API: Add WorkerStatus to GetMetrics and remove unused items (#110)

* update API Signed-off-by: YujiOshima <yuji.oshima0x3fd@gmail.com> * add status to GetMetrics and delete unused item in API Signed-off-by: YujiOshima <yuji.oshima0x3fd@gmail.com>
kubeflow · Jun 13, 2018 · 14963b0 · 14963b0
1 parent aa1d592
commit 14963b0
Show file tree

Hide file tree

Showing 13 changed files with 2,221 additions and 781 deletions.
diff --git a/cmd/manager/main.go b/cmd/manager/main.go
@@ -87,21 +87,10 @@ func (s *server) GetTrials(ctx context.Context, in *pb.GetTrialsRequest) (*pb.Ge
 }
 
 func (s *server) GetSuggestions(ctx context.Context, in *pb.GetSuggestionsRequest) (*pb.GetSuggestionsReply, error) {
-	suggestAlgorithm := ""
-	if in.SuggestionAlgorithm != "" {
-		suggestAlgorithm = in.SuggestionAlgorithm
-	} else {
-		sc, err := dbIf.GetStudyConfig(in.StudyId)
-		if err != nil {
-			return &pb.GetSuggestionsReply{Trials: []*pb.Trial{}}, err
-		}
-		suggestAlgorithm = sc.DefaultSuggestionAlgorithm
-	}
-	if suggestAlgorithm == "" {
+	if in.SuggestionAlgorithm == "" {
 		return &pb.GetSuggestionsReply{Trials: []*pb.Trial{}}, errors.New("No suggest algorithm specified")
 	}
-
-	conn, err := grpc.Dial("vizier-suggestion-"+suggestAlgorithm+":6789", grpc.WithInsecure())
+	conn, err := grpc.Dial("vizier-suggestion-"+in.SuggestionAlgorithm+":6789", grpc.WithInsecure())
 	if err != nil {
 		return &pb.GetSuggestionsReply{Trials: []*pb.Trial{}}, err
 	}
@@ -149,20 +138,10 @@ func (s *server) GetWorkers(ctx context.Context, in *pb.GetWorkersRequest) (*pb.
 }
 
 func (s *server) GetShouldStopWorkers(ctx context.Context, in *pb.GetShouldStopWorkersRequest) (*pb.GetShouldStopWorkersReply, error) {
-	EarlyStoppingAlgorithm := ""
-	if in.EarlyStoppingAlgorithm != "" {
-		EarlyStoppingAlgorithm = in.EarlyStoppingAlgorithm
-	} else {
-		sc, err := dbIf.GetStudyConfig(in.StudyId)
-		if err != nil {
-			return &pb.GetShouldStopWorkersReply{}, err
-		}
-		EarlyStoppingAlgorithm = sc.DefaultEarlyStoppingAlgorithm
-	}
-	if EarlyStoppingAlgorithm == "" {
+	if in.EarlyStoppingAlgorithm == "" {
 		return &pb.GetShouldStopWorkersReply{}, errors.New("No EarlyStopping Algorithm specified")
 	}
-	conn, err := grpc.Dial("vizier-earlystopping-"+EarlyStoppingAlgorithm+":6789", grpc.WithInsecure())
+	conn, err := grpc.Dial("vizier-earlystopping-"+in.EarlyStoppingAlgorithm+":6789", grpc.WithInsecure())
 	if err != nil {
 		return &pb.GetShouldStopWorkersReply{}, err
 	}
@@ -188,9 +167,14 @@ func (s *server) GetMetrics(ctx context.Context, in *pb.GetMetricsRequest) (*pb.
 	}
 	mls := make([]*pb.MetricsLogSet, len(in.WorkerIds))
 	for i, w := range in.WorkerIds {
+		wr, err := s.GetWorkers(ctx, &pb.GetWorkersRequest{WorkerId: w})
+		if err != nil {
+			return &pb.GetMetricsReply{}, err
+		}
 		mls[i] = &pb.MetricsLogSet{
-			WorkerId:    w,
-			MetricsLogs: make([]*pb.MetricsLog, len(mNames)),
+			WorkerId:     w,
+			MetricsLogs:  make([]*pb.MetricsLog, len(mNames)),
+			WorkerStatus: wr.Workers[0].Status,
 		}
 		for j, m := range mNames {
 			ls, err := dbIf.GetWorkerLogs(w, &kdb.GetWorkerLogOpts{Name: m})

diff --git a/examples/GKEDemo/git-issue-summarize-demo.go b/examples/GKEDemo/git-issue-summarize-demo.go
@@ -11,13 +11,11 @@ import (
 )
 
 var studyConfig = api.StudyConfig{
-	Name:                          "grid-demo",
-	Owner:                         "katib",
-	OptimizationType:              api.OptimizationType_MAXIMIZE,
-	OptimizationGoal:              0.99,
-	DefaultSuggestionAlgorithm:    "grid",
-	DefaultEarlyStoppingAlgorithm: "medianstopping",
-	ObjectiveValueName:            "Validation-accuracy",
+	Name:               "grid-demo",
+	Owner:              "katib",
+	OptimizationType:   api.OptimizationType_MAXIMIZE,
+	OptimizationGoal:   0.99,
+	ObjectiveValueName: "Validation-accuracy",
 	Metrics: []string{
 		"accuracy",
 	},

diff --git a/examples/client-example.go b/examples/client-example.go
@@ -39,7 +39,7 @@ func main() {
 	//RunTrials
 	workerIds := runTrials(c, studyId, getSuggestReply)
 
-	for !isCompletedAllWorker(c, studyId) {
+	for true {
 		time.Sleep(10 * time.Second)
 		getMetricsRequest := &api.GetMetricsRequest{
 			StudyId:   studyId,
@@ -52,6 +52,9 @@ func main() {
 		}
 		//Save or Update model on ModelDB
 		SaveOrUpdateModel(c, getMetricsReply)
+		if isCompletedAllWorker(c, getMetricsReply.MetricsLogSets) {
+			break
+		}
 	}
 }
 
@@ -217,15 +220,15 @@ func SaveOrUpdateModel(c api.ManagerClient, getMetricsReply *api.GetMetricsReply
 	}
 }
 
-func isCompletedAllWorker(c api.ManagerClient, studyId string) bool {
-	ctx := context.Background()
-	getWorkerRequest := &api.GetWorkersRequest{StudyId: studyId}
-	getWorkerReply, err := c.GetWorkers(ctx, getWorkerRequest)
-	if err != nil {
-		log.Fatalf("GetWorker Error %v", err)
-	}
-	for _, w := range getWorkerReply.Workers {
-		if w.Status != api.State_COMPLETED {
+func isCompletedAllWorker(c api.ManagerClient, ms []*api.MetricsLogSet) bool {
+	//	ctx := context.Background()
+	//	getWorkerRequest := &api.GetWorkersRequest{StudyId: studyId}
+	//	getWorkerReply, err := c.GetWorkers(ctx, getWorkerRequest)
+	//	if err != nil {
+	//		log.Fatalf("GetWorker Error %v", err)
+	//	}
+	for _, mls := range ms {
+		if mls.WorkerStatus != api.State_COMPLETED {
 			return false
 		}
 	}

diff --git a/examples/study-config.yml b/examples/study-config.yml
@@ -7,8 +7,8 @@ parameterconfigs:
   - name: --lr
     parametertype: 1
     feasible:
-      min: "0.03"
-      max: "0.07"
+      min: "0.01"
+      max: "0.03"
   - name: --num-layers
     parametertype: 2
     feasible:
@@ -21,8 +21,6 @@ parameterconfigs:
       - sgd
       - adam
       - ftrl
-defaultsuggestionalgorithm: random
-defaultearlystoppingalgorithm: medianstopping
 objectivevaluename: Validation-accuracy
 metrics:
 - accuracy