"driver/driver.hip.cpp" did not exist on "120ab94aa18d00bf5fdf6b77b512a7f702425e80"
Commit d68ca0b3 authored by Shinai Yang's avatar Shinai Yang
Browse files

fix hdfsOUtputDir

parent 755ac5f0
...@@ -77,8 +77,6 @@ class PAITrainingService implements TrainingService { ...@@ -77,8 +77,6 @@ class PAITrainingService implements TrainingService {
private versionCheck: boolean = true; private versionCheck: boolean = true;
private logCollection: string; private logCollection: string;
private isMultiPhase: boolean = false; private isMultiPhase: boolean = false;
private hdfsCodeDir?: string;
private hdfsOutputDir?: string;
constructor() { constructor() {
this.log = getLogger(); this.log = getLogger();
...@@ -154,13 +152,13 @@ class PAITrainingService implements TrainingService { ...@@ -154,13 +152,13 @@ class PAITrainingService implements TrainingService {
//TODO: use HDFS working folder instead //TODO: use HDFS working folder instead
const trialWorkingFolder: string = path.join(this.expRootDir, 'trials', trialJobId); const trialWorkingFolder: string = path.join(this.expRootDir, 'trials', trialJobId);
const paiJobName: string = `nni_exp_${this.experimentId}_trial_${trialJobId}`; const paiJobName: string = `nni_exp_${this.experimentId}_trial_${trialJobId}`;
this.hdfsCodeDir = HDFSClientUtility.getHdfsTrialWorkDir(this.paiClusterConfig.userName, trialJobId); const hdfsCodeDir: string = HDFSClientUtility.getHdfsTrialWorkDir(this.paiClusterConfig.userName, trialJobId);
this.hdfsOutputDir = unixPathJoin(this.hdfsCodeDir, 'nnioutput'); const hdfsOutputDir: string = unixPathJoin(hdfsCodeDir, 'nnioutput');
const hdfsLogPath : string = String.Format( const hdfsLogPath : string = String.Format(
PAI_LOG_PATH_FORMAT, PAI_LOG_PATH_FORMAT,
this.paiClusterConfig.host, this.paiClusterConfig.host,
this.hdfsOutputDir hdfsOutputDir
); );
const trialJobDetail: PAITrialJobDetail = new PAITrialJobDetail( const trialJobDetail: PAITrialJobDetail = new PAITrialJobDetail(
...@@ -365,14 +363,6 @@ class PAITrainingService implements TrainingService { ...@@ -365,14 +363,6 @@ class PAITrainingService implements TrainingService {
throw new Error('PAI token is not initialized'); throw new Error('PAI token is not initialized');
} }
if (this.hdfsCodeDir === undefined) {
throw new Error('hdfsCodeDir is not initialized');
}
if (this.hdfsOutputDir === undefined) {
throw new Error('hdfsOutputDir is not initialized');
}
if (this.paiRestServerPort === undefined) { if (this.paiRestServerPort === undefined) {
const restServer: PAIJobRestServer = component.get(PAIJobRestServer); const restServer: PAIJobRestServer = component.get(PAIJobRestServer);
this.paiRestServerPort = restServer.clusterRestServerPort; this.paiRestServerPort = restServer.clusterRestServerPort;
...@@ -401,7 +391,8 @@ class PAITrainingService implements TrainingService { ...@@ -401,7 +391,8 @@ class PAITrainingService implements TrainingService {
trialForm.hyperParameters.value, { encoding: 'utf8' } trialForm.hyperParameters.value, { encoding: 'utf8' }
); );
} }
const hdfsCodeDir: string = HDFSClientUtility.getHdfsTrialWorkDir(this.paiClusterConfig.userName, trialJobId);
const hdfsOutputDir: string = unixPathJoin(hdfsCodeDir, 'nnioutput');
// tslint:disable-next-line: strict-boolean-expressions // tslint:disable-next-line: strict-boolean-expressions
const nniManagerIp: string = this.nniManagerIpConfig ? this.nniManagerIpConfig.nniManagerIp : getIPV4Address(); const nniManagerIp: string = this.nniManagerIpConfig ? this.nniManagerIpConfig.nniManagerIp : getIPV4Address();
const version: string = this.versionCheck ? await getVersion() : ''; const version: string = this.versionCheck ? await getVersion() : '';
...@@ -417,7 +408,7 @@ class PAITrainingService implements TrainingService { ...@@ -417,7 +408,7 @@ class PAITrainingService implements TrainingService {
this.paiTrialConfig.command, this.paiTrialConfig.command,
nniManagerIp, nniManagerIp,
this.paiRestServerPort, this.paiRestServerPort,
this.hdfsOutputDir, hdfsOutputDir,
this.paiClusterConfig.host, this.paiClusterConfig.host,
this.paiClusterConfig.userName, this.paiClusterConfig.userName,
HDFSClientUtility.getHdfsExpCodeDir(this.paiClusterConfig.userName), HDFSClientUtility.getHdfsExpCodeDir(this.paiClusterConfig.userName),
...@@ -452,7 +443,7 @@ class PAITrainingService implements TrainingService { ...@@ -452,7 +443,7 @@ class PAITrainingService implements TrainingService {
// Docker image // Docker image
this.paiTrialConfig.image, this.paiTrialConfig.image,
// codeDir // codeDir
`$PAI_DEFAULT_FS_URI${this.hdfsCodeDir}`, `$PAI_DEFAULT_FS_URI${hdfsCodeDir}`,
// PAI Task roles // PAI Task roles
paiTaskRoles, paiTaskRoles,
// Add Virutal Cluster // Add Virutal Cluster
...@@ -463,9 +454,9 @@ class PAITrainingService implements TrainingService { ...@@ -463,9 +454,9 @@ class PAITrainingService implements TrainingService {
// Step 2. Upload code files in codeDir onto HDFS // Step 2. Upload code files in codeDir onto HDFS
try { try {
await HDFSClientUtility.copyDirectoryToHdfs(trialLocalTempFolder, this.hdfsCodeDir, this.hdfsClient); await HDFSClientUtility.copyDirectoryToHdfs(trialLocalTempFolder, hdfsCodeDir, this.hdfsClient);
} catch (error) { } catch (error) {
this.log.error(`PAI Training service: copy ${this.paiTrialConfig.codeDir} to HDFS ${this.hdfsCodeDir} failed, error is ${error}`); this.log.error(`PAI Training service: copy ${this.paiTrialConfig.codeDir} to HDFS ${hdfsCodeDir} failed, error is ${error}`);
trialJobDetail.status = 'FAILED'; trialJobDetail.status = 'FAILED';
deferred.resolve(true); deferred.resolve(true);
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment