Support multiphase in PAIK8S (#2368)

microsoft · Apr 30, 2020 · 2b77ab2 · 2b77ab2
1 parent 8630055
commit 2b77ab2
Show file tree

Hide file tree

Showing 3 changed files with 16 additions and 12 deletions.
diff --git a/src/nni_manager/training_service/pai/paiK8S/paiK8STrainingService.ts b/src/nni_manager/training_service/pai/paiK8S/paiK8STrainingService.ts
@@ -100,12 +100,15 @@ class PAIK8STrainingService extends PAITrainingService {
         }
     }
 
-    //TODO: update trial parameters
+    // update trial parameters for multi-phase
     public async updateTrialJob(trialJobId: string, form: TrialJobApplicationForm): Promise<TrialJobDetail> {
-        const trialJobDetail: undefined | TrialJobDetail = this.trialJobsMap.get(trialJobId);
+        const trialJobDetail: PAITrialJobDetail | undefined = this.trialJobsMap.get(trialJobId);
         if (trialJobDetail === undefined) {
             throw new Error(`updateTrialJob failed: ${trialJobId} not found`);
         }
+        // Write file content ( parameter.cfg ) to working folders
+        await this.writeParameterFile(trialJobDetail.logPath, form.hyperParameters);
+
         return trialJobDetail;
     }
 
@@ -230,24 +233,20 @@ class PAIK8STrainingService extends PAITrainingService {
         this.paiRestServerPort = this.paiJobRestServer.clusterRestServerPort;
 
         // Step 1. Prepare PAI job configuration
-        const trialLocalFolder: string = path.join(this.paiTrialConfig.nniManagerNFSMountPath, this.experimentId, trialJobId);
         //create trial local working folder locally.
-        await execMkdir(trialLocalFolder);
+        await execMkdir(trialJobDetail.logPath);
 
         const runScriptContent: string = CONTAINER_INSTALL_NNI_SHELL_FORMAT;
         // Write NNI installation file to local files
-        await fs.promises.writeFile(path.join(trialLocalFolder, 'install_nni.sh'), runScriptContent, { encoding: 'utf8' });
+        await fs.promises.writeFile(path.join(trialJobDetail.logPath, 'install_nni.sh'), runScriptContent, { encoding: 'utf8' });
 
         // Write file content ( parameter.cfg ) to local working folders
         if (trialJobDetail.form !== undefined) {
-            await fs.promises.writeFile(
-                path.join(trialLocalFolder, generateParamFileName(trialJobDetail.form.hyperParameters)),
-                trialJobDetail.form.hyperParameters.value, { encoding: 'utf8' }
-            );
+            await this.writeParameterFile(trialJobDetail.logPath, trialJobDetail.form.hyperParameters);
         }
 
         //Copy codeDir files to local working folder
-        await execCopydir(this.paiTrialConfig.codeDir, trialLocalFolder);
+        await execCopydir(this.paiTrialConfig.codeDir, trialJobDetail.logPath);
 
         const nniManagerIp: string = this.nniManagerIpConfig ? this.nniManagerIpConfig.nniManagerIp : getIPV4Address();
         const version: string = this.versionCheck ? await getVersion() : '';
@@ -298,6 +297,11 @@ class PAIK8STrainingService extends PAITrainingService {
 
         return deferred.promise;
     }
+
+    private async writeParameterFile(directory: string, hyperParameters: HyperParameters): Promise<void> {
+        const filepath: string = path.join(directory, generateParamFileName(hyperParameters));
+        await fs.promises.writeFile(filepath, hyperParameters.value, { encoding: 'utf8' });
+    }
 }
 
 export { PAIK8STrainingService };
diff --git a/test/pipelines/pipelines-it-pai-windows.yml b/test/pipelines/pipelines-it-pai-windows.yml
@@ -70,5 +70,5 @@ jobs:
       python --version
       mount -o anon $(pai_nfs_uri) $(local_nfs_uri)
       python nni_test/nnitest/generate_ts_config.py --ts pai --pai_token $(pai_token) --pai_host $(pai_host) --pai_user $(pai_user) --nni_docker_image $(docker_image) --pai_storage_plugin $(pai_storage_plugin) --nni_manager_nfs_mount_path $(nni_manager_nfs_mount_path) --container_nfs_mount_path $(container_nfs_mount_path) --nni_manager_ip $(nni_manager_ip)
-      python nni_test/nnitest/run_tests.py --config config/integration_tests.yml --ts pai --exclude multi-phase
+      python nni_test/nnitest/run_tests.py --config config/integration_tests.yml --ts pai
     displayName: 'Examples and advanced features tests on pai'
diff --git a/test/pipelines/pipelines-it-pai.yml b/test/pipelines/pipelines-it-pai.yml
@@ -57,5 +57,5 @@ jobs:
       cd test
       python3 nni_test/nnitest/generate_ts_config.py --ts pai --pai_host $(pai_host) --pai_user $(pai_user) --nni_docker_image $TEST_IMG --pai_storage_plugin $(pai_storage_plugin)\
        --pai_token $(pai_token) --nni_manager_nfs_mount_path $(nni_manager_nfs_mount_path) --container_nfs_mount_path $(container_nfs_mount_path) --nni_manager_ip $(nni_manager_ip)
-      PATH=$HOME/.local/bin:$PATH python3 nni_test/nnitest/run_tests.py --config config/integration_tests.yml --ts pai --exclude multi-phase
+      PATH=$HOME/.local/bin:$PATH python3 nni_test/nnitest/run_tests.py --config config/integration_tests.yml --ts pai
     displayName: 'integration test'