Merge pull request #197 from microsoft/master

merge master

examples/trials/auto-feature-engineering/README.md

+ **Automatic Feature Engineering in nni**	
+ ===	
+ 
+  Now we have an [example](https://github.com/SpongebBob/tabular_automl_NNI), which could automaticlly do feature engineering in nni. 
+  
+  These code come from our contributors. And thanks our lovely contributors!
+  
+ And welcome more and more people to join us!

examples/tuners/random_nas_tuner/random_nas_tuner.py

@@ -7,7 +7,9 @@ def random_archi_generator(nas_ss, random_state):
     '''
     chosen_archi = {}
     print("zql: nas search space: ", nas_ss)
-    for block_name, block in nas_ss.items():
+    for block_name, block_value in nas_ss.items():
+        assert block_value['_type'] == "mutable_layer", "Random NAS Tuner only receives NAS search space whose _type is 'mutable_layer'"
+        block = block_value['_value']
         tmp_block = {}
         for layer_name, layer in block.items():
             tmp_layer = {}

setup.py

@@ -35,9 +35,10 @@ setup(
     license = 'MIT',
     url = 'https://github.com/Microsoft/nni',
 
-    packages = find_packages('src/sdk/pynni', exclude=['tests']) + find_packages('tools'),
+    packages = find_packages('src/sdk/pynni', exclude=['tests']) + find_packages('src/sdk/pycli') + find_packages('tools'),
     package_dir = {
         'nni': 'src/sdk/pynni/nni',
+        'nnicli': 'src/sdk/pycli/nnicli',
         'nni_annotation': 'tools/nni_annotation',
         'nni_cmd': 'tools/nni_cmd',
         'nni_trial_tool':'tools/nni_trial_tool',

src/nni_manager/rest_server/restValidationSchemas.ts

@@ -51,10 +51,12 @@ export namespace ValidationSchemas {
                 command: joi.string().min(1),
                 virtualCluster: joi.string(),
                 shmMB: joi.number(),
+                authFile: joi.string(),
                 nasMode: joi.string().valid('classic_mode', 'enas_mode', 'oneshot_mode'),
                 worker: joi.object({
                     replicas: joi.number().min(1).required(),
                     image: joi.string().min(1),
+                    privateRegistryAuthPath: joi.string().min(1),
                     outputDir: joi.string(),
                     cpuNum: joi.number().min(1),
                     memoryMB: joi.number().min(100),
@@ -64,6 +66,7 @@ export namespace ValidationSchemas {
                 ps: joi.object({
                         replicas: joi.number().min(1).required(),
                         image: joi.string().min(1),
+                        privateRegistryAuthPath: joi.string().min(1),
                         outputDir: joi.string(),
                         cpuNum: joi.number().min(1),
                         memoryMB: joi.number().min(100),
@@ -73,6 +76,7 @@ export namespace ValidationSchemas {
                 master: joi.object({
                     replicas: joi.number().min(1).required(),
                     image: joi.string().min(1),
+                    privateRegistryAuthPath: joi.string().min(1),
                     outputDir: joi.string(),
                     cpuNum: joi.number().min(1),
                     memoryMB: joi.number().min(100),
@@ -83,6 +87,7 @@ export namespace ValidationSchemas {
                     name: joi.string().min(1),
                     taskNum: joi.number().min(1).required(),
                     image: joi.string().min(1),
+                    privateRegistryAuthPath: joi.string().min(1),
                     outputDir: joi.string(),
                     cpuNum: joi.number().min(1),
                     memoryMB: joi.number().min(100),

src/nni_manager/training_service/kubernetes/frameworkcontroller/frameworkcontrollerConfig.ts

@@ -43,8 +43,8 @@ export class FrameworkControllerTrialConfigTemplate extends KubernetesTrialConfi
     public readonly taskNum: number;
     constructor(taskNum: number, command : string, gpuNum : number,
                 cpuNum: number, memoryMB: number, image: string,
-                frameworkAttemptCompletionPolicy: FrameworkAttemptCompletionPolicy) {
-        super(command, gpuNum, cpuNum, memoryMB, image);
+                frameworkAttemptCompletionPolicy: FrameworkAttemptCompletionPolicy, privateRegistryFilePath?: string | undefined) {
+        super(command, gpuNum, cpuNum, memoryMB, image, privateRegistryFilePath);
         this.frameworkAttemptCompletionPolicy = frameworkAttemptCompletionPolicy;
         this.name = name;
         this.taskNum = taskNum;

src/nni_manager/training_service/kubernetes/frameworkcontroller/frameworkcontrollerTrainingService.ts

@@ -305,7 +305,7 @@ class FrameworkControllerTrainingService extends KubernetesTrainingService imple
         }
         // Generate frameworkcontroller job resource config object
         const frameworkcontrollerJobConfig: any =
-          this.generateFrameworkControllerJobConfig(trialJobId, trialWorkingFolder, frameworkcontrollerJobName, podResources);
+          await this.generateFrameworkControllerJobConfig(trialJobId, trialWorkingFolder, frameworkcontrollerJobName, podResources);
 
         return Promise.resolve(frameworkcontrollerJobConfig);
     }
@@ -329,8 +329,8 @@ class FrameworkControllerTrainingService extends KubernetesTrainingService imple
      * @param frameworkcontrollerJobName job name
      * @param podResources  pod template
      */
-    private generateFrameworkControllerJobConfig(trialJobId: string, trialWorkingFolder: string,
-                                                 frameworkcontrollerJobName : string, podResources : any) : any {
+    private async generateFrameworkControllerJobConfig(trialJobId: string, trialWorkingFolder: string,
+                                                 frameworkcontrollerJobName : string, podResources : any) : Promise<any> {
         if (this.fcClusterConfig === undefined) {
             throw new Error('frameworkcontroller Cluster config is not initialized');
         }
@@ -345,12 +345,14 @@ class FrameworkControllerTrainingService extends KubernetesTrainingService imple
             if (containerPort === undefined) {
                 throw new Error('Container port is not initialized');
             }
+            
             const taskRole: any = this.generateTaskRoleConfig(
                 trialWorkingFolder,
                 this.fcTrialConfig.taskRoles[index].image,
                 `run_${this.fcTrialConfig.taskRoles[index].name}.sh`,
                 podResources[index],
-                containerPort
+                containerPort,
+                await this.createRegistrySecret(this.fcTrialConfig.taskRoles[index].privateRegistryAuthPath)
             );
             taskRoles.push({
                 name: this.fcTrialConfig.taskRoles[index].name,
@@ -363,7 +365,7 @@ class FrameworkControllerTrainingService extends KubernetesTrainingService imple
             });
         }
 
-        return {
+        return Promise.resolve({
             apiVersion: `frameworkcontroller.microsoft.com/v1`,
             kind: 'Framework',
             metadata: {
@@ -379,11 +381,11 @@ class FrameworkControllerTrainingService extends KubernetesTrainingService imple
                 executionType: 'Start',
                 taskRoles: taskRoles
             }
-        };
+        });
     }
 
-    private generateTaskRoleConfig(trialWorkingFolder: string, replicaImage: string, runScriptFile: string,
-                                   podResources: any, containerPort: number): any {
+    private  generateTaskRoleConfig(trialWorkingFolder: string, replicaImage: string, runScriptFile: string,
+                                   podResources: any, containerPort: number, privateRegistrySecretName: string | undefined): any {
         if (this.fcClusterConfig === undefined) {
             throw new Error('frameworkcontroller Cluster config is not initialized');
         }
@@ -451,13 +453,22 @@ class FrameworkControllerTrainingService extends KubernetesTrainingService imple
                     mountPath: '/mnt/frameworkbarrier'
                 }]
         }];
-        const spec: any = {
-                containers: containers,
-                initContainers: initContainers,
-                restartPolicy: 'OnFailure',
-                volumes: volumeSpecMap.get('nniVolumes'),
-                hostNetwork: false
+        
+        let spec: any = {
+            containers: containers,
+            initContainers: initContainers,
+            restartPolicy: 'OnFailure',
+            volumes: volumeSpecMap.get('nniVolumes'),
+            hostNetwork: false
         };
+        if(privateRegistrySecretName) {
+            spec.imagePullSecrets = [
+                {
+                    name: privateRegistrySecretName 
+                }
+             ]
+        }
+
         if (this.fcClusterConfig.serviceAccountName !== undefined) {
             spec.serviceAccountName = this.fcClusterConfig.serviceAccountName;
         }

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

@@ -135,8 +135,8 @@ export class KubeflowTrialConfig extends KubernetesTrialConfig {
 export class KubeflowTrialConfigTemplate extends KubernetesTrialConfigTemplate {
     public readonly replicas: number;
     constructor(replicas: number, command : string, gpuNum : number,
-                cpuNum: number, memoryMB: number, image: string) {
-        super(command, gpuNum, cpuNum, memoryMB, image);
+                cpuNum: number, memoryMB: number, image: string, privateRegistryAuthPath?: string) {
+        super(command, gpuNum, cpuNum, memoryMB, image, privateRegistryAuthPath);
         this.replicas = replicas;
     }
 }

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

@@ -347,7 +347,7 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
         }
 
         // Generate kubeflow job resource config object
-        const kubeflowJobConfig: any = this.generateKubeflowJobConfig(trialJobId, trialWorkingFolder, kubeflowJobName, workerPodResources,
+        const kubeflowJobConfig: any = await this.generateKubeflowJobConfig(trialJobId, trialWorkingFolder, kubeflowJobName, workerPodResources,
                                                                       nonWorkerResources);
 
         return Promise.resolve(kubeflowJobConfig);
@@ -361,8 +361,8 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
      * @param workerPodResources worker pod template
      * @param nonWorkerPodResources non-worker pod template, like ps or master
      */
-    private generateKubeflowJobConfig(trialJobId: string, trialWorkingFolder: string, kubeflowJobName : string, workerPodResources : any,
-                                      nonWorkerPodResources?: any) : any {
+    private async generateKubeflowJobConfig(trialJobId: string, trialWorkingFolder: string, kubeflowJobName : string, workerPodResources : any,
+                                            nonWorkerPodResources?: any) : Promise<any> {
         if (this.kubeflowClusterConfig === undefined) {
             throw new Error('Kubeflow Cluster config is not initialized');
         }
@@ -377,29 +377,32 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
 
         const replicaSpecsObj: any = {};
         const replicaSpecsObjMap: Map<string, object> = new Map<string, object>();
-
         if (this.kubeflowTrialConfig.operatorType === 'tf-operator') {
             const tensorflowTrialConfig: KubeflowTrialConfigTensorflow = <KubeflowTrialConfigTensorflow>this.kubeflowTrialConfig;
+            let privateRegistrySecretName = await this.createRegistrySecret(tensorflowTrialConfig.worker.privateRegistryAuthPath);
             replicaSpecsObj.Worker = this.generateReplicaConfig(trialWorkingFolder, tensorflowTrialConfig.worker.replicas,
-                                                                tensorflowTrialConfig.worker.image, 'run_worker.sh', workerPodResources);
+                                                                tensorflowTrialConfig.worker.image, 'run_worker.sh', workerPodResources, privateRegistrySecretName);
             if (tensorflowTrialConfig.ps !== undefined) {
+                let privateRegistrySecretName: string | undefined = await this.createRegistrySecret(tensorflowTrialConfig.ps.privateRegistryAuthPath);
                 replicaSpecsObj.Ps = this.generateReplicaConfig(trialWorkingFolder, tensorflowTrialConfig.ps.replicas,
-                                                                tensorflowTrialConfig.ps.image, 'run_ps.sh', nonWorkerPodResources);
+                                                                tensorflowTrialConfig.ps.image, 'run_ps.sh', nonWorkerPodResources, privateRegistrySecretName);
             }
             replicaSpecsObjMap.set(this.kubernetesCRDClient.jobKind, {tfReplicaSpecs: replicaSpecsObj});
         } else if (this.kubeflowTrialConfig.operatorType === 'pytorch-operator') {
             const pytorchTrialConfig: KubeflowTrialConfigPytorch = <KubeflowTrialConfigPytorch>this.kubeflowTrialConfig;
             if (pytorchTrialConfig.worker !== undefined) {
+                let privateRegistrySecretName: string | undefined = await this.createRegistrySecret(pytorchTrialConfig.worker.privateRegistryAuthPath);
                 replicaSpecsObj.Worker = this.generateReplicaConfig(trialWorkingFolder, pytorchTrialConfig.worker.replicas,
-                                                                    pytorchTrialConfig.worker.image, 'run_worker.sh', workerPodResources);
+                                                                    pytorchTrialConfig.worker.image, 'run_worker.sh', workerPodResources, privateRegistrySecretName);
             }
+            let privateRegistrySecretName: string | undefined = await this.createRegistrySecret(pytorchTrialConfig.master.privateRegistryAuthPath);
             replicaSpecsObj.Master = this.generateReplicaConfig(trialWorkingFolder, pytorchTrialConfig.master.replicas,
-                                                                pytorchTrialConfig.master.image, 'run_master.sh', nonWorkerPodResources);
+                                                                pytorchTrialConfig.master.image, 'run_master.sh', nonWorkerPodResources, privateRegistrySecretName);
 
             replicaSpecsObjMap.set(this.kubernetesCRDClient.jobKind, {pytorchReplicaSpecs: replicaSpecsObj});
         }
 
-        return {
+        return Promise.resolve({
             apiVersion: `kubeflow.org/${this.kubernetesCRDClient.apiVersion}`,
             kind: this.kubernetesCRDClient.jobKind,
             metadata: {
@@ -412,7 +415,7 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
                 }
             },
             spec: replicaSpecsObjMap.get(this.kubernetesCRDClient.jobKind)
-        };
+        });
     }
 
     /**
@@ -424,7 +427,7 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
      * @param podResources pod resource config section
      */
     private generateReplicaConfig(trialWorkingFolder: string, replicaNumber: number, replicaImage: string, runScriptFile: string,
-                                  podResources: any): any {
+                                  podResources: any, privateRegistrySecretName: string | undefined): any {
         if (this.kubeflowClusterConfig === undefined) {
             throw new Error('Kubeflow Cluster config is not initialized');
         }
@@ -436,7 +439,7 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
         if (this.kubernetesCRDClient === undefined) {
             throw new Error('Kubeflow operator client is not initialized');
         }
-
+        // The config spec for volume field
         const volumeSpecMap: Map<string, object> = new Map<string, object>();
         if (this.kubeflowClusterConfig.storageType === 'azureStorage') {
             volumeSpecMap.set('nniVolumes', [
@@ -459,7 +462,34 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
                 }
             }]);
         }
-
+        // The config spec for container field
+        const containersSpecMap: Map<string, object> = new Map<string, object>(); 
+        containersSpecMap.set('containers', [
+        {
+                // Kubeflow tensorflow operator requires that containers' name must be tensorflow
+                // TODO: change the name based on operator's type
+                name: this.kubernetesCRDClient.containerName,
+                image: replicaImage,
+                args: ['sh', `${path.join(trialWorkingFolder, runScriptFile)}`],
+                volumeMounts: [
+                {
+                    name: 'nni-vol',
+                    mountPath: this.CONTAINER_MOUNT_PATH
+                }],
+                resources: podResources
+            }
+        ]);
+        let spec: any = {
+            containers: containersSpecMap.get('containers'),
+            restartPolicy: 'ExitCode',
+            volumes: volumeSpecMap.get('nniVolumes')
+        }
+        if (privateRegistrySecretName) {
+            spec.imagePullSecrets = [
+                {
+                    name: privateRegistrySecretName
+                }]
+        }
         return {
             replicas: replicaNumber,
             template: {
@@ -467,26 +497,9 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
                     // tslint:disable-next-line:no-null-keyword
                     creationTimestamp: null
                 },
-                spec: {
-                    containers: [
-                    {
-                        // Kubeflow tensorflow operator requires that containers' name must be tensorflow
-                        // TODO: change the name based on operator's type
-                        name: this.kubernetesCRDClient.containerName,
-                        image: replicaImage,
-                        args: ['sh', `${path.join(trialWorkingFolder, runScriptFile)}`],
-                        volumeMounts: [
-                        {
-                            name: 'nni-vol',
-                            mountPath: this.CONTAINER_MOUNT_PATH
-                        }],
-                        resources: podResources
-                    }],
-                    restartPolicy: 'ExitCode',
-                    volumes: volumeSpecMap.get('nniVolumes')
-                }
+                spec: spec
             }
-        };
+        }
     }
 }
 // tslint:enable: no-unsafe-any no-any

src/nni_manager/training_service/kubernetes/kubernetesConfig.ts

@@ -179,6 +179,9 @@ export class KubernetesTrialConfigTemplate {
     // Docker image
     public readonly image: string;
 
+    // Private registry config file path to download docker iamge
+    public readonly privateRegistryAuthPath?: string;
+
     // Trail command
     public readonly command : string;
 
@@ -186,12 +189,13 @@ export class KubernetesTrialConfigTemplate {
     public readonly gpuNum : number;
 
     constructor(command : string, gpuNum : number,
-                cpuNum: number, memoryMB: number, image: string) {
+                cpuNum: number, memoryMB: number, image: string, privateRegistryAuthPath?: string) {
         this.command = command;
         this.gpuNum = gpuNum;
         this.cpuNum = cpuNum;
         this.memoryMB = memoryMB;
         this.image = image;
+        this.privateRegistryAuthPath = privateRegistryAuthPath;
     }
 }
 

src/nni_manager/training_service/kubernetes/kubernetesTrainingService.ts

@@ -38,6 +38,8 @@ import { KubernetesClusterConfig } from './kubernetesConfig';
 import { kubernetesScriptFormat, KubernetesTrialJobDetail } from './kubernetesData';
 import { KubernetesJobRestServer } from './kubernetesJobRestServer';
 
+var fs = require('fs');
+
 /**
  * Training Service implementation for Kubernetes
  */
@@ -327,5 +329,34 @@ abstract class KubernetesTrainingService {
 
         return Promise.resolve();
     }
+
+    protected async createRegistrySecret(filePath: string | undefined): Promise<string | undefined> {
+        if(filePath === undefined || filePath === '') {
+            return undefined;
+        }
+        let body = fs.readFileSync(filePath).toString('base64');
+        let registrySecretName = String.Format('nni-secret-{0}', uniqueString(8)
+                                                                            .toLowerCase());
+        await this.genericK8sClient.createSecret(
+            {
+                apiVersion: 'v1',
+                kind: 'Secret',
+                metadata: {
+                    name: registrySecretName,
+                    namespace: 'default',
+                    labels: {
+                        app: this.NNI_KUBERNETES_TRIAL_LABEL,
+                        expId: getExperimentId()
+                    }
+                },
+                type: 'kubernetes.io/dockerconfigjson',
+                data: {
+                    '.dockerconfigjson': body
+                }
+            }
+        );
+        return registrySecretName;
+    }
+     
 }
 export { KubernetesTrainingService };

src/nni_manager/training_service/pai/paiConfig.ts

@@ -71,6 +71,8 @@ export class PAIJobConfig {
     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[];
@@ -87,12 +89,13 @@ export class PAIJobConfig {
      * @param taskRoles List of taskRole, one task role at least
      */
     constructor(jobName: string, image : string, codeDir : string,
-                taskRoles : PAITaskRole[], virtualCluster: 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;
     }
 }
 
@@ -129,14 +132,17 @@ export class NNIPAITrialConfig extends TrialConfig {
     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;
 
     constructor(command : string, codeDir : string, gpuNum : number, cpuNum: number, memoryMB: number,
-                image: string, virtualCluster?: string, shmMB?: number) {
+                image: string, virtualCluster?: string, shmMB?: number, authFile?: string) {
         super(command, codeDir, gpuNum);
         this.cpuNum = cpuNum;
         this.memoryMB = memoryMB;
         this.image = image;
         this.virtualCluster = virtualCluster;
         this.shmMB = shmMB;
+        this.authFile = authFile;
     }
 }

src/nni_manager/training_service/pai/paiTrainingService.ts

@@ -442,7 +442,7 @@ class PAITrainingService implements TrainingService {
                 // Task command
                 nniPaiTrialCommand,
                 // Task shared memory
-                this.paiTrialConfig.shmMB
+                this.paiTrialConfig.shmMB,
             )
         ];
 
@@ -456,7 +456,9 @@ class PAITrainingService implements TrainingService {
             // PAI Task roles
             paiTaskRoles,
             // Add Virutal Cluster
-            this.paiTrialConfig.virtualCluster === undefined ? 'default' : this.paiTrialConfig.virtualCluster.toString()
+            this.paiTrialConfig.virtualCluster === undefined ? 'default' : this.paiTrialConfig.virtualCluster.toString(),
+            //Task auth File
+            this.paiTrialConfig.authFile
         );
 
         // Step 2. Upload code files in codeDir onto HDFS

src/sdk/pycli/nnicli/__init__.py

+# Copyright (c) Microsoft Corporation. All rights reserved.
+#
+# MIT License
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
+# associated documentation files (the "Software"), to deal in the Software without restriction,
+# including without limitation the rights to use, copy, modify, merge, publish, distribute,
+# sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all copies or
+# substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED *AS IS*, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT
+# NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+# DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT
+# OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+# ==================================================================================================
+
+from .nni_client import *

src/sdk/pycli/nnicli/nni_client.py

+# Copyright (c) Microsoft Corporation. All rights reserved.
+#
+# MIT License
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy of this software and
+# associated documentation files (the "Software"), to deal in the Software without restriction,
+# including without limitation the rights to use, copy, modify, merge, publish, distribute,
+# sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in all copies or
+# substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED *AS IS*, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT
+# NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+# DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT
+# OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+# ==================================================================================================
+
+""" A python wrapper for nni rest api
+
+Example:
+
+import nnicli as nc
+
+nc.start_nni('../../../../examples/trials/mnist/config.yml')
+
+nc.set_endpoint('http://localhost:8080')
+
+print(nc.version())
+print(nc.get_experiment_status())
+
+print(nc.get_job_statistics())
+print(nc.list_trial_jobs())
+
+nc.stop_nni()
+
+"""
+
+import sys
+import os
+import subprocess
+import requests
+
+__all__ = [
+    'start_nni',
+    'stop_nni',
+    'set_endpoint',
+    'version',
+    'get_experiment_status',
+    'get_experiment_profile',
+    'get_trial_job',
+    'list_trial_jobs',
+    'get_job_statistics',
+    'get_job_metrics',
+    'export_data'
+]
+
+EXPERIMENT_PATH = 'experiment'
+VERSION_PATH = 'version'
+STATUS_PATH = 'check-status'
+JOB_STATISTICS_PATH = 'job-statistics'
+TRIAL_JOBS_PATH = 'trial-jobs'
+METRICS_PATH = 'metric-data'
+EXPORT_DATA_PATH = 'export-data'
+
+API_ROOT_PATH = 'api/v1/nni'
+
+_api_endpoint = None
+
+def set_endpoint(endpoint):
+    """set endpoint of nni rest server for nnicli, for example:
+    http://localhost:8080
+    """
+    global _api_endpoint
+    _api_endpoint = endpoint
+
+def _check_endpoint():
+    if _api_endpoint is None:
+        raise AssertionError("Please call set_endpoint to specify nni endpoint")
+
+def _nni_rest_get(api_path, response_type='json'):
+    _check_endpoint()
+    uri = '{}/{}/{}'.format(_api_endpoint, API_ROOT_PATH, api_path)
+    res = requests.get(uri)
+    if _http_succeed(res.status_code):
+        if response_type == 'json':
+            return res.json()
+        elif response_type == 'text':
+            return res.text
+        else:
+            raise AssertionError('Incorrect response_type')
+    else:
+        return None
+
+def _http_succeed(status_code):
+    return status_code // 100 == 2
+
+def _create_process(cmd):
+    if sys.platform == 'win32':
+        process = subprocess.Popen(cmd, stdout=subprocess.PIPE, creationflags=subprocess.CREATE_NEW_PROCESS_GROUP)
+    else:
+        process = subprocess.Popen(cmd, stdout=subprocess.PIPE)
+
+    while process.poll() is None:
+        output = process.stdout.readline()
+        if output:
+            print(output.decode('utf-8').strip())
+    return process.returncode
+
+def start_nni(config_file):
+    """start nni experiment with specified configuration file"""
+    cmd = 'nnictl create --config {}'.format(config_file).split(' ')
+    if _create_process(cmd) != 0:
+        raise RuntimeError('Failed to start nni.')
+
+def stop_nni():
+    """stop nni experiment"""
+    cmd = 'nnictl stop'.split(' ')
+    if _create_process(cmd) != 0:
+        raise RuntimeError('Failed to stop nni.')
+
+def version():
+    """return version of nni"""
+    return _nni_rest_get(VERSION_PATH, 'text')
+
+def get_experiment_status():
+    """return experiment status as a dict"""
+    return _nni_rest_get(STATUS_PATH)
+
+def get_experiment_profile():
+    """return experiment profile as a dict"""
+    return _nni_rest_get(EXPERIMENT_PATH)
+
+def get_trial_job(trial_job_id):
+    """return trial job information as a dict"""
+    assert trial_job_id is not None
+    return _nni_rest_get(os.path.join(TRIAL_JOBS_PATH, trial_job_id))
+
+def list_trial_jobs():
+    """return information for all trial jobs as a list"""
+    return _nni_rest_get(TRIAL_JOBS_PATH)
+
+def get_job_statistics():
+    """return trial job statistics information as a dict"""
+    return _nni_rest_get(JOB_STATISTICS_PATH)
+
+def get_job_metrics(trial_job_id=None):
+    """return trial job metrics"""
+    api_path = METRICS_PATH if trial_job_id is None else os.path.join(METRICS_PATH, trial_job_id)
+    return _nni_rest_get(api_path)
+
+def export_data():
+    """return exported information for all trial jobs"""
+    return _nni_rest_get(EXPORT_DATA_PATH)

src/sdk/pycli/setup.py

+import setuptools
+
+setuptools.setup(
+    name = 'nnicli',
+    version = '999.0.0-developing',
+    packages = setuptools.find_packages(),
+
+    python_requires = '>=3.5',
+    install_requires = [
+        'requests'
+    ],
+
+    author = 'Microsoft NNI Team',
+    author_email = 'nni@microsoft.com',
+    description = 'nnicli for Neural Network Intelligence project',
+    license = 'MIT',
+    url = 'https://github.com/Microsoft/nni',
+)

src/sdk/pynni/nni/hyperopt_tuner/hyperopt_tuner.py

@@ -190,13 +190,19 @@ class HyperoptTuner(Tuner):
     HyperoptTuner is a tuner which using hyperopt algorithm.
     """
 
-    def __init__(self, algorithm_name, optimize_mode='minimize'):
+    def __init__(self, algorithm_name, optimize_mode='minimize', 
+                 parallel_optimize=False, constant_liar_type='min'):
         """
         Parameters
         ----------
         algorithm_name : str
             algorithm_name includes "tpe", "random_search" and anneal".
         optimize_mode : str
+        parallel_optimize : bool 
+            More detail could reference: docs/en_US/Tuner/HyperoptTuner.md
+        constant_liar_type : str
+            constant_liar_type including "min", "max" and "mean"
+            More detail could reference: docs/en_US/Tuner/HyperoptTuner.md
         """
         self.algorithm_name = algorithm_name
         self.optimize_mode = OptimizeMode(optimize_mode)
@@ -205,6 +211,13 @@ class HyperoptTuner(Tuner):
         self.rval = None
         self.supplement_data_num = 0
 
+        self.parallel = parallel_optimize
+        if self.parallel:
+            self.CL_rval = None
+            self.constant_liar_type = constant_liar_type
+            self.running_data = []
+            self.optimal_y = None
+
     def _choose_tuner(self, algorithm_name):
         """
         Parameters
@@ -266,6 +279,10 @@ class HyperoptTuner(Tuner):
             # but it can cause deplicate parameter rarely
             total_params = self.get_suggestion(random_search=True)
         self.total_data[parameter_id] = total_params
+
+        if self.parallel:
+            self.running_data.append(parameter_id)
+        
         params = split_index(total_params)
         return params
 
@@ -287,10 +304,39 @@ class HyperoptTuner(Tuner):
             raise RuntimeError('Received parameter_id not in total_data.')
         params = self.total_data[parameter_id]
 
+        # code for parallel
+        if self.parallel:
+            constant_liar = kwargs.get('constant_liar', False)
+
+            if constant_liar:
+                rval = self.CL_rval
+            else:
+                rval = self.rval
+                self.running_data.remove(parameter_id)
+
+                # update the reward of optimal_y
+                if self.optimal_y is None:
+                    if self.constant_liar_type == 'mean':
+                        self.optimal_y = [reward, 1]
+                    else:
+                        self.optimal_y = reward
+                else:
+                    if self.constant_liar_type == 'mean':
+                        _sum = self.optimal_y[0] + reward
+                        _number = self.optimal_y[1] + 1
+                        self.optimal_y = [_sum, _number]
+                    elif self.constant_liar_type == 'min':
+                        self.optimal_y = min(self.optimal_y, reward)
+                    elif self.constant_liar_type == 'max':
+                        self.optimal_y = max(self.optimal_y, reward)
+                logger.debug("Update optimal_y with reward, optimal_y = %s", self.optimal_y)
+        else:
+            rval = self.rval
+
+
         if self.optimize_mode is OptimizeMode.Maximize:
             reward = -reward
 
-        rval = self.rval
         domain = rval.domain
         trials = rval.trials
 
@@ -375,13 +421,26 @@ class HyperoptTuner(Tuner):
         total_params : dict
             parameter suggestion
         """
+        if self.parallel and len(self.total_data)>20 and len(self.running_data) and self.optimal_y is not None:
+            self.CL_rval = copy.deepcopy(self.rval)
+            if self.constant_liar_type == 'mean':
+                _constant_liar_y = self.optimal_y[0] / self.optimal_y[1]
+            else:
+                _constant_liar_y = self.optimal_y
+            for _parameter_id in self.running_data:
+                self.receive_trial_result(parameter_id=_parameter_id, parameters=None, value=_constant_liar_y, constant_liar=True)
+            rval = self.CL_rval
 
-        rval = self.rval
+            random_state = np.random.randint(2**31 - 1)
+        else:
+            rval = self.rval
+            random_state = rval.rstate.randint(2**31 - 1)
+    
         trials = rval.trials
         algorithm = rval.algo
         new_ids = rval.trials.new_trial_ids(1)
         rval.trials.refresh()
-        random_state = rval.rstate.randint(2**31 - 1)
+
         if random_search:
             new_trials = hp.rand.suggest(new_ids, rval.domain, trials,
                                          random_state)

test/cli_test.py

+# Copyright (c) Microsoft Corporation
+# All rights reserved.
+#
+# MIT License
+#
+# Permission is hereby granted, free of charge,
+# to any person obtaining a copy of this software and associated
+# documentation files (the "Software"), to deal in the Software without restriction,
+# including without limitation the rights to use, copy, modify, merge, publish,
+# distribute, sublicense, and/or sell copies of the Software, and
+# to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+# The above copyright notice and this permission notice shall be included
+# in all copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED *AS IS*, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING
+# BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
+# DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+import sys
+import time
+import traceback
+from utils import GREEN, RED, CLEAR, setup_experiment
+
+def test_nni_cli():
+    import nnicli as nc
+
+    config_file = 'config_test/examples/mnist.test.yml'
+
+    try:
+        # Sleep here to make sure previous stopped exp has enough time to exit to avoid port conflict
+        time.sleep(6)
+        print(GREEN + 'Testing nnicli:' + config_file + CLEAR)
+        nc.start_nni(config_file)
+        time.sleep(3)
+        nc.set_endpoint('http://localhost:8080')
+        print(nc.version())
+        print(nc.get_job_statistics())
+        print(nc.get_experiment_status())
+        nc.list_trial_jobs()
+
+        print(GREEN + 'Test nnicli {}: TEST PASS'.format(config_file) + CLEAR)
+    except Exception as error:
+        print(RED + 'Test nnicli {}: TEST FAIL'.format(config_file) + CLEAR)
+        print('%r' % error)
+        traceback.print_exc()
+        raise error
+    finally:
+        nc.stop_nni()
+
+if __name__ == '__main__':
+    installed = (sys.argv[-1] != '--preinstall')
+    setup_experiment(installed)
+
+    test_nni_cli()