.url = `${paiTrialJob.logPath}`;
+                            }
                         }
                     }
                 }