KEP-2170: Rename TrainingRuntimeRef to RuntimeRef API (#2283)

* KEP-2170: Rename TrainingRuntimeRef to RuntimeRef API Signed-off-by: Andrey Velichkevich <[email protected]> * Rename RuntimeRef in runtime framework Signed-off-by: Andrey Velichkevich <[email protected]> --------- Signed-off-by: Andrey Velichkevich <[email protected]>
kubeflow · Oct 17, 2024 · 6965c1a · 6965c1a
1 parent 2d58b49
commit 6965c1a
Show file tree

Hide file tree

Showing 15 changed files with 134 additions and 134 deletions.
diff --git a/docs/proposals/2170-kubeflow-training-v2/README.md b/docs/proposals/2170-kubeflow-training-v2/README.md
@@ -281,7 +281,7 @@ type TrainJob struct {
 
 type TrainJobSpec struct {
 	// Reference to the training runtime.
-	TrainingRuntimeRef TrainingRuntimeRef `json:"trainingRuntimeRef"`
+	RuntimeRef RuntimeRef `json:"runtimeRef"`
 
 	// Configuration of the desired trainer.
 	Trainer *Trainer `json:"trainer,omitempty"`
@@ -317,7 +317,7 @@ type TrainJobSpec struct {
 	ManagedBy *string `json:"managedBy,omitempty"`
 }
 
-type TrainingRuntimeRef struct {
+type RuntimeRef struct {
 	// Name of the runtime being referenced.
 	// When namespaced-scoped TrainingRuntime is used, the TrainJob must have
 	// the same namespace as the deployed runtime.
@@ -375,7 +375,7 @@ This table explains the rationale for each `TrainJob` parameter:
    </td>
   </tr>
   <tr>
-   <td><code>TrainingRuntimeRef</code>
+   <td><code>RuntimeRef</code>
    </td>
    <td>Reference to the existing <code>TrainingRuntime</code> that is pre-deployed by platform engineers
    </td>
@@ -430,7 +430,7 @@ metadata:
   name: torch-ddp
   namespace: tenant-alpha
 spec:
-  trainingRuntimeRef:
+  runtimeRef:
     name: torch-distributed-multi-node
   trainer:
     image: docker.io/custom-training
@@ -488,7 +488,7 @@ metadata:
   name: tune-llama-with-yelp
   namespace: tenant-alpha
 spec:
-  trainingRuntimeRef:
+  runtimeRef:
     name: torch-tune-llama-7b
   datasetConfig:
     storageUri: s3://dataset/custom-dataset/yelp-review
@@ -890,7 +890,7 @@ metadata:
   name: pytorch-distributed
   namespace: tenant-alpha
 spec:
-  trainingRuntimeRef:
+  runtimeRef:
     name: pytorch-distributed-gpu
   trainer:
     image: docker.io/custom-training
@@ -939,7 +939,7 @@ to control versions of `TrainingRuntime` and enable rolling updates.
 
 We are going to create two CRDs: `TrainingRuntime` and `ClusterTrainingRuntime`. These runtimes have
 exactly the same APIs, but the first one is the namespace-scoped, the second is the cluster-scoped.
-User can set the `kind` and `apiGroup` parameters in the `trainingRuntimeRef` to use
+User can set the `kind` and `apiGroup` parameters in the `runtimeRef` to use
 the `TrainingRuntime` from the `TrainJob's` namespace, otherwise the `ClusterTrainingRuntime` will
 be used.
 
@@ -1228,7 +1228,7 @@ metadata:
   name: torch-test
   namespace: tenant-alpha
 spec:
-  trainingRuntimeRef:
+  runtimeRef:
     name: torch-distributed-multi-node
   trainer:
     resourcesPerNode:
@@ -1698,15 +1698,15 @@ Note that we should implement the status transitions validations to once we supp
 
 ### Support Multiple API Versions of TrainingRuntime
 
-We can consider to introduce the `version` field for runtime API version to the `.spec.trainingRuntimeRef`
+We can consider to introduce the `version` field for runtime API version to the `.spec.runtimeRef`
 so that we can support multiple API versions of TrainingRuntime.
 
 It could mitigate the pain points when users upgrade the older API Version to newer API Version like alpha to beta.
 But, we do not aim to support both Alpha and Beta versions or both first Alpha and second Alpha versions in the specific training-operator release.
 Hence, the `version` field was not introduced.
 
 ```go
-type TrainingRuntimeRef struct {
+type RuntimeRef struct {
 	[...]
 
 	// APIVersion is the apiVersion for the runtime.

diff --git a/manifests/v2/base/crds/kubeflow.org_clustertrainingruntimes.yaml b/manifests/v2/base/crds/kubeflow.org_clustertrainingruntimes.yaml
@@ -19,7 +19,7 @@ spec:
       openAPIV3Schema:
         description: |-
           ClusterTrainingRuntime represents a training runtime which can be referenced as part of
-          `trainingRuntimeRef` API in TrainJob. This resource is a cluster-scoped and can be referenced
+          `runtimeRef` API in TrainJob. This resource is a cluster-scoped and can be referenced
           by TrainJob that created in *any* namespace.
         properties:
           apiVersion:

diff --git a/manifests/v2/base/crds/kubeflow.org_trainingruntimes.yaml b/manifests/v2/base/crds/kubeflow.org_trainingruntimes.yaml
@@ -19,7 +19,7 @@ spec:
       openAPIV3Schema:
         description: |-
           TrainingRuntime represents a training runtime which can be referenced as part of
-          `trainingRuntimeRef` API in TrainJob. This resource is a namespaced-scoped and can be referenced
+          `runtimeRef` API in TrainJob. This resource is a namespaced-scoped and can be referenced
           by TrainJob that created in the *same* namespace as the TrainingRuntime.
         properties:
           apiVersion:

diff --git a/manifests/v2/base/crds/kubeflow.org_trainjobs.yaml b/manifests/v2/base/crds/kubeflow.org_trainjobs.yaml
@@ -2732,6 +2732,29 @@ spec:
                   - targetReplicatedJobs
                   type: object
                 type: array
+              runtimeRef:
+                description: Reference to the training runtime.
+                properties:
+                  apiGroup:
+                    description: |-
+                      APIGroup of the runtime being referenced.
+                      Defaults to `kubeflow.org`.
+                    type: string
+                  kind:
+                    description: |-
+                      Kind of the runtime being referenced.
+                      It must be one of TrainingRuntime or ClusterTrainingRuntime.
+                      Defaults to ClusterTrainingRuntime.
+                    type: string
+                  name:
+                    description: |-
+                      Name of the runtime being referenced.
+                      When namespaced-scoped TrainingRuntime is used, the TrainJob must have
+                      the same namespace as the deployed runtime.
+                    type: string
+                required:
+                - name
+                type: object
               suspend:
                 description: |-
                   Whether the controller should suspend the running TrainJob.
@@ -2937,31 +2960,8 @@ spec:
                         type: object
                     type: object
                 type: object
-              trainingRuntimeRef:
-                description: Reference to the training runtime.
-                properties:
-                  apiGroup:
-                    description: |-
-                      APIGroup of the runtime being referenced.
-                      Defaults to `kubeflow.org`.
-                    type: string
-                  kind:
-                    description: |-
-                      Kind of the runtime being referenced.
-                      It must be one of TrainingRuntime or ClusterTrainingRuntime.
-                      Defaults to ClusterTrainingRuntime.
-                    type: string
-                  name:
-                    description: |-
-                      Name of the runtime being referenced.
-                      When namespaced-scoped TrainingRuntime is used, the TrainJob must have
-                      the same namespace as the deployed runtime.
-                    type: string
-                required:
-                - name
-                type: object
             required:
-            - trainingRuntimeRef
+            - runtimeRef
             type: object
           status:
             description: Current status of TrainJob.

diff --git a/pkg/apis/kubeflow.org/v2alpha1/openapi_generated.go b/pkg/apis/kubeflow.org/v2alpha1/openapi_generated.go
diff --git a/pkg/apis/kubeflow.org/v2alpha1/trainingruntime_types.go b/pkg/apis/kubeflow.org/v2alpha1/trainingruntime_types.go
@@ -38,7 +38,7 @@ const (
 // +kubebuilder:resource:scope=Cluster
 
 // ClusterTrainingRuntime represents a training runtime which can be referenced as part of
-// `trainingRuntimeRef` API in TrainJob. This resource is a cluster-scoped and can be referenced
+// `runtimeRef` API in TrainJob. This resource is a cluster-scoped and can be referenced
 // by TrainJob that created in *any* namespace.
 type ClusterTrainingRuntime struct {
 	metav1.TypeMeta `json:",inline"`
@@ -72,7 +72,7 @@ type ClusterTrainingRuntimeList struct {
 // +kubebuilder:storageversion
 
 // TrainingRuntime represents a training runtime which can be referenced as part of
-// `trainingRuntimeRef` API in TrainJob. This resource is a namespaced-scoped and can be referenced
+// `runtimeRef` API in TrainJob. This resource is a namespaced-scoped and can be referenced
 // by TrainJob that created in the *same* namespace as the TrainingRuntime.
 type TrainingRuntime struct {
 	metav1.TypeMeta `json:",inline"`

diff --git a/pkg/apis/kubeflow.org/v2alpha1/trainjob_types.go b/pkg/apis/kubeflow.org/v2alpha1/trainjob_types.go
@@ -63,7 +63,7 @@ type TrainJobList struct {
 // TrainJobSpec represents specification of the desired TrainJob.
 type TrainJobSpec struct {
 	// Reference to the training runtime.
-	TrainingRuntimeRef TrainingRuntimeRef `json:"trainingRuntimeRef"`
+	RuntimeRef RuntimeRef `json:"runtimeRef"`
 
 	// Configuration of the desired trainer.
 	Trainer *Trainer `json:"trainer,omitempty"`
@@ -99,8 +99,8 @@ type TrainJobSpec struct {
 	ManagedBy *string `json:"managedBy,omitempty"`
 }
 
-// TrainingRuntimeRef represents the reference to the existing training runtime.
-type TrainingRuntimeRef struct {
+// RuntimeRef represents the reference to the existing training runtime.
+type RuntimeRef struct {
 	// Name of the runtime being referenced.
 	// When namespaced-scoped TrainingRuntime is used, the TrainJob must have
 	// the same namespace as the deployed runtime.