Merge pull request #1520 from suiguoxin/v1.0-conf-resolve

merge v1.0 back to master (conflicts resolved)

src/nni_manager/rest_server/restValidationSchemas.ts

@@ -53,6 +53,11 @@ export namespace ValidationSchemas {
                 shmMB: joi.number(),
                 authFile: joi.string(),
                 nasMode: joi.string().valid('classic_mode', 'enas_mode', 'oneshot_mode', 'darts_mode'),
+                portList: joi.array().items(joi.object({
+                    label: joi.string().required(),
+                    beginAt: joi.number().required(),
+                    portNumber: joi.number().required(),
+                })),
                 worker: joi.object({
                     replicas: joi.number().min(1).required(),
                     image: joi.string().min(1),
@@ -120,7 +125,8 @@ export namespace ValidationSchemas {
                 azureStorage: joi.object({
                     accountName: joi.string().regex(/^([0-9]|[a-z]|[A-Z]|-){3,31}$/),
                     azureShare: joi.string().regex(/^([0-9]|[a-z]|[A-Z]|-){3,63}$/)
-                })
+                }),
+                uploadRetryCount: joi.number().min(1)
             }),
             frameworkcontroller_config: joi.object({
                 storage: joi.string().min(1),
@@ -136,7 +142,8 @@ export namespace ValidationSchemas {
                 azureStorage: joi.object({
                     accountName: joi.string().regex(/^([0-9]|[a-z]|[A-Z]|-){3,31}$/),
                     azureShare: joi.string().regex(/^([0-9]|[a-z]|[A-Z]|-){3,63}$/)
-                })
+                }),
+                uploadRetryCount: joi.number().min(1)
             }),
             nni_manager_ip: joi.object({
                 nniManagerIp: joi.string().min(1)

src/nni_manager/training_service/kubernetes/azureStorageClientUtils.ts

@@ -35,15 +35,15 @@ export namespace AzureStorageClientUtility {
      * @param fileServerClient
      * @param azureShare
      */
-    export async function createShare(fileServerClient: any, azureShare: any): Promise<void> {
-        const deferred: Deferred<void> = new Deferred<void>();
+    export async function createShare(fileServerClient: any, azureShare: any): Promise<boolean> {
+        const deferred: Deferred<boolean> = new Deferred<boolean>();
         fileServerClient.createShareIfNotExists(azureShare, (error: any, result: any, response: any) => {
             if (error) {
                 getLogger()
                   .error(`Create share failed:, ${error}`);
-                deferred.reject(error);
+                deferred.resolve(false);
             } else {
-                deferred.resolve();
+                deferred.resolve(true);
             }
         });
 
@@ -56,18 +56,17 @@ export namespace AzureStorageClientUtility {
      * @param azureFoler
      * @param azureShare
      */
-    export async function createDirectory(fileServerClient: azureStorage.FileService, azureFoler: any, azureShare: any): Promise<void> {
-        const deferred: Deferred<void> = new Deferred<void>();
+    export async function createDirectory(fileServerClient: azureStorage.FileService, azureFoler: any, azureShare: any): Promise<boolean> {
+        const deferred: Deferred<boolean> = new Deferred<boolean>();
         fileServerClient.createDirectoryIfNotExists(azureShare, azureFoler, (error: any, result: any, response: any) => {
             if (error) {
                 getLogger()
                   .error(`Create directory failed:, ${error}`);
-                deferred.reject(error);
+                deferred.resolve(false);
             } else {
-                deferred.resolve();
+                deferred.resolve(true);
             }
         });
-
         return deferred.promise;
     }
 
@@ -77,16 +76,20 @@ export namespace AzureStorageClientUtility {
      * @param azureDirectory
      */
     export async function createDirectoryRecursive(fileServerClient: azureStorage.FileService, azureDirectory: string,
-                                                   azureShare: any): Promise<void> {
-        const deferred: Deferred<void> = new Deferred<void>();
+                                                   azureShare: any): Promise<boolean> {
+        const deferred: Deferred<boolean> = new Deferred<boolean>();
         const directories: string[] = azureDirectory.split('/');
         let rootDirectory: string = '';
         for (const directory of directories) {
             rootDirectory += directory;
-            await createDirectory(fileServerClient, rootDirectory, azureShare);
+            let result:boolean = await createDirectory(fileServerClient, rootDirectory, azureShare);
+            if (!result) {
+                deferred.resolve(false);
+                return deferred.promise;
+            }
             rootDirectory += '/';
         }
-        deferred.resolve();
+        deferred.resolve(true);
 
         return deferred.promise;
     }
@@ -100,16 +103,16 @@ export namespace AzureStorageClientUtility {
      * @param localFilePath
      */
     async function uploadFileToAzure(fileServerClient: any, azureDirectory: string, azureFileName: any, azureShare: any,
-                                     localFilePath: string): Promise<void> {
-        const deferred: Deferred<void> = new Deferred<void>();
+                                     localFilePath: string): Promise<boolean> {
+        const deferred: Deferred<boolean> = new Deferred<boolean>();
         await fileServerClient.createFileFromLocalFile(azureShare, azureDirectory, azureFileName, localFilePath,
                                                        (error: any, result: any, response: any) => {
             if (error) {
                 getLogger()
                   .error(`Upload file failed:, ${error}`);
-                deferred.reject(error);
+                deferred.resolve(false);
             } else {
-                deferred.resolve();
+                deferred.resolve(true);
             }
         });
 
@@ -125,17 +128,17 @@ export namespace AzureStorageClientUtility {
      * @param localFilePath
      */
     async function downloadFile(fileServerClient: any, azureDirectory: string, azureFileName: any, azureShare: any,
-                                localFilePath: string): Promise<void> {
-        const deferred: Deferred<void> = new Deferred<void>();
+                                localFilePath: string): Promise<boolean> {
+        const deferred: Deferred<boolean> = new Deferred<boolean>();
         // tslint:disable-next-line:non-literal-fs-path
         await fileServerClient.getFileToStream(azureShare, azureDirectory, azureFileName, fs.createWriteStream(localFilePath),
                                                (error: any, result: any, response: any) => {
             if (error) {
                 getLogger()
                   .error(`Download file failed:, ${error}`);
-                deferred.reject(error);
+                deferred.resolve(false);
             } else {
-                deferred.resolve();
+                deferred.resolve(true);
             }
         });
 
@@ -151,28 +154,38 @@ export namespace AzureStorageClientUtility {
      */
     // tslint:disable:non-literal-fs-path
     export async function uploadDirectory(fileServerClient: azureStorage.FileService, azureDirectory: string, azureShare: any,
-                                          localDirectory: string): Promise<void> {
-        const deferred: Deferred<void> = new Deferred<void>();
+                                          localDirectory: string): Promise<boolean> {
+        const deferred: Deferred<boolean> = new Deferred<boolean>();
         const fileNameArray: string[] = fs.readdirSync(localDirectory);
-        await createDirectoryRecursive(fileServerClient, azureDirectory, azureShare);
+        let result: boolean = await createDirectoryRecursive(fileServerClient, azureDirectory, azureShare);
+        if (!result) {
+            deferred.resolve(false);
+            return deferred.promise;
+        }
         for (const fileName of fileNameArray) {
             const fullFilePath: string = path.join(localDirectory, fileName);
             try {
+                let resultUploadFile: boolean = true;
+                let resultUploadDir: boolean = true;
                 if (fs.lstatSync(fullFilePath)
                       .isFile()) {
-                    await uploadFileToAzure(fileServerClient, azureDirectory, fileName, azureShare, fullFilePath);
+                    resultUploadFile = await uploadFileToAzure(fileServerClient, azureDirectory, fileName, azureShare, fullFilePath);
                 } else {
                     // If filePath is a directory, recuisively copy it to azure
-                    await uploadDirectory(fileServerClient, String.Format('{0}/{1}', azureDirectory, fileName), azureShare, fullFilePath);
+                    resultUploadDir = await uploadDirectory(fileServerClient, String.Format('{0}/{1}', azureDirectory, fileName), azureShare, fullFilePath);
+                }
+                if (!(resultUploadFile && resultUploadDir)) {
+                    deferred.resolve(false);
+                    return deferred.promise;
                 }
             } catch (error) {
-                deferred.reject(error);
+                deferred.resolve(false);
 
                 return deferred.promise;
             }
         }
         // All files/directories are copied successfully, resolve
-        deferred.resolve();
+        deferred.resolve(true);
 
         return deferred.promise;
     }

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

@@ -25,7 +25,7 @@ import * as path from 'path';
 import * as component from '../../../common/component';
 import { getExperimentId } from '../../../common/experimentStartupInfo';
 import {
-    JobApplicationForm, NNIManagerIpConfig, TrialJobApplicationForm, TrialJobDetail
+    JobApplicationForm, NNIManagerIpConfig, TrialJobApplicationForm, TrialJobDetail, TrialJobStatus
 } from '../../../common/trainingService';
 import { delay, generateParamFileName, getExperimentRootDir, uniqueString } from '../../../common/utils';
 import { CONTAINER_INSTALL_NNI_SHELL_FORMAT } from '../../common/containerJobData';
@@ -102,10 +102,13 @@ class FrameworkControllerTrainingService extends KubernetesTrainingService imple
 
         //upload code files
         const trialJobOutputUrl: string = await this.uploadCodeFiles(trialJobId, trialLocalTempFolder);
-
+        let initStatus: TrialJobStatus = 'WAITING';
+        if (!trialJobOutputUrl) {
+            initStatus = 'FAILED';
+        }
         const trialJobDetail: KubernetesTrialJobDetail = new KubernetesTrialJobDetail(
             trialJobId,
-            'WAITING',
+            initStatus,
             Date.now(),
             trialWorkingFolder,
             form,
@@ -208,24 +211,10 @@ class FrameworkControllerTrainingService extends KubernetesTrainingService imple
         let trialJobOutputUrl: string = '';
 
         if (this.fcClusterConfig.storageType === 'azureStorage') {
-            if (this.azureStorageClient === undefined) {
-                throw new Error('azureStorageClient is not initialized');
-            }
-            try {
-                //upload local files, including scripts for running the trial and configuration (e.g., hyperparameters) for the trial, to azure storage
-                await AzureStorageClientUtility.uploadDirectory(
-                    this.azureStorageClient, `nni/${getExperimentId()}/${trialJobId}`, this.azureStorageShare, `${trialLocalTempFolder}`);
-                //upload code files to azure storage
-                await AzureStorageClientUtility.uploadDirectory(
-                    this.azureStorageClient, `nni/${getExperimentId()}/${trialJobId}`, this.azureStorageShare, `${this.fcTrialConfig.codeDir}`);
-
-                trialJobOutputUrl = `https://${this.azureStorageAccountName}.file.core.windows.net/` + 
-                                    `${this.azureStorageShare}/${path.join('nni', getExperimentId(), trialJobId, 'output')}`;
-            } catch (error) {
-                this.log.error(error);
-
-                return Promise.reject(error);
-            }
+            const azureFrameworkControllerClusterConfig: FrameworkControllerClusterConfigAzure =
+                      <FrameworkControllerClusterConfigAzure>this.fcClusterConfig;
+            trialJobOutputUrl = await this.uploadFilesToAzureStorage(trialJobId, trialLocalTempFolder, this.fcTrialConfig.codeDir,
+                 azureFrameworkControllerClusterConfig.uploadRetryCount);
         } else if (this.fcClusterConfig.storageType === 'nfs') {
             const nfsFrameworkControllerClusterConfig: FrameworkControllerClusterConfigNFS =
               <FrameworkControllerClusterConfigNFS>this.fcClusterConfig;

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

@@ -27,7 +27,7 @@ import * as component from '../../../common/component';
 
 import { getExperimentId } from '../../../common/experimentStartupInfo';
 import {
-    JobApplicationForm, NNIManagerIpConfig, TrialJobApplicationForm, TrialJobDetail
+    JobApplicationForm, NNIManagerIpConfig, TrialJobApplicationForm, TrialJobDetail, TrialJobStatus
 } from '../../../common/trainingService';
 import { delay, generateParamFileName, getExperimentRootDir, uniqueString } from '../../../common/utils';
 import { CONTAINER_INSTALL_NNI_SHELL_FORMAT } from '../../common/containerJobData';
@@ -102,9 +102,13 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
         await this.prepareRunScript(trialLocalTempFolder, trialJobId, trialWorkingFolder, curTrialSequenceId, form);
         //upload files to sotrage
         const trialJobOutputUrl: string = await this.uploadCodeFiles(trialJobId, trialLocalTempFolder);
+        let initStatus: TrialJobStatus = 'WAITING';
+        if (!trialJobOutputUrl) {
+            initStatus = 'FAILED';
+        }
         const trialJobDetail: KubernetesTrialJobDetail = new KubernetesTrialJobDetail(
             trialJobId,
-            'WAITING',
+            initStatus,
             Date.now(),
             trialWorkingFolder,
             form,
@@ -215,23 +219,8 @@ class KubeflowTrainingService extends KubernetesTrainingService implements Kuber
             if (this.azureStorageClient === undefined) {
                 throw new Error('azureStorageClient is not initialized');
             }
-            try {
-                //upload local files, including scripts for running the trial and configuration (e.g., hyperparameters) for the trial, to azure storage
-                await AzureStorageClientUtility.uploadDirectory(this.azureStorageClient,
-                                                                `nni/${getExperimentId()}/${trialJobId}`, this.azureStorageShare,
-                                                                `${trialLocalTempFolder}`);
-                //upload code files to azure storage
-                await AzureStorageClientUtility.uploadDirectory(this.azureStorageClient,
-                                                                `nni/${getExperimentId()}/${trialJobId}`, this.azureStorageShare,
-                                                                `${this.kubeflowTrialConfig.codeDir}`);
-
-                trialJobOutputUrl = `https://${this.azureStorageAccountName}.file.core.windows.net/${this.azureStorageShare}` + 
-                                    `/${path.join('nni', getExperimentId(), trialJobId, 'output')}`;
-            } catch (error) {
-                this.log.error(error);
-
-                return Promise.reject(error);
-            }
+            const azureKubeflowClusterConfig: KubeflowClusterConfigAzure = <KubeflowClusterConfigAzure>this.kubeflowClusterConfig;
+            trialJobOutputUrl = await this.uploadFilesToAzureStorage(trialJobId, trialLocalTempFolder, this.kubeflowTrialConfig.codeDir, azureKubeflowClusterConfig.uploadRetryCount);
         } else if (this.kubeflowClusterConfig.storage === 'nfs' || this.kubeflowClusterConfig.storage === undefined) {
             const nfsKubeflowClusterConfig: KubeflowClusterConfigNFS = <KubeflowClusterConfigNFS>this.kubeflowClusterConfig;
             // Creat work dir for current trial in NFS directory

src/nni_manager/training_service/kubernetes/kubernetesConfig.ts

@@ -75,16 +75,19 @@ export class KubernetesClusterConfigNFS extends KubernetesClusterConfig {
 export class KubernetesClusterConfigAzure extends KubernetesClusterConfig {
     public readonly keyVault: KeyVaultConfig;
     public readonly azureStorage: AzureStorage;
+    public readonly uploadRetryCount: number | undefined;
 
     constructor(
             apiVersion: string,
             keyVault: KeyVaultConfig,
             azureStorage: AzureStorage,
-            storage?: KubernetesStorageKind
+            storage?: KubernetesStorageKind,
+            uploadRetryCount?: number
         ) {
         super(apiVersion, storage);
         this.keyVault = keyVault;
         this.azureStorage = azureStorage;
+        this.uploadRetryCount = uploadRetryCount;
     }
 
     public get storageType(): KubernetesStorageKind {
@@ -98,7 +101,8 @@ export class KubernetesClusterConfigAzure extends KubernetesClusterConfig {
             kubernetesClusterConfigObjectAzure.apiVersion,
             kubernetesClusterConfigObjectAzure.keyVault,
             kubernetesClusterConfigObjectAzure.azureStorage,
-            kubernetesClusterConfigObjectAzure.storage
+            kubernetesClusterConfigObjectAzure.storage,
+            kubernetesClusterConfigObjectAzure.uploadRetryCount
         );
     }
 }

src/nni_manager/training_service/kubernetes/kubernetesTrainingService.ts

@@ -31,13 +31,14 @@ import { getLogger, Logger } from '../../common/log';
 import {
     NNIManagerIpConfig, TrialJobDetail, TrialJobMetric
 } from '../../common/trainingService';
-import { getExperimentRootDir, getIPV4Address, getJobCancelStatus, getVersion, uniqueString } from '../../common/utils';
+import { delay, getExperimentRootDir, getIPV4Address, getJobCancelStatus, getVersion, uniqueString } from '../../common/utils';
 import { AzureStorageClientUtility } from './azureStorageClientUtils';
 import { GeneralK8sClient, KubernetesCRDClient } from './kubernetesApiClient';
 import { KubernetesClusterConfig } from './kubernetesConfig';
 import { kubernetesScriptFormat, KubernetesTrialJobDetail } from './kubernetesData';
 import { KubernetesJobRestServer } from './kubernetesJobRestServer';
 
+var yaml = require('js-yaml');
 var fs = require('fs');
 
 /**
@@ -357,6 +358,52 @@ abstract class KubernetesTrainingService {
         );
         return registrySecretName;
     }
+
+    protected async uploadFilesToAzureStorage(trialJobId: string, trialLocalTempFolder: String, codeDir: String, uploadRetryCount: number | undefined): Promise<string> {
+        if (this.azureStorageClient === undefined) {
+            throw new Error('azureStorageClient is not initialized');
+        }
+        let trialJobOutputUrl: string = '';
+        let retryCount: number = 1;
+        if(uploadRetryCount) {
+            retryCount = uploadRetryCount;
+        }
+        let resultUploadNNIScript: boolean = false;
+        let resultUploadCodeFile: boolean = false;
+        try {
+            do {
+                //upload local files, including scripts for running the trial and configuration (e.g., hyperparameters) for the trial, to azure storage
+                if(!resultUploadNNIScript) {
+                    resultUploadNNIScript = await AzureStorageClientUtility.uploadDirectory(this.azureStorageClient,
+                        `nni/${getExperimentId()}/${trialJobId}`, this.azureStorageShare,
+                        `${trialLocalTempFolder}`);
+                }
+                //upload code files to azure storage
+                if(!resultUploadCodeFile) {
+                    resultUploadCodeFile = await AzureStorageClientUtility.uploadDirectory(this.azureStorageClient,
+                        `nni/${getExperimentId()}/${trialJobId}`, this.azureStorageShare,
+                        `${codeDir}`);
+                }
+                if (resultUploadNNIScript && resultUploadCodeFile) {
+                    trialJobOutputUrl = `https://${this.azureStorageAccountName}.file.core.windows.net/${this.azureStorageShare}` + 
+                    `/${path.join('nni', getExperimentId(), trialJobId, 'output')}`;
+                    break;
+                } else {
+                    //wait for 5 seconds to re-upload files
+                    await delay(5000);
+                    this.log.info('Upload failed, Retry: upload files to azure-storage');
+                }
+            } while (retryCount-- >= 0)
+        } catch (error) {
+            this.log.error(error);
+            //return a empty url when got error
+            return Promise.resolve("");
+        }
+        if(!trialJobOutputUrl) {
+            this.log.info(`Retry-count is used up, upload files to azureStorage for trial ${trialJobId} failed!`);
+        }
+        return Promise.resolve(trialJobOutputUrl);
+    }
      
 }
 export { KubernetesTrainingService };

src/nni_manager/training_service/pai/paiConfig.ts

@@ -39,6 +39,8 @@ export class PAITaskRole {
     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
@@ -50,7 +52,7 @@ export class PAITaskRole {
      * @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) {
+                command : string, shmMB?: number, portList?: portListMetaData[]) {
         this.name = name;
         this.taskNumber = taskNumber;
         this.cpuNumber = cpuNumber;
@@ -58,6 +60,7 @@ export class PAITaskRole {
         this.gpuNumber = gpuNumber;
         this.command = command;
         this.shmMB = shmMB;
+        this.portList = portList;
     }
 }
 
@@ -120,6 +123,16 @@ export class PAIClusterConfig {
     }
 }
 
+/**
+ * portList data structure used in PAI taskRole
+ */
+export class portListMetaData {
+    public readonly label : string = '';
+    public readonly beginAt: number = 0;
+    public readonly portNumber: number = 0;
+}
+  
+
 /**
  * PAI trial configuration
  */
@@ -134,9 +147,11 @@ export class NNIPAITrialConfig extends TrialConfig {
     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) {
+                image: string, virtualCluster?: string, shmMB?: number, authFile?: string, portList?: portListMetaData[]) {
         super(command, codeDir, gpuNum);
         this.cpuNum = cpuNum;
         this.memoryMB = memoryMB;
@@ -144,5 +159,6 @@ export class NNIPAITrialConfig extends TrialConfig {
         this.virtualCluster = virtualCluster;
         this.shmMB = shmMB;
         this.authFile = authFile;
+        this.portList = portList;
     }
 }

src/nni_manager/training_service/pai/paiTrainingService.ts

@@ -79,6 +79,7 @@ class PAITrainingService implements TrainingService {
     private logCollection: string;
     private isMultiPhase: boolean = false;
     private authFileHdfsPath: string | undefined = undefined;
+    private portList?: string | undefined;
 
     constructor() {
         this.log = getLogger();
@@ -446,6 +447,8 @@ class PAITrainingService implements TrainingService {
                 nniPaiTrialCommand,
                 // Task shared memory
                 this.paiTrialConfig.shmMB,
+                // Task portList
+                this.paiTrialConfig.portList
             )
         ];
 

src/nni_manager/yarn.lock

@@ -1410,14 +1410,7 @@ js-tokens@^4.0.0:
   version "4.0.0"
   resolved "https://registry.yarnpkg.com/js-tokens/-/js-tokens-4.0.0.tgz#19203fb59991df98e3a287050d4647cdeaf32499"
 
-js-yaml@^3.10.0:
-  version "3.12.0"
-  resolved "https://registry.yarnpkg.com/js-yaml/-/js-yaml-3.12.0.tgz#eaed656ec8344f10f527c6bfa1b6e2244de167d1"
-  dependencies:
-    argparse "^1.0.7"
-    esprima "^4.0.0"
-
-js-yaml@^3.13.1:
+js-yaml@^3.10.0, js-yaml@^3.13.1:
   version "3.13.1"
   resolved "https://registry.yarnpkg.com/js-yaml/-/js-yaml-3.13.1.tgz#aff151b30bfdfa8e49e05da22e7415e9dfa37847"
   dependencies:

src/sdk/pynni/nni/nas_utils.py

@@ -32,7 +32,7 @@ def classic_mode(
     '''Execute the chosen function and inputs directly.
     In this mode, the trial code is only running the chosen subgraph (i.e., the chosen ops and inputs),
     without touching the full model graph.'''
-    if trial._params is None:
+    if trial.get_current_parameter() is None:
         trial.get_next_parameter()
     mutable_block = trial.get_current_parameter(mutable_id)
     chosen_layer = mutable_block[mutable_layer_id]["chosen_layer"]
@@ -118,7 +118,7 @@ def oneshot_mode(
     The difference is that oneshot mode does not receive subgraph.
     Instead, it uses dropout to randomly dropout inputs and ops.'''
     # NNI requires to get_next_parameter before report a result. But the parameter will not be used in this mode
-    if trial._params is None:
+    if trial.get_current_parameter() is None:
         trial.get_next_parameter()
     optional_inputs = list(optional_inputs.values())
     inputs_num = len(optional_inputs)

src/sdk/pynni/nni/smartparam.py

@@ -189,6 +189,6 @@ else:
         raise RuntimeError('Unrecognized mode: %s' % mode)
 
     def _get_param(key):
-        if trial._params is None:
+        if trial.get_current_parameter() is None:
             trial.get_next_parameter()
         return trial.get_current_parameter(key)

src/sdk/pynni/nni/trial.py

@@ -50,10 +50,12 @@ def get_next_parameter():
         return None
     return _params['parameters']
 
-def get_current_parameter(tag):
+def get_current_parameter(tag=None):
     global _params
     if _params is None:
         return None
+    if tag is None:
+        return _params['parameters']
     return _params['parameters'][tag]
 
 def get_experiment_id():

src/webui/src/components/trial-detail/DefaultMetricPoint.tsx

@@ -85,7 +85,9 @@ class DefaultPoint extends React.Component<DefaultPointProps, DefaultPointState>
             });
             // deal with best metric line
             const bestCurve: Array<number | object>[] = []; // best curve data source
-            bestCurve.push([lineListDefault[0][0], lineListDefault[0][1], accSource[0].searchSpace]);
+            if (lineListDefault[0] !== undefined) {
+                bestCurve.push([lineListDefault[0][0], lineListDefault[0][1], accSource[0].searchSpace]);
+            }
             if (optimize === 'maximize') {
                 for (let i = 1; i < lineListDefault.length; i++) {
                     const val = lineListDefault[i][1];

src/webui/src/static/style/table.scss

@@ -115,6 +115,7 @@
 }
 
 #detail-button{
+    margin: 2px 0;
     .common-style, .common-style:visited, .common-style:focus{
         height: 26px;
         border: none;
@@ -131,7 +132,7 @@
     .common-style:disabled{
         background-color: #f4f4f4;
     }
-    .special, .special:visited, .special:focus{
+    .special, .special:visited, .special:focus, .special button{
         height: 26px;
         border: none;
         border-radius: 0;
@@ -146,7 +147,7 @@
         background-color: #c8c8c8;
         outline: 0;
     }
-    .special:disabled{
+    .special:disabled, .special button:disabled{
         background-color: #f4f4f4;
         color: #d9d9d9;
     }

tools/nni_cmd/config_schema.py

@@ -240,7 +240,12 @@ pai_trial_schema = {
     Optional('outputDir'): And(Regex(r'hdfs://(([0-9]{1,3}.){3}[0-9]{1,3})(:[0-9]{2,5})?(/.*)?'),\
                          error='ERROR: outputDir format error, outputDir format is hdfs://xxx.xxx.xxx.xxx:xxx'),
     Optional('virtualCluster'): setType('virtualCluster', str),
-    Optional('nasMode'): setChoice('nasMode', 'classic_mode', 'enas_mode', 'oneshot_mode', 'darts_mode')
+    Optional('nasMode'): setChoice('nasMode', 'classic_mode', 'enas_mode', 'oneshot_mode', 'darts_mode'),
+    Optional('portList'): [{
+        "label": setType('label', str),
+        "beginAt": setType('beginAt', int),
+        "portNumber": setType('portNumber', int)
+    }]
     }
 }
 
@@ -310,7 +315,8 @@ kubeflow_config_schema = {
                            error='ERROR: accountName format error, accountName support using (0-9|a-z|A-Z|-)'),
             'azureShare': And(Regex('([0-9]|[a-z]|[A-Z]|-){3,63}'),\
                           error='ERROR: azureShare format error, azureShare support using (0-9|a-z|A-Z|-)')
-        }
+        },
+        Optional('uploadRetryCount'): setNumberRange('uploadRetryCount', int, 1, 99999)
     })
 }
 
@@ -356,7 +362,8 @@ frameworkcontroller_config_schema = {
                            error='ERROR: accountName format error, accountName support using (0-9|a-z|A-Z|-)'),
             'azureShare': And(Regex('([0-9]|[a-z]|[A-Z]|-){3,63}'),\
                           error='ERROR: azureShare format error, azureShare support using (0-9|a-z|A-Z|-)')
-        }
+        },
+        Optional('uploadRetryCount'): setNumberRange('uploadRetryCount', int, 1, 99999)
     })
 }
 

tools/nni_cmd/launcher_utils.py

@@ -198,10 +198,7 @@ def validate_common_content(experiment_config):
                     Schema({**separate_schema_dict[separate_key]['customized']}).validate(experiment_config[separate_key])
     except SchemaError as error:
         print_error('Your config file is not correct, please check your config file content!')
-        if error.__str__().__contains__('Wrong key'):
-            print_error(' '.join(error.__str__().split()[:3]))
-        else:
-            print_error(error)
+        print_error(error.code)
         exit(1)
 
     #set default value