modify run/resume/view api & fix logDir rewrite (#3545)

nni/experiment/experiment.py

@@ -40,7 +40,7 @@ class Experiment:
         """
         Prepare an experiment.
 
-        Use `Experiment.start()` to launch it.
+        Use `Experiment.run()` to launch it.
 
         Parameters
         ----------
@@ -60,7 +60,7 @@ class Experiment:
             experiment.config.trial_command = 'python3 trial.py'
             experiment.config.machines.append(RemoteMachineConfig(ip=..., user_name=...))
             ...
-            experiment.start(8080)
+            experiment.run(8080)
 
         Parameters
         ----------
@@ -149,27 +149,30 @@ class Experiment:
         self._proc = None
         _logger.info('Experiment stopped')
 
-    def run(self, port: int = 8080, debug: bool = False) -> bool:
+    def run(self, port: int = 8080, wait_completion: bool = True, debug: bool = False) -> bool:
         """
         Run the experiment.
 
-        This function will block until experiment finish or error.
+        If wait_completion is True, this function will block until experiment finish or error.
 
         Return `True` when experiment done; or return `False` when experiment failed.
+
+        Else if wait_completion is False, this function will non-block and return None immediately.
         """
         self.start(port, debug)
-        try:
-            while True:
-                time.sleep(10)
-                status = self.get_status()
-                if status == 'DONE' or status == 'STOPPED':
-                    return True
-                if status == 'ERROR':
-                    return False
-        except KeyboardInterrupt:
-            _logger.warning('KeyboardInterrupt detected')
-        finally:
-            self.stop()
+        if wait_completion:
+            try:
+                while True:
+                    time.sleep(10)
+                    status = self.get_status()
+                    if status == 'DONE' or status == 'STOPPED':
+                        return True
+                    if status == 'ERROR':
+                        return False
+            except KeyboardInterrupt:
+                _logger.warning('KeyboardInterrupt detected')
+            finally:
+                self.stop()
 
     @classmethod
     def connect(cls, port: int):
@@ -194,7 +197,7 @@ class Experiment:
         return experiment
 
     @classmethod
-    def resume(cls, experiment_id: str, port: int, wait_completion: bool = True, debug: bool = False):
+    def resume(cls, experiment_id: str, port: int = 8080, wait_completion: bool = True, debug: bool = False):
         """
         Resume a stopped experiment.
 
@@ -212,14 +215,12 @@ class Experiment:
         experiment = Experiment()
         experiment.id = experiment_id
         experiment.mode = 'resume'
-        if wait_completion:
-            experiment.run(port, debug)
-        else:
-            experiment.start(port, debug)
+        experiment.run(port=port, wait_completion=wait_completion, debug=debug)
+        if not wait_completion:
             return experiment
 
     @classmethod
-    def view(cls, experiment_id: str, port: int, wait_completion: bool = True):
+    def view(cls, experiment_id: str, port: int = 8080, non_blocking: bool = False):
         """
         View a stopped experiment.
 
@@ -229,18 +230,24 @@ class Experiment:
             The stopped experiment id.
         port
             The port of web UI.
-        wait_completion
-            If true, run in the foreground. If false, run in the background.
+        non_blocking
+            If false, run in the foreground. If true, run in the background.
         """
         debug = False
         experiment = Experiment()
         experiment.id = experiment_id
         experiment.mode = 'view'
-        if wait_completion:
-            experiment.run(port, debug)
-        else:
-            experiment.start(port, debug)
+        experiment.start(port=port, debug=debug)
+        if non_blocking:
             return experiment
+        else:
+            try:
+                while True:
+                    time.sleep(10)
+            except KeyboardInterrupt:
+                _logger.warning('KeyboardInterrupt detected')
+            finally:
+                experiment.stop()
 
     def get_status(self) -> str:
         """

nni/tools/nnictl/launcher.py

@@ -319,7 +319,10 @@ def launch_experiment(args, experiment_config, mode, experiment_id, config_versi
             if package_name in ['SMAC', 'BOHB', 'PPOTuner']:
                 print_error(f'The dependencies for {package_name} can be installed through pip install nni[{package_name}]')
             raise
-    log_dir = experiment_config['logDir'] if experiment_config.get('logDir') else NNI_HOME_DIR
+    if config_version == 1:
+        log_dir = experiment_config['logDir'] if experiment_config.get('logDir') else NNI_HOME_DIR
+    else:
+        log_dir = experiment_config['experimentWorkingDirectory'] if experiment_config.get('experimentWorkingDirectory') else NNI_HOME_DIR
     log_level = experiment_config['logLevel'] if experiment_config.get('logLevel') else None
     #view experiment mode do not need debug function, when view an experiment, there will be no new logs created
     foreground = False
@@ -486,8 +489,10 @@ def manage_stopped_experiment(args, mode):
     assert 'trainingService' in experiment_config or 'trainingServicePlatform' in experiment_config
     try:
         if 'trainingService' in experiment_config:
+            experiment_config['experimentWorkingDirectory'] = experiments_dict[args.id]['logDir']
             launch_experiment(args, experiment_config, mode, experiment_id, 2)
         else:
+            experiment_config['logDir'] = experiments_dict[args.id]['logDir']
             launch_experiment(args, experiment_config, mode, experiment_id, 1)
     except Exception as exception:
         restServerPid = Experiments().get_all_experiments().get(experiment_id, {}).get('pid')

ts/nni_manager/common/manager.ts

@@ -8,7 +8,7 @@ import { TrialJobStatus, LogType } from './trainingService';
 import { ExperimentConfig } from './experimentConfig';
 
 type ProfileUpdateType = 'TRIAL_CONCURRENCY' | 'MAX_EXEC_DURATION' | 'SEARCH_SPACE' | 'MAX_TRIAL_NUM';
-type ExperimentStatus = 'INITIALIZED' | 'RUNNING' | 'ERROR' | 'STOPPING' | 'STOPPED' | 'DONE' | 'NO_MORE_TRIAL' | 'TUNER_NO_MORE_TRIAL';
+type ExperimentStatus = 'INITIALIZED' | 'RUNNING' | 'ERROR' | 'STOPPING' | 'STOPPED' | 'DONE' | 'NO_MORE_TRIAL' | 'TUNER_NO_MORE_TRIAL' | 'VIEWED';
 namespace ExperimentStartUpMode {
     export const NEW = 'new';
     export const RESUME = 'resume';

ts/nni_manager/core/nnimanager.ts

@@ -204,6 +204,7 @@ class NNIManager implements Manager {
         this.experimentProfile = await this.dataStore.getExperimentProfile(experimentId);
         this.readonly = readonly;
         if (readonly) {
+            this.setStatus('VIEWED');
             return Promise.resolve();
         }
 

ts/webui/src/App.tsx

@@ -237,7 +237,7 @@ class App extends React.Component<{}, AppState> {
         }
 
         // experiment status and /trial-jobs api's status could decide website update
-        if (['DONE', 'ERROR', 'STOPPED'].includes(EXPERIMENT.status) || TRIALS.jobListError()) {
+        if (['DONE', 'ERROR', 'STOPPED', 'VIEWED'].includes(EXPERIMENT.status) || TRIALS.jobListError()) {
             // experiment finished, refresh once more to ensure consistency
             this.setState(() => ({ interval: 0, isUpdate: false }));
             return;

ts/webui/src/components/modals/ExperimentSummaryPanel.tsx

@@ -54,7 +54,7 @@ class ExperimentSummaryPanel extends React.Component<ExpDrawerProps, ExpDrawerSt
             this.setState({ experiment: JSON.stringify(result, null, 4) });
         }
 
-        if (['DONE', 'ERROR', 'STOPPED'].includes(EXPERIMENT.status)) {
+        if (['DONE', 'ERROR', 'STOPPED', 'VIEWED'].includes(EXPERIMENT.status)) {
             if (this.refreshId !== null || this.refreshId !== undefined) {
                 window.clearInterval(this.refreshId);
             }

ts/webui/src/components/trial-detail/TableList.tsx

@@ -419,7 +419,7 @@ class TableList extends React.Component<TableListProps, TableListState> {
 
     private _renderOperationColumn(record: any): React.ReactNode {
         const runningTrial: boolean = ['RUNNING', 'UNKNOWN'].includes(record.status) ? false : true;
-        const disabledAddCustomizedTrial = ['DONE', 'ERROR', 'STOPPED'].includes(EXPERIMENT.status);
+        const disabledAddCustomizedTrial = ['DONE', 'ERROR', 'STOPPED', 'VIEWED'].includes(EXPERIMENT.status);
         return (
             <Stack className='detail-button' horizontal>
                 <PrimaryButton

ts/webui/src/static/const.ts

@@ -21,6 +21,7 @@ const EXPERIMENTSTATUS = [
     'ERROR',
     'STOPPING',
     'STOPPED',
+    'VIEWED',
     'DONE',
     'NO_MORE_TRIAL',
     'TUNER_NO_MORE_TRIAL'

ts/webui/src/static/style/overview/probar.scss

@@ -19,7 +19,8 @@ $error: #a4262c;
 }
 
 .DONE,
-.STOPPED {
+.STOPPED,
+.VIEWED {
     color: $done;
 
     .ms-ProgressIndicator-progressBar {
@@ -37,7 +38,8 @@ $error: #a4262c;
 
 .bestMetric {
     .DONE,
-    .STOPPED {
+    .STOPPED,
+    .VIEWED {
         color: $done;
     }