openai 4.97.0 → 4.99.0

package/src/resources/fine-tuning/index.ts

@@ -1,5 +1,6 @@
 // File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
 
+export { Alpha } from './alpha/index';
 export { Checkpoints } from './checkpoints/index';
 export { FineTuning } from './fine-tuning';
 export {
@@ -15,3 +16,12 @@ export {
   type JobListParams,
   type JobListEventsParams,
 } from './jobs/index';
+export {
+  Methods,
+  type DpoHyperparameters,
+  type DpoMethod,
+  type ReinforcementHyperparameters,
+  type ReinforcementMethod,
+  type SupervisedHyperparameters,
+  type SupervisedMethod,
+} from './methods';

package/src/resources/fine-tuning/jobs/checkpoints.ts

@@ -8,6 +8,16 @@ import { CursorPage, type CursorPageParams } from '../../../pagination';
 export class Checkpoints extends APIResource {
   /**
    * List checkpoints for a fine-tuning job.
+   *
+   * @example
+   * ```ts
+   * // Automatically fetches more pages as needed.
+   * for await (const fineTuningJobCheckpoint of client.fineTuning.jobs.checkpoints.list(
+   *   'ft-AF1WoRqd3aJAHsqc9NY7iL8F',
+   * )) {
+   *   // ...
+   * }
+   * ```
    */
   list(
     fineTuningJobId: string,

package/src/resources/fine-tuning/jobs/jobs.ts

@@ -3,6 +3,7 @@
 import { APIResource } from '../../../resource';
 import { isRequestOptions } from '../../../core';
 import * as Core from '../../../core';
+import * as MethodsAPI from '../methods';
 import * as CheckpointsAPI from './checkpoints';
 import {
   CheckpointListParams,
@@ -23,6 +24,14 @@ export class Jobs extends APIResource {
    * of the fine-tuned models once complete.
    *
    * [Learn more about fine-tuning](https://platform.openai.com/docs/guides/fine-tuning)
+   *
+   * @example
+   * ```ts
+   * const fineTuningJob = await client.fineTuning.jobs.create({
+   *   model: 'gpt-4o-mini',
+   *   training_file: 'file-abc123',
+   * });
+   * ```
    */
   create(body: JobCreateParams, options?: Core.RequestOptions): Core.APIPromise<FineTuningJob> {
     return this._client.post('/fine_tuning/jobs', { body, ...options });
@@ -32,6 +41,13 @@ export class Jobs extends APIResource {
    * Get info about a fine-tuning job.
    *
    * [Learn more about fine-tuning](https://platform.openai.com/docs/guides/fine-tuning)
+   *
+   * @example
+   * ```ts
+   * const fineTuningJob = await client.fineTuning.jobs.retrieve(
+   *   'ft-AF1WoRqd3aJAHsqc9NY7iL8F',
+   * );
+   * ```
    */
   retrieve(fineTuningJobId: string, options?: Core.RequestOptions): Core.APIPromise<FineTuningJob> {
     return this._client.get(`/fine_tuning/jobs/${fineTuningJobId}`, options);
@@ -39,6 +55,14 @@ export class Jobs extends APIResource {
 
   /**
    * List your organization's fine-tuning jobs
+   *
+   * @example
+   * ```ts
+   * // Automatically fetches more pages as needed.
+   * for await (const fineTuningJob of client.fineTuning.jobs.list()) {
+   *   // ...
+   * }
+   * ```
    */
   list(
     query?: JobListParams,
@@ -57,6 +81,13 @@ export class Jobs extends APIResource {
 
   /**
    * Immediately cancel a fine-tune job.
+   *
+   * @example
+   * ```ts
+   * const fineTuningJob = await client.fineTuning.jobs.cancel(
+   *   'ft-AF1WoRqd3aJAHsqc9NY7iL8F',
+   * );
+   * ```
    */
   cancel(fineTuningJobId: string, options?: Core.RequestOptions): Core.APIPromise<FineTuningJob> {
     return this._client.post(`/fine_tuning/jobs/${fineTuningJobId}/cancel`, options);
@@ -64,6 +95,16 @@ export class Jobs extends APIResource {
 
   /**
    * Get status updates for a fine-tuning job.
+   *
+   * @example
+   * ```ts
+   * // Automatically fetches more pages as needed.
+   * for await (const fineTuningJobEvent of client.fineTuning.jobs.listEvents(
+   *   'ft-AF1WoRqd3aJAHsqc9NY7iL8F',
+   * )) {
+   *   // ...
+   * }
+   * ```
    */
   listEvents(
     fineTuningJobId: string,
@@ -87,6 +128,34 @@ export class Jobs extends APIResource {
       ...options,
     });
   }
+
+  /**
+   * Pause a fine-tune job.
+   *
+   * @example
+   * ```ts
+   * const fineTuningJob = await client.fineTuning.jobs.pause(
+   *   'ft-AF1WoRqd3aJAHsqc9NY7iL8F',
+   * );
+   * ```
+   */
+  pause(fineTuningJobId: string, options?: Core.RequestOptions): Core.APIPromise<FineTuningJob> {
+    return this._client.post(`/fine_tuning/jobs/${fineTuningJobId}/pause`, options);
+  }
+
+  /**
+   * Resume a fine-tune job.
+   *
+   * @example
+   * ```ts
+   * const fineTuningJob = await client.fineTuning.jobs.resume(
+   *   'ft-AF1WoRqd3aJAHsqc9NY7iL8F',
+   * );
+   * ```
+   */
+  resume(fineTuningJobId: string, options?: Core.RequestOptions): Core.APIPromise<FineTuningJob> {
+    return this._client.post(`/fine_tuning/jobs/${fineTuningJobId}/resume`, options);
+  }
 }
 
 export class FineTuningJobsPage extends CursorPage<FineTuningJob> {}
@@ -233,7 +302,7 @@ export namespace FineTuningJob {
      * Number of examples in each batch. A larger batch size means that model
      * parameters are updated less frequently, but with lower variance.
      */
-    batch_size?: 'auto' | number;
+    batch_size?: unknown | 'auto' | number | null;
 
     /**
      * Scaling factor for the learning rate. A smaller learning rate may be useful to
@@ -253,97 +322,24 @@ export namespace FineTuningJob {
    */
   export interface Method {
     /**
-     * Configuration for the DPO fine-tuning method.
+     * The type of method. Is either `supervised`, `dpo`, or `reinforcement`.
      */
-    dpo?: Method.Dpo;
+    type: 'supervised' | 'dpo' | 'reinforcement';
 
     /**
-     * Configuration for the supervised fine-tuning method.
-     */
-    supervised?: Method.Supervised;
-
-    /**
-     * The type of method. Is either `supervised` or `dpo`.
+     * Configuration for the DPO fine-tuning method.
      */
-    type?: 'supervised' | 'dpo';
-  }
+    dpo?: MethodsAPI.DpoMethod;
 
-  export namespace Method {
     /**
-     * Configuration for the DPO fine-tuning method.
+     * Configuration for the reinforcement fine-tuning method.
      */
-    export interface Dpo {
-      /**
-       * The hyperparameters used for the fine-tuning job.
-       */
-      hyperparameters?: Dpo.Hyperparameters;
-    }
-
-    export namespace Dpo {
-      /**
-       * The hyperparameters used for the fine-tuning job.
-       */
-      export interface Hyperparameters {
-        /**
-         * Number of examples in each batch. A larger batch size means that model
-         * parameters are updated less frequently, but with lower variance.
-         */
-        batch_size?: 'auto' | number;
-
-        /**
-         * The beta value for the DPO method. A higher beta value will increase the weight
-         * of the penalty between the policy and reference model.
-         */
-        beta?: 'auto' | number;
-
-        /**
-         * Scaling factor for the learning rate. A smaller learning rate may be useful to
-         * avoid overfitting.
-         */
-        learning_rate_multiplier?: 'auto' | number;
-
-        /**
-         * The number of epochs to train the model for. An epoch refers to one full cycle
-         * through the training dataset.
-         */
-        n_epochs?: 'auto' | number;
-      }
-    }
+    reinforcement?: MethodsAPI.ReinforcementMethod;
 
     /**
      * Configuration for the supervised fine-tuning method.
      */
-    export interface Supervised {
-      /**
-       * The hyperparameters used for the fine-tuning job.
-       */
-      hyperparameters?: Supervised.Hyperparameters;
-    }
-
-    export namespace Supervised {
-      /**
-       * The hyperparameters used for the fine-tuning job.
-       */
-      export interface Hyperparameters {
-        /**
-         * Number of examples in each batch. A larger batch size means that model
-         * parameters are updated less frequently, but with lower variance.
-         */
-        batch_size?: 'auto' | number;
-
-        /**
-         * Scaling factor for the learning rate. A smaller learning rate may be useful to
-         * avoid overfitting.
-         */
-        learning_rate_multiplier?: 'auto' | number;
-
-        /**
-         * The number of epochs to train the model for. An epoch refers to one full cycle
-         * through the training dataset.
-         */
-        n_epochs?: 'auto' | number;
-      }
-    }
+    supervised?: MethodsAPI.SupervisedMethod;
   }
 }
 
@@ -597,97 +593,24 @@ export namespace JobCreateParams {
    */
   export interface Method {
     /**
-     * Configuration for the DPO fine-tuning method.
+     * The type of method. Is either `supervised`, `dpo`, or `reinforcement`.
      */
-    dpo?: Method.Dpo;
+    type: 'supervised' | 'dpo' | 'reinforcement';
 
     /**
-     * Configuration for the supervised fine-tuning method.
+     * Configuration for the DPO fine-tuning method.
      */
-    supervised?: Method.Supervised;
+    dpo?: MethodsAPI.DpoMethod;
 
     /**
-     * The type of method. Is either `supervised` or `dpo`.
+     * Configuration for the reinforcement fine-tuning method.
      */
-    type?: 'supervised' | 'dpo';
-  }
-
-  export namespace Method {
-    /**
-     * Configuration for the DPO fine-tuning method.
-     */
-    export interface Dpo {
-      /**
-       * The hyperparameters used for the fine-tuning job.
-       */
-      hyperparameters?: Dpo.Hyperparameters;
-    }
-
-    export namespace Dpo {
-      /**
-       * The hyperparameters used for the fine-tuning job.
-       */
-      export interface Hyperparameters {
-        /**
-         * Number of examples in each batch. A larger batch size means that model
-         * parameters are updated less frequently, but with lower variance.
-         */
-        batch_size?: 'auto' | number;
-
-        /**
-         * The beta value for the DPO method. A higher beta value will increase the weight
-         * of the penalty between the policy and reference model.
-         */
-        beta?: 'auto' | number;
-
-        /**
-         * Scaling factor for the learning rate. A smaller learning rate may be useful to
-         * avoid overfitting.
-         */
-        learning_rate_multiplier?: 'auto' | number;
-
-        /**
-         * The number of epochs to train the model for. An epoch refers to one full cycle
-         * through the training dataset.
-         */
-        n_epochs?: 'auto' | number;
-      }
-    }
+    reinforcement?: MethodsAPI.ReinforcementMethod;
 
     /**
      * Configuration for the supervised fine-tuning method.
      */
-    export interface Supervised {
-      /**
-       * The hyperparameters used for the fine-tuning job.
-       */
-      hyperparameters?: Supervised.Hyperparameters;
-    }
-
-    export namespace Supervised {
-      /**
-       * The hyperparameters used for the fine-tuning job.
-       */
-      export interface Hyperparameters {
-        /**
-         * Number of examples in each batch. A larger batch size means that model
-         * parameters are updated less frequently, but with lower variance.
-         */
-        batch_size?: 'auto' | number;
-
-        /**
-         * Scaling factor for the learning rate. A smaller learning rate may be useful to
-         * avoid overfitting.
-         */
-        learning_rate_multiplier?: 'auto' | number;
-
-        /**
-         * The number of epochs to train the model for. An epoch refers to one full cycle
-         * through the training dataset.
-         */
-        n_epochs?: 'auto' | number;
-      }
-    }
+    supervised?: MethodsAPI.SupervisedMethod;
   }
 }
 

package/src/resources/fine-tuning/methods.ts

@@ -0,0 +1,152 @@
+// File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
+
+import { APIResource } from '../../resource';
+import * as GraderModelsAPI from '../graders/grader-models';
+
+export class Methods extends APIResource {}
+
+/**
+ * The hyperparameters used for the DPO fine-tuning job.
+ */
+export interface DpoHyperparameters {
+  /**
+   * Number of examples in each batch. A larger batch size means that model
+   * parameters are updated less frequently, but with lower variance.
+   */
+  batch_size?: 'auto' | number;
+
+  /**
+   * The beta value for the DPO method. A higher beta value will increase the weight
+   * of the penalty between the policy and reference model.
+   */
+  beta?: 'auto' | number;
+
+  /**
+   * Scaling factor for the learning rate. A smaller learning rate may be useful to
+   * avoid overfitting.
+   */
+  learning_rate_multiplier?: 'auto' | number;
+
+  /**
+   * The number of epochs to train the model for. An epoch refers to one full cycle
+   * through the training dataset.
+   */
+  n_epochs?: 'auto' | number;
+}
+
+/**
+ * Configuration for the DPO fine-tuning method.
+ */
+export interface DpoMethod {
+  /**
+   * The hyperparameters used for the DPO fine-tuning job.
+   */
+  hyperparameters?: DpoHyperparameters;
+}
+
+/**
+ * The hyperparameters used for the reinforcement fine-tuning job.
+ */
+export interface ReinforcementHyperparameters {
+  /**
+   * Number of examples in each batch. A larger batch size means that model
+   * parameters are updated less frequently, but with lower variance.
+   */
+  batch_size?: 'auto' | number;
+
+  /**
+   * Multiplier on amount of compute used for exploring search space during training.
+   */
+  compute_multiplier?: 'auto' | number;
+
+  /**
+   * The number of training steps between evaluation runs.
+   */
+  eval_interval?: 'auto' | number;
+
+  /**
+   * Number of evaluation samples to generate per training step.
+   */
+  eval_samples?: 'auto' | number;
+
+  /**
+   * Scaling factor for the learning rate. A smaller learning rate may be useful to
+   * avoid overfitting.
+   */
+  learning_rate_multiplier?: 'auto' | number;
+
+  /**
+   * The number of epochs to train the model for. An epoch refers to one full cycle
+   * through the training dataset.
+   */
+  n_epochs?: 'auto' | number;
+
+  /**
+   * Level of reasoning effort.
+   */
+  reasoning_effort?: 'default' | 'low' | 'medium' | 'high';
+}
+
+/**
+ * Configuration for the reinforcement fine-tuning method.
+ */
+export interface ReinforcementMethod {
+  /**
+   * The grader used for the fine-tuning job.
+   */
+  grader:
+    | GraderModelsAPI.StringCheckGrader
+    | GraderModelsAPI.TextSimilarityGrader
+    | GraderModelsAPI.PythonGrader
+    | GraderModelsAPI.ScoreModelGrader
+    | GraderModelsAPI.MultiGrader;
+
+  /**
+   * The hyperparameters used for the reinforcement fine-tuning job.
+   */
+  hyperparameters?: ReinforcementHyperparameters;
+}
+
+/**
+ * The hyperparameters used for the fine-tuning job.
+ */
+export interface SupervisedHyperparameters {
+  /**
+   * Number of examples in each batch. A larger batch size means that model
+   * parameters are updated less frequently, but with lower variance.
+   */
+  batch_size?: 'auto' | number;
+
+  /**
+   * Scaling factor for the learning rate. A smaller learning rate may be useful to
+   * avoid overfitting.
+   */
+  learning_rate_multiplier?: 'auto' | number;
+
+  /**
+   * The number of epochs to train the model for. An epoch refers to one full cycle
+   * through the training dataset.
+   */
+  n_epochs?: 'auto' | number;
+}
+
+/**
+ * Configuration for the supervised fine-tuning method.
+ */
+export interface SupervisedMethod {
+  /**
+   * The hyperparameters used for the fine-tuning job.
+   */
+  hyperparameters?: SupervisedHyperparameters;
+}
+
+export declare namespace Methods {
+  export {
+    type DpoHyperparameters as DpoHyperparameters,
+    type DpoMethod as DpoMethod,
+    type ReinforcementHyperparameters as ReinforcementHyperparameters,
+    type ReinforcementMethod as ReinforcementMethod,
+    type SupervisedHyperparameters as SupervisedHyperparameters,
+    type SupervisedMethod as SupervisedMethod,
+  };
+}