Merge image namespace to 1.9.2 (#418)

* fix: use opencsghq namespace

* Remove unused type parameter for deploy uages

* support score model inference for dataflow

* support ernie and hunyuan model with cuda 11.8

* Fix test

---------

Co-authored-by: James <[email protected]>
Co-authored-by: zhzhang <[email protected]>

Author: 3 people

api/handler/monitor.go

@@ -218,18 +218,13 @@ func getRequestParameters(ctx *gin.Context) (*types.MonitorReq, error) {
 	if len(instance) < 1 {
 		return nil, fmt.Errorf("bad request format for instance")
 	}
-	deployType := ctx.Param("type")
-	if len(deployType) < 1 {
-		return nil, fmt.Errorf("bad request format for type")
-	}
 	lastDuration, timeRange := common.GetValidTimeDuration(ctx)
 	req := &types.MonitorReq{
 		CurrentUser:  currentUser,
 		Namespace:    namespace,
 		Name:         name,
 		RepoType:     repoType,
 		DeployID:     deployID,
-		DeployType:   deployType,
 		Instance:     instance,
 		LastDuration: lastDuration,
 		TimeRange:    timeRange,

api/handler/monitor_test.go

@@ -54,7 +54,6 @@ func TestMonitorHandler_CPUUsage(t *testing.T) {
 		Instance:     "test-instance",
 		LastDuration: "30m",
 		TimeRange:    "1m",
-		DeployType:   "inference",
 	}
 
 	tester.mocks.component.EXPECT().CPUUsage(tester.Ctx(), req).Return(&types.MonitorCPUResp{}, nil)
@@ -78,7 +77,6 @@ func TestMonitorHandler_MemoryUsage(t *testing.T) {
 		Instance:     "test-instance",
 		LastDuration: "30m",
 		TimeRange:    "1m",
-		DeployType:   "inference",
 	}
 
 	tester.mocks.component.EXPECT().MemoryUsage(tester.Ctx(), req).Return(&types.MonitorMemoryResp{}, nil)
@@ -102,7 +100,6 @@ func TestMonitorHandler_RequestCount(t *testing.T) {
 		Instance:     "test-instance",
 		LastDuration: "30m",
 		TimeRange:    "1m",
-		DeployType:   "inference",
 	}
 
 	tester.mocks.component.EXPECT().RequestCount(tester.Ctx(), req).Return(&types.MonitorRequestCountResp{}, nil)
@@ -125,7 +122,6 @@ func TestMonitorHandler_RequestLatency(t *testing.T) {
 		Instance:     "test-instance",
 		LastDuration: "30m",
 		TimeRange:    "1m",
-		DeployType:   "inference",
 	}
 
 	tester.mocks.component.EXPECT().RequestLatency(tester.Ctx(), req).Return(&types.MonitorRequestLatencyResp{}, nil)

builder/deploy/scheduler/deploy_runner.go

@@ -348,18 +348,19 @@ func (t *DeployRunner) makeDeployEnv(
 
 	if deploy.SpaceID > 0 {
 		// sdk port for space
-		if t.repo.Sdk == types.GRADIO.Name {
+		switch t.repo.Sdk {
+		case types.GRADIO.Name:
 			envMap["port"] = strconv.Itoa(types.GRADIO.Port)
-		} else if t.repo.Sdk == types.STREAMLIT.Name {
+		case types.STREAMLIT.Name:
 			envMap["port"] = strconv.Itoa(types.STREAMLIT.Port)
-		} else if t.repo.Sdk == types.NGINX.Name {
+		case types.NGINX.Name:
 			envMap["port"] = strconv.Itoa(types.NGINX.Port)
-		} else if t.repo.Sdk == types.DOCKER.Name {
+		case types.DOCKER.Name:
 			envMap["port"] = strconv.Itoa(deploy.ContainerPort)
 			envMap["HF_ENDPOINT"] = t.deployCfg.ModelDownloadEndpoint
-		} else if t.repo.Sdk == types.MCPSERVER.Name {
+		case types.MCPSERVER.Name:
 			envMap["port"] = strconv.Itoa(types.MCPSERVER.Port)
-		} else {
+		default:
 			envMap["port"] = strconv.Itoa(types.DefaultContainerPort)
 		}
 	}
@@ -378,6 +379,8 @@ func (t *DeployRunner) makeDeployEnv(
 		envMap["HF_TOKEN"] = token.Token
 		envMap["USE_CSGHUB_MODEL"] = "1"
 		envMap["USE_CSGHUB_DATASET"] = "1"
+		envMap["JUPYTER_ENABLE_LAB"] = "yes"
+		envMap["TERM"] = "xterm-256color"
 	}
 
 	if t.deployCfg.PublicRootDomain == "" {

builder/store/database/cluster.go

@@ -47,17 +47,16 @@ type ClusterInfo struct {
 func (r *clusterInfoStoreImpl) Add(ctx context.Context, clusterConfig string, region string) (*ClusterInfo, error) {
 	cluster, err := r.ByClusterConfig(ctx, clusterConfig)
 	if errors.Is(err, sql.ErrNoRows) {
-		newCluster := &ClusterInfo{
+		cluster = ClusterInfo{
 			ClusterID:     uuid.New().String(),
 			ClusterConfig: clusterConfig,
 			Region:        region,
 			Enable:        true,
 		}
-		_, err = r.db.Operator.Core.NewInsert().Model(newCluster).Exec(ctx)
+		_, err = r.db.Operator.Core.NewInsert().Model(&cluster).Exec(ctx)
 		if err != nil {
 			return nil, err
 		}
-		return newCluster, nil
 	}
 	return &cluster, err
 }

builder/store/database/migrations/20250717025330_migrate_runtime_opencsghq.down.sql

@@ -0,0 +1,6 @@
+SET statement_timeout = 0;
+
+--bun:split
+
+SELECT 1
+

builder/store/database/migrations/20250717025330_migrate_runtime_opencsghq.up.sql

@@ -0,0 +1,10 @@
+SET statement_timeout = 0;
+
+--bun:split
+
+-- migrate all image to opencsghq namespace since csghub 2.0
+DELETE FROM runtime_frameworks WHERE frame_image NOT LIKE '%/%';
+
+--bun:split
+
+DELETE FROM runtime_architectures WHERE runtime_framework_id NOT IN (SELECT id FROM runtime_frameworks);

common/config/config.go

@@ -143,7 +143,7 @@ type Config struct {
 	Model struct {
 		DeployTimeoutInMin      int    `env:"STARHUB_SERVER_MODEL_DEPLOY_TIMEOUT_IN_MINUTES" default:"60"`
 		DownloadEndpoint        string `env:"STARHUB_SERVER_MODEL_DOWNLOAD_ENDPOINT" default:"https://hub.opencsg.com"`
-		DockerRegBase           string `env:"STARHUB_SERVER_MODEL_DOCKER_REG_BASE" default:"opencsg-registry.cn-beijing.cr.aliyuncs.com/public/"`
+		DockerRegBase           string `env:"STARHUB_SERVER_MODEL_DOCKER_REG_BASE" default:"opencsg-registry.cn-beijing.cr.aliyuncs.com"`
 		NimDockerSecretName     string `env:"STARHUB_SERVER_MODEL_NIM_DOCKER_SECRET_NAME" default:"ngc-secret"`
 		NimNGCSecretName        string `env:"STARHUB_SERVER_MODEL_NIM_NGC_SECRET_NAME" default:"nvidia-nim-secrets"`
 		MinContextForEstimation int    `env:"STARHUB_SERVER_MODEL_MIN_CONTEXT_FOR_ESTIMATION" default:"8192"`

configs/evaluation/evalscope.json

@@ -5,12 +5,12 @@
   "engine_images": [
     {
       "compute_type": "cpu",
-      "image": "evalscope:0.15.1-cpu",
+      "image": "opencsghq/evalscope:0.15.1-cpu",
       "engine_version": "0.15.1"
     },
     {
       "compute_type": "gpu",
-      "image": "evalscope:0.15.1-cu120",
+      "image": "opencsghq/evalscope:0.15.1-cu120",
       "driver_version": "12.1",
       "engine_version": "0.15.1"
     }

configs/evaluation/lm-evaluation-harness.json

@@ -6,7 +6,7 @@
   "engine_images": [
     {
       "compute_type": "gpu",
-      "image": "lm-evaluation-harness:0.4.9",
+      "image": "opencsghq/lm-evaluation-harness:0.4.9",
       "driver_version": "12.1",
       "engine_version": "0.4.9"
     }

configs/evaluation/opencompass.json

@@ -6,7 +6,7 @@
   "engine_images": [
     {
       "compute_type": "gpu",
-      "image": "opencompass:0.4.2",
+      "image": "opencsghq/opencompass:0.4.2",
       "driver_version": "12.1",
       "engine_version": "0.4.2"
     }