feat: support hyperparameters in distillation tuning

PiperOrigin-RevId: 882708166

Author: speedstorm1

GeneratedFirebaseAI/Sources/Types.swift

@@ -19483,9 +19483,16 @@ extension PreferenceOptimizationSpec: Codable {
   }
 }
 
-/// Hyperparameters for Distillation. This data type is not supported in Gemini API.
+/// Hyperparameters for distillation.
 @available(iOS 15.0, macOS 13.0, macCatalyst 15.0, tvOS 15.0, watchOS 8.0, *)
 public struct DistillationHyperParameters: Sendable {
+  /// The batch size hyperparameter for tuning.
+  /// This is only supported for OSS models in Vertex.
+  public let batchSize: Int32?
+
+  /// The learning rate for tuning. OSS models only.
+  public let learningRate: Float?
+
   /// Optional. Adapter size for distillation.
   public let adapterSize: AdapterSize?
 
@@ -19498,10 +19505,14 @@ public struct DistillationHyperParameters: Sendable {
 
   /// Default initializer.
   public init(
+    batchSize: Int32? = nil,
+    learningRate: Float? = nil,
     adapterSize: AdapterSize? = nil,
     epochCount: Int64? = nil,
     learningRateMultiplier: Double? = nil
   ) {
+    self.batchSize = batchSize
+    self.learningRate = learningRate
     self.adapterSize = adapterSize
     self.epochCount = epochCount
     self.learningRateMultiplier = learningRateMultiplier
@@ -19513,6 +19524,8 @@ extension DistillationHyperParameters: Codable {
 
   // MARK: - Codable
   public enum VertexKeys: String, CodingKey {
+    case batchSize = "batchSize"
+    case learningRate = "learningRate"
     case adapterSize = "adapterSize"
     case epochCount = "epochCount"
     case learningRateMultiplier = "learningRateMultiplier"
@@ -19522,6 +19535,16 @@ extension DistillationHyperParameters: Codable {
     let configuration: APIClient = try decoder.userInfoOrThrow(.configuration)
 
     let VertexKeysContainer = try decoder.container(keyedBy: VertexKeys.self)
+    batchSize = try VertexKeysContainer.decodeIfPresent(
+      Int32.self,
+      forKey: .batchSize
+    )
+
+    learningRate = try VertexKeysContainer.decodeIfPresent(
+      Float.self,
+      forKey: .learningRate
+    )
+
     adapterSize = try VertexKeysContainer.decodeIfPresent(
       AdapterSize.self,
       forKey: .adapterSize
@@ -19544,6 +19567,16 @@ extension DistillationHyperParameters: Codable {
     if configuration.isVertexAI() {
 
       var VertexKeysContainer = encoder.container(keyedBy: VertexKeys.self)
+      try VertexKeysContainer.encodeIfPresent(
+        batchSize,
+        forKey: .batchSize
+      )
+
+      try VertexKeysContainer.encodeIfPresent(
+        learningRate,
+        forKey: .learningRate
+      )
+
       try VertexKeysContainer.encodeIfPresent(
         adapterSize,
         forKey: .adapterSize
@@ -19569,14 +19602,17 @@ public struct DistillationSpec: Sendable {
   /// The GCS URI of the prompt dataset to use during distillation.
   public let promptDatasetUri: String?
 
+  /// Tuning mode for tuning.
+  public let tuningMode: TuningMode?
+
+  /// Optional. Hyperparameters for Distillation.
+  public let hyperParameters: DistillationHyperParameters?
+
   /// The base teacher model that is being distilled. See [Supported
   /// models](https://cloud.google.com/vertex-ai/generative-ai/docs/model-
   /// reference/tuning#supported_models).
   public let baseTeacherModel: String?
 
-  /// Optional. Hyperparameters for Distillation.
-  public let hyperParameters: DistillationHyperParameters?
-
   /// Deprecated. A path in a Cloud Storage bucket, which will be treated as the root
   /// output directory of the distillation pipeline. It is used by the system to
   /// generate the paths of output artifacts.
@@ -19601,17 +19637,19 @@ public struct DistillationSpec: Sendable {
   /// Default initializer.
   public init(
     promptDatasetUri: String? = nil,
-    baseTeacherModel: String? = nil,
+    tuningMode: TuningMode? = nil,
     hyperParameters: DistillationHyperParameters? = nil,
+    baseTeacherModel: String? = nil,
     pipelineRootDirectory: String? = nil,
     studentModel: String? = nil,
     trainingDatasetUri: String? = nil,
     tunedTeacherModelSource: String? = nil,
     validationDatasetUri: String? = nil
   ) {
     self.promptDatasetUri = promptDatasetUri
-    self.baseTeacherModel = baseTeacherModel
+    self.tuningMode = tuningMode
     self.hyperParameters = hyperParameters
+    self.baseTeacherModel = baseTeacherModel
     self.pipelineRootDirectory = pipelineRootDirectory
     self.studentModel = studentModel
     self.trainingDatasetUri = trainingDatasetUri
@@ -19626,8 +19664,9 @@ extension DistillationSpec: Codable {
   // MARK: - Codable
   public enum VertexKeys: String, CodingKey {
     case promptDatasetUri = "promptDatasetUri"
-    case baseTeacherModel = "baseTeacherModel"
+    case tuningMode = "tuningMode"
     case hyperParameters = "hyperParameters"
+    case baseTeacherModel = "baseTeacherModel"
     case pipelineRootDirectory = "pipelineRootDirectory"
     case studentModel = "studentModel"
     case trainingDatasetUri = "trainingDatasetUri"
@@ -19644,16 +19683,21 @@ extension DistillationSpec: Codable {
       forKey: .promptDatasetUri
     )
 
-    baseTeacherModel = try VertexKeysContainer.decodeIfPresent(
-      String.self,
-      forKey: .baseTeacherModel
+    tuningMode = try VertexKeysContainer.decodeIfPresent(
+      TuningMode.self,
+      forKey: .tuningMode
     )
 
     hyperParameters = try VertexKeysContainer.decodeIfPresent(
       DistillationHyperParameters.self,
       forKey: .hyperParameters
     )
 
+    baseTeacherModel = try VertexKeysContainer.decodeIfPresent(
+      String.self,
+      forKey: .baseTeacherModel
+    )
+
     pipelineRootDirectory = try VertexKeysContainer.decodeIfPresent(
       String.self,
       forKey: .pipelineRootDirectory
@@ -19692,15 +19736,20 @@ extension DistillationSpec: Codable {
       )
 
       try VertexKeysContainer.encodeIfPresent(
-        baseTeacherModel,
-        forKey: .baseTeacherModel
+        tuningMode,
+        forKey: .tuningMode
       )
 
       try VertexKeysContainer.encodeIfPresent(
         hyperParameters,
         forKey: .hyperParameters
       )
 
+      try VertexKeysContainer.encodeIfPresent(
+        baseTeacherModel,
+        forKey: .baseTeacherModel
+      )
+
       try VertexKeysContainer.encodeIfPresent(
         pipelineRootDirectory,
         forKey: .pipelineRootDirectory
@@ -23809,7 +23858,7 @@ public struct CreateTuningJobConfig: Sendable {
   /// Adapter size for tuning.
   public let adapterSize: AdapterSize?
 
-  /// Tuning mode for SFT tuning.
+  /// Tuning mode for tuning.
   public let tuningMode: TuningMode?
 
   /// Custom base model for tuning. This is only supported for OSS models in Vertex.