> ## Documentation Index > Fetch the complete documentation index at: https://docs.together.ai/llms.txt > Use this file to discover all available pages before exploring further. # Optimizer step > Submits an optimizer step operation that will asynchronously apply accumulated gradients to update model parameters. Does not make the updated parameters available for sampling; call `weights-sync` afterwards when you want subsequent samples to use the updated policy. ## OpenAPI ````yaml openapi.yaml POST /rl/training-sessions/{session_id}/operations/optim-step openapi: 3.1.0 info: title: Together APIs description: The Together REST API. See https://docs.together.ai for more details. version: 2.0.0 termsOfService: https://www.together.ai/terms-of-service contact: name: Together Support url: https://www.together.ai/contact license: name: MIT url: https://github.com/togethercomputer/openapi/blob/main/LICENSE servers: - url: https://api.together.ai/v1 description: Default environment for APIs - url: https://api-inference.together.ai/v2 description: Optimized environment for inference security: - bearerAuth: [] paths: /rl/training-sessions/{session_id}/operations/optim-step: post: tags: - RL summary: Optimizer step description: >- Submits an optimizer step operation that will asynchronously apply accumulated gradients to update model parameters. Does not make the updated parameters available for sampling; call `weights-sync` afterwards when you want subsequent samples to use the updated policy. operationId: OptimStep parameters: - name: session_id in: path required: true schema: description: Training session ID type: string - name: Idempotency-Key in: header required: true schema: description: >- Required key that makes retries return the original operation; use a new key for changed request bodies. type: string requestBody: content: application/json: schema: $ref: '#/components/schemas/RL.OptimStepBody' required: true responses: '200': description: Optimizer step operation details content: application/json: schema: $ref: '#/components/schemas/RL.OptimStepOperation' default: description: An unexpected error response. content: application/json: schema: $ref: '#/components/schemas/ErrorData' components: schemas: RL.OptimStepBody: type: object description: Request body for an optimizer step. properties: adam_params: $ref: '#/components/schemas/RL.AdamParams' description: Adam optimizer overrides for this step. muon_params: $ref: '#/components/schemas/RL.MuonParams' description: Muon optimizer overrides for this step. RL.OptimStepOperation: type: object description: Async optimizer step operation required: - id - status properties: id: type: string example: 550e8400-e29b-41d4-a716-446655440000 description: Operation ID status: $ref: '#/components/schemas/RL.TrainingOperationStatus' description: Operation status output: $ref: '#/components/schemas/RL.OptimStepResult' description: Result on success error: $ref: '#/components/schemas/RL.TrainingOperationError' description: Error details on failure ErrorData: type: object required: - error properties: error: type: object properties: message: type: string nullable: false type: type: string nullable: false param: type: string nullable: true default: null code: type: string nullable: true default: null required: - type - message RL.AdamParams: description: Per-step Adam optimizer overrides. type: object properties: learning_rate: description: Learning rate for the Adam-tuned parameters type: number example: 0.0001 beta1: description: Exponential decay rate for the first-moment estimate type: number example: 0.9 default: 0.9 beta2: description: Exponential decay rate for the second-moment estimate type: number example: 0.95 default: 0.95 eps: description: Epsilon for numerical stability type: number example: 1.e-8 default: 1.e-8 weight_decay: description: Weight decay coefficient type: number example: 0.1 default: 0 grad_clip_norm: description: >- Maximum gradient norm for this step, gradients across all model parameters are clipped to this value. Set to 0 to disable gradient clipping. When unset, gradients are clipped to the session default (1.0). type: number example: 10 default: 1 RL.MuonParams: type: object description: Per-step Muon optimizer overrides properties: learning_rate: type: number example: 0.02 description: Learning rate for this Muon optimizer step. momentum: type: number example: 0.95 default: 0.95 description: Momentum coefficient newton_schulz_steps: type: integer example: 5 default: 5 description: Number of Newton-Schulz iterations weight_decay: type: number example: 0 default: 0 description: Weight decay coefficient grad_clip_norm: description: >- Maximum gradient norm for this step, gradients across all model parameters are clipped to this value. Set to 0 to disable gradient clipping. When unset, gradients are clipped to the session default (1.0). type: number example: 10 default: 1 adam: $ref: '#/components/schemas/RL.AdamParams' description: >- Per-step Adam optimizer overrides for the Adam-tuned parameters in a Muon-tuned optimizer session. RL.TrainingOperationStatus: type: string enum: - TRAINING_OPERATION_STATUS_UNSPECIFIED - TRAINING_OPERATION_STATUS_PENDING - TRAINING_OPERATION_STATUS_RUNNING - TRAINING_OPERATION_STATUS_COMPLETED - TRAINING_OPERATION_STATUS_FAILED default: TRAINING_OPERATION_STATUS_UNSPECIFIED RL.OptimStepResult: type: object description: Result of an optimizer step operation required: - step properties: step: description: Step number oneOf: - type: string - type: integer example: '100' RL.TrainingOperationError: type: object description: Error details for a failed training operation properties: code: $ref: '#/components/schemas/RL.TrainingOperationErrorCode' description: Application error code example: TRAINING_OPERATION_ERROR_CODE_TIMEOUT message: description: Human-readable error message type: string example: Operation timed out RL.TrainingOperationErrorCode: type: string description: Application error code for a failed training operation enum: - TRAINING_OPERATION_ERROR_CODE_UNSPECIFIED - TRAINING_OPERATION_ERROR_CODE_RESOURCE_EXHAUSTED - TRAINING_OPERATION_ERROR_CODE_TIMEOUT - TRAINING_OPERATION_ERROR_CODE_INTERNAL_ERROR - TRAINING_OPERATION_ERROR_CODE_SESSION_NOT_ACTIVE - TRAINING_OPERATION_ERROR_CODE_INVALID_INPUT - TRAINING_OPERATION_ERROR_CODE_NON_FINITE_LOSS default: TRAINING_OPERATION_ERROR_CODE_UNSPECIFIED securitySchemes: bearerAuth: type: http scheme: bearer x-bearer-format: bearer x-default: default ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.