> ## Documentation Index > Fetch the complete documentation index at: https://docs.baseten.co/llms.txt > Use this file to discover all available pages before exploring further. # Create training job > Creates a training job with the specified configuration. ## OpenAPI ````yaml post /v1/training_projects/{training_project_id}/jobs openapi: 3.1.0 info: description: REST API for management of Baseten resources title: Baseten management API version: 1.0.0 servers: - url: https://api.baseten.co security: - ApiKeyAuth: [] paths: /v1/training_projects/{training_project_id}/jobs: parameters: - $ref: '#/components/parameters/training_project_id' post: summary: Create a training job. description: Creates a training job with the specified configuration. requestBody: content: application/json: schema: $ref: '#/components/schemas/CreateTrainingJobRequestV1' required: true responses: '200': description: '' content: application/json: schema: $ref: '#/components/schemas/CreateTrainingJobResponseV1' x-codeSamples: - lang: bash source: >- curl --request POST \ --url https://api.baseten.co/v1/training_projects/{training_project_id}/jobs \ --header "Authorization: Api-Key $BASETEN_API_KEY" \ --data '{ "training_job": { "image": { "base_image": "hello-world", "docker_auth": null }, "compute": { "node_count": 1, "cpu_count": 1, "memory": "2Gi", "accelerator": { "accelerator": "H100", "count": 2 } }, "runtime": { "start_commands": [ "python main.py" ], "environment_variables": { "API_KEY": "your_api_key_here", "PATH": "/usr/bin" }, "artifacts": null, "enable_cache": true, "cache_config": { "enable_legacy_hf_mount": true, "enabled": true, "mount_base_path": "/root/.cache", "require_cache_affinity": true }, "checkpointing_config": { "enabled": true, "checkpoint_path": "/mnt/ckpts", "volume_size_gib": 10 }, "load_checkpoint_config": null }, "name": "gpt-oss-job", "truss_user_env": null, "interactive_session": null, "weights": [ { "allow_patterns": null, "auth": null, "auth_secret_name": null, "ignore_patterns": null, "mount_location": "/app/models/base", "source": "hf://meta-llama/Llama-3-8B@main" } ], "enable_baseten_workdir": false, "priority": 0 } }' - lang: python source: >- import requests import os API_KEY = os.environ.get("BASETEN_API_KEY", "") url = "https://api.baseten.co/v1/training_projects/{training_project_id}/jobs" headers = {"Authorization": f"Api-Key {API_KEY}"} response = requests.request( "POST", url, headers=headers, json={'training_job': {'image': {'base_image': 'hello-world', 'docker_auth': None}, 'compute': {'node_count': 1, 'cpu_count': 1, 'memory': '2Gi', 'accelerator': {'accelerator': 'H100', 'count': 2}}, 'runtime': {'start_commands': ['python main.py'], 'environment_variables': {'API_KEY': 'your_api_key_here', 'PATH': '/usr/bin'}, 'artifacts': None, 'enable_cache': True, 'cache_config': {'enable_legacy_hf_mount': True, 'enabled': True, 'mount_base_path': '/root/.cache', 'require_cache_affinity': True}, 'checkpointing_config': {'enabled': True, 'checkpoint_path': '/mnt/ckpts', 'volume_size_gib': 10}, 'load_checkpoint_config': None}, 'name': 'gpt-oss-job', 'truss_user_env': None, 'interactive_session': None, 'weights': [{'allow_patterns': None, 'auth': None, 'auth_secret_name': None, 'ignore_patterns': None, 'mount_location': '/app/models/base', 'source': 'hf://meta-llama/Llama-3-8B@main'}], 'enable_baseten_workdir': False, 'priority': 0}} ) print(response.text) components: parameters: training_project_id: schema: type: string name: training_project_id in: path required: true schemas: CreateTrainingJobRequestV1: description: A request to create a training job. properties: training_job: $ref: '#/components/schemas/CreateTrainingJobV1' description: The training job to create. required: - training_job title: CreateTrainingJobRequestV1 type: object CreateTrainingJobResponseV1: description: A response to creating a training job. properties: training_job: $ref: '#/components/schemas/TrainingJobV1' description: The created training job. required: - training_job title: CreateTrainingJobResponseV1 type: object CreateTrainingJobV1: description: Configuration for a training job. properties: image: $ref: '#/components/schemas/CreateTrainingJobImageV1' examples: - base_image: hello-world docker_auth: null compute: $ref: '#/components/schemas/CreateTrainingJobComputeV1' examples: - accelerator: accelerator: H100 count: 2 cpu_count: 1 memory: 2Gi node_count: 1 runtime: $ref: '#/components/schemas/CreateTrainingJobRuntimeV1' description: Configuration for the runtime environment of the training job. examples: - artifacts: [] cache_config: null checkpointing_config: checkpoint_path: null enabled: false volume_size_gib: null enable_cache: null environment_variables: API_KEY: your_api_key_here PATH: /usr/bin load_checkpoint_config: null start_commands: - python main.py name: anyOf: - type: string - type: 'null' default: null description: Name of the training job. examples: - gpt-oss-job title: Name truss_user_env: anyOf: - $ref: '#/components/schemas/TrussUserEnv' - type: 'null' default: null description: Truss user environment information interactive_session: anyOf: - $ref: '#/components/schemas/InteractiveSessionConfigV1' - type: 'null' default: null description: Configuration for interactive debugging sessions. weights: description: >- MDN weight sources to mount in the training container. Weights are mirrored and cached for fast startup. examples: - - allow_patterns: null auth: null auth_secret_name: null ignore_patterns: null mount_location: /app/models/base source: hf://meta-llama/Llama-3-8B@main items: $ref: '#/components/schemas/CreateJobWeightConfigV1' title: Weights type: array enable_baseten_workdir: default: false description: >- When enabled, uses /b10/workspace as the working directory instead of the image WORKDIR. examples: - false - true title: Enable Baseten Workdir type: boolean priority: anyOf: - type: integer - type: 'null' default: null description: Queue priority. Higher values are dequeued first. Defaults to 0. examples: - 0 - 10 - 100 title: Priority required: - image title: CreateTrainingJobV1 type: object TrainingJobV1: properties: id: description: Unique identifier of the training job. title: Id type: string created_at: description: Time the job was created in ISO 8601 format. format: date-time title: Created At type: string current_status: description: Current status of the training job. title: Current Status type: string error_message: anyOf: - type: string - type: 'null' default: null description: Error message if the training job failed. title: Error Message instance_type: $ref: '#/components/schemas/InstanceTypeV1' description: Instance type of the training job. updated_at: description: Time the job was updated in ISO 8601 format. format: date-time title: Updated At type: string training_project_id: description: ID of the training project. title: Training Project Id type: string training_project: $ref: '#/components/schemas/TrainingProjectSummaryV1' description: Summary of the training project. name: anyOf: - type: string - type: 'null' default: null description: Name of the training job. examples: - gpt-oss-job title: Name checkpoint_sync_status: anyOf: - $ref: '#/components/schemas/CheckpointSyncStatus' - type: 'null' default: null description: Checkpoint sync status of the training job. priority: default: 0 description: >- Queue priority. Higher values are dequeued first. NULL is treated as 0. title: Priority type: integer user: anyOf: - $ref: '#/components/schemas/UserV1' - type: 'null' default: null description: The user who created the training job. required: - id - created_at - current_status - instance_type - updated_at - training_project_id - training_project title: TrainingJobV1 type: object CreateTrainingJobImageV1: description: Configuration to create a training job image. properties: base_image: description: Base image for the training job. examples: - hello-world title: Base Image type: string docker_auth: anyOf: - $ref: '#/components/schemas/DockerAuthV1' - type: 'null' default: null description: Docker authentication credentials required: - base_image title: CreateTrainingJobImageV1 type: object CreateTrainingJobComputeV1: description: Configuration to specify the compute for a training job. properties: node_count: default: 1 description: Number of nodes for the training job. examples: - 1 title: Node Count type: integer cpu_count: default: 1 description: Number of cpus for the training job. examples: - 1 title: Cpu Count type: integer memory: default: 2Gi description: Memory for the training job. examples: - 2Gi title: Memory type: string accelerator: anyOf: - $ref: '#/components/schemas/CreateTrainingJobAcceleratorV1' - type: 'null' default: null description: GPU specification for the training job examples: - accelerator: H100 count: 2 title: CreateTrainingJobComputeV1 type: object CreateTrainingJobRuntimeV1: description: Configuration to specify the runtime environment for a training job. properties: start_commands: description: Commands to execute when starting the runtime. examples: - - python main.py items: type: string title: Start Commands type: array environment_variables: additionalProperties: anyOf: - type: string - $ref: '#/components/schemas/SecretReferenceV1' description: Environment variables to set in the runtime. examples: - API_KEY: your_api_key_here PATH: /usr/bin title: Environment Variables type: object artifacts: description: Runtime artifacts for the training job. items: $ref: '#/components/schemas/CreateTrainingJobS3Artifact' title: Artifacts type: array enable_cache: anyOf: - type: boolean - type: 'null' default: null description: Deprecated. Use cache_config instead. examples: - true title: Enable Cache cache_config: anyOf: - $ref: '#/components/schemas/CreateTrainingJobCacheConfig' - type: 'null' default: null description: Configuration for the read-write cache. examples: - enable_legacy_hf_mount: true enabled: true mount_base_path: /root/.cache require_cache_affinity: true checkpointing_config: $ref: '#/components/schemas/CreateTrainingJobCheckpointingConfig' description: Configuration for checkpointing. examples: - checkpoint_path: /mnt/ckpts enabled: true volume_size_gib: null load_checkpoint_config: anyOf: - $ref: '#/components/schemas/LoadCheckpointConfig' - type: 'null' default: null description: Configuration for loading checkpoints title: CreateTrainingJobRuntimeV1 type: object TrussUserEnv: description: >- This data models is used to flexibly store info alongside oracle versions. There is a corresponding data model in the truss client. In contrast, here all fields are optional for backwards compatibility with old clients. properties: truss_client_version: anyOf: - type: string - type: 'null' default: null title: Truss Client Version python_version: anyOf: - type: string - type: 'null' default: null title: Python Version pydantic_version: anyOf: - type: string - type: 'null' default: null title: Pydantic Version mypy_version: anyOf: - type: string - type: 'null' default: null title: Mypy Version is_library_deployment: default: false title: Is Library Deployment type: boolean is_frontend_deployment: default: false title: Is Frontend Deployment type: boolean git_info: anyOf: - $ref: '#/components/schemas/GitInfo' - type: 'null' default: null title: TrussUserEnv type: object InteractiveSessionConfigV1: description: Configuration for interactive debugging sessions on training jobs. properties: trigger: $ref: '#/components/schemas/V1InteractiveSessionTrigger' default: on_demand description: >- When to create the interactive session. 'on_startup' creates on job start, 'on_failure' creates on job failure, 'on_demand' bypasses automatic session creation. timeout_minutes: default: 480 description: Number of minutes before the interactive session times out. examples: - 480 - 1440 - 10080 title: Timeout Minutes type: integer session_provider: $ref: '#/components/schemas/V1InteractiveSessionProvider' default: vs_code description: The IDE client for the interactive session. auth_provider: $ref: '#/components/schemas/V1InteractiveSessionAuthProvider' default: github description: The authentication provider for the interactive session. title: InteractiveSessionConfigV1 type: object CreateJobWeightConfigV1: description: >- Weight source configuration for MDN (Model Distribution Network). Enables training jobs to mount external model weights from HuggingFace, S3, GCS, or R2 via MDN's caching and CSI mounting infrastructure. Weights are mirrored once and deduplicated across training jobs. properties: source: description: 'Weight source URI. Supported formats: hf://, s3://, gs://, r2://' examples: - hf://meta-llama/Llama-3-8B@main - s3://my-bucket/models/llama - gs://my-bucket/models/llama - r2://account_id.bucket/models/llama title: Source type: string mount_location: description: Path where weights will be mounted in the container examples: - /app/models/base - /models/llama title: Mount Location type: string allow_patterns: anyOf: - items: type: string type: array - type: 'null' default: null description: File patterns to include (Unix-style shell patterns) examples: - - '*.safetensors' - config.json title: Allow Patterns ignore_patterns: anyOf: - items: type: string type: array - type: 'null' default: null description: File patterns to exclude (Unix-style shell patterns) examples: - - '*.bin' - '*.h5' title: Ignore Patterns auth_secret_name: anyOf: - type: string - type: 'null' default: null description: >- Name of the workspace secret for authentication (e.g., HuggingFace token) examples: - hf_token - aws_credentials title: Auth Secret Name auth: anyOf: - additionalProperties: true type: object - type: 'null' default: null description: Authentication configuration for the weight source. examples: - auth_method: CUSTOM_SECRET auth_secret_name: hf_token title: Auth required: - source - mount_location title: CreateJobWeightConfigV1 type: object InstanceTypeV1: description: An instance type. properties: id: description: Identifier string for the instance type title: Id type: string name: description: Display name of the instance type title: Name type: string memory_limit_mib: description: Memory limit of the instance type in Mebibytes title: Memory Limit Mib type: integer millicpu_limit: description: CPU limit of the instance type in millicpu title: Millicpu Limit type: integer gpu_count: description: Number of GPUs on the instance type title: Gpu Count type: integer gpu_type: anyOf: - type: string - type: 'null' description: Type of GPU on the instance type title: Gpu Type gpu_memory_limit_mib: anyOf: - type: integer - type: 'null' description: Memory limit of the GPU on the instance type in Mebibytes title: Gpu Memory Limit Mib required: - id - name - memory_limit_mib - millicpu_limit - gpu_count - gpu_type - gpu_memory_limit_mib title: InstanceTypeV1 type: object TrainingProjectSummaryV1: description: A summary of a training project. properties: id: description: Unique identifier of the training project. title: Id type: string name: description: Name of the training project. title: Name type: string required: - id - name title: TrainingProjectSummaryV1 type: object CheckpointSyncStatus: description: Lifecycle state for the checkpoint uploader. enum: - SYNCING - COMPLETED title: CheckpointSyncStatus type: string UserV1: description: A user. properties: email: anyOf: - type: string - type: 'null' default: null description: Email of the user. title: Email title: UserV1 type: object DockerAuthV1: description: Docker authentication credentials. properties: registry: description: Registry to authenticate with title: Registry type: string auth_method: $ref: '#/components/schemas/DockerAuthType' description: Method to authenticate with the registry examples: - GCP_SERVICE_ACCOUNT_JSON - AWS_IAM - AWS_OIDC - GCP_OIDC - REGISTRY_SECRET gcp_service_account_json_docker_auth: anyOf: - $ref: '#/components/schemas/GcpServiceAccountJsonDockerAuthV1' - type: 'null' default: null description: GCP service account details for the registry aws_iam_docker_auth: anyOf: - $ref: '#/components/schemas/AwsIamDockerAuthV1' - type: 'null' default: null description: AWS details for the registry aws_oidc_docker_auth: anyOf: - $ref: '#/components/schemas/AwsOidcDockerAuthV1' - type: 'null' default: null description: AWS OIDC details for the registry gcp_oidc_docker_auth: anyOf: - $ref: '#/components/schemas/GcpOidcDockerAuthV1' - type: 'null' default: null description: GCP OIDC details for the registry registry_secret_docker_auth: anyOf: - $ref: '#/components/schemas/RegistrySecretDockerAuthV1' - type: 'null' default: null description: >- Required when auth_method is REGISTRY_SECRET. Supports any Docker registry (Docker Hub, GHCR, NGC, etc.) via username:password credentials stored as a Baseten secret. required: - registry - auth_method title: DockerAuthV1 type: object CreateTrainingJobAcceleratorV1: properties: accelerator: description: GPU type for the training job. examples: - H100 title: Accelerator type: string count: description: GPUs needed for the training job. examples: - 2 title: Count type: integer required: - accelerator - count title: CreateTrainingJobAcceleratorV1 type: object SecretReferenceV1: properties: name: description: Name of the secret to reference. examples: - hf_token title: Name type: string required: - name title: SecretReferenceV1 type: object CreateTrainingJobS3Artifact: properties: s3_bucket: description: S3 bucket for the uploaded runtime artifact. examples: - my-s3-bucket title: S3 Bucket type: string s3_key: description: S3 key for the uploaded runtime artifact. examples: - my-s3-key title: S3 Key type: string required: - s3_bucket - s3_key title: CreateTrainingJobS3Artifact type: object CreateTrainingJobCacheConfig: properties: enable_legacy_hf_mount: default: false description: Whether to enable the legacy Hugging Face cache. examples: - true title: Enable Legacy Hf Mount type: boolean enabled: default: false description: Whether to enable the read-write cache. examples: - true title: Enabled type: boolean require_cache_affinity: default: true description: >- Whether to require region affinity for the read-write cache. If False, the resulting job is not guaranteed to be deployed alongside the previous cache. examples: - true - false title: Require Cache Affinity type: boolean mount_base_path: default: /root/.cache description: >- Mount base path for the cache directory. The project cache and team cache will be mounted under this path. examples: - /workspace/.cache - /root/.cache title: Mount Base Path type: string title: CreateTrainingJobCacheConfig type: object CreateTrainingJobCheckpointingConfig: properties: enabled: default: false description: Whether checkpointing is enabled. examples: - true title: Enabled type: boolean checkpoint_path: anyOf: - type: string - type: 'null' default: null description: path where checkpoints will be saved. examples: - /mnt/ckpts title: Checkpoint Path volume_size_gib: anyOf: - type: integer - type: 'null' default: null description: >- Size of the volume in gibibytes. If not provided, the default size will be used examples: - 10 title: Volume Size Gib title: CreateTrainingJobCheckpointingConfig type: object LoadCheckpointConfig: properties: enabled: default: false description: Whether checkpoint loading is enabled title: Enabled type: boolean download_folder: default: /tmp/loaded_checkpoints description: Folder where checkpoints will be downloaded title: Download Folder type: string checkpoints: description: List of checkpoint configurations items: discriminator: mapping: baseten_latest_checkpoint: $ref: '#/components/schemas/BasetenLatestCheckpointConfig' baseten_named_checkpoint: $ref: '#/components/schemas/BasetenNamedCheckpointConfig' propertyName: typ oneOf: - $ref: '#/components/schemas/BasetenLatestCheckpointConfig' - $ref: '#/components/schemas/BasetenNamedCheckpointConfig' title: Checkpoints type: array title: LoadCheckpointConfig type: object GitInfo: properties: latest_commit_sha: title: Latest Commit Sha type: string latest_tag: anyOf: - type: string - type: 'null' title: Latest Tag commits_since_tag: anyOf: - type: integer - type: 'null' title: Commits Since Tag has_uncommitted_changes: title: Has Uncommitted Changes type: boolean required: - latest_commit_sha - latest_tag - commits_since_tag - has_uncommitted_changes title: GitInfo type: object V1InteractiveSessionTrigger: enum: - on_startup - on_failure - on_demand title: V1InteractiveSessionTrigger type: string V1InteractiveSessionProvider: enum: - vs_code - cursor - ssh title: V1InteractiveSessionProvider type: string V1InteractiveSessionAuthProvider: enum: - github - microsoft title: V1InteractiveSessionAuthProvider type: string DockerAuthType: enum: - GCP_SERVICE_ACCOUNT_JSON - AWS_IAM - AWS_OIDC - GCP_OIDC - REGISTRY_SECRET title: DockerAuthType type: string GcpServiceAccountJsonDockerAuthV1: description: GCP details for the registry. properties: service_account_json_secret_ref: $ref: '#/components/schemas/SecretReferenceV1' description: Name of the service account secret required: - service_account_json_secret_ref title: GcpServiceAccountJsonDockerAuthV1 type: object AwsIamDockerAuthV1: description: AWS details for the registry. properties: access_key_secret_ref: $ref: '#/components/schemas/SecretReferenceV1' description: Name of the access key secret secret_access_key_secret_ref: $ref: '#/components/schemas/SecretReferenceV1' description: Name of the secret key secret required: - access_key_secret_ref - secret_access_key_secret_ref title: AwsIamDockerAuthV1 type: object AwsOidcDockerAuthV1: description: AWS OIDC details for the registry. properties: role_arn: description: AWS IAM role ARN for OIDC authentication title: Role Arn type: string region: description: AWS region for OIDC authentication title: Region type: string required: - role_arn - region title: AwsOidcDockerAuthV1 type: object GcpOidcDockerAuthV1: description: GCP OIDC details for the registry. properties: service_account: description: GCP service account name for OIDC authentication title: Service Account type: string workload_identity_provider: description: GCP workload identity provider for OIDC authentication title: Workload Identity Provider type: string required: - service_account - workload_identity_provider title: GcpOidcDockerAuthV1 type: object RegistrySecretDockerAuthV1: description: >- Authentication via a Baseten secret for any Docker registry (Docker Hub, GHCR, NGC, etc.). The referenced secret must contain credentials in the format 'username:password'. For Docker Hub, set registry to 'https://index.docker.io/v1/'. For GHCR, use 'ghcr.io'. properties: secret_ref: $ref: '#/components/schemas/SecretReferenceV1' description: >- Reference to a Baseten secret containing credentials in the format 'username:password' required: - secret_ref title: RegistrySecretDockerAuthV1 type: object BasetenLatestCheckpointConfig: properties: project_name: anyOf: - type: string - type: 'null' default: null description: Name of the project to load the checkpoint from title: Project Name job_id: anyOf: - type: string - type: 'null' default: null description: ID of the job to load the checkpoint from title: Job Id typ: const: baseten_latest_checkpoint default: baseten_latest_checkpoint title: Typ type: string title: BasetenLatestCheckpointConfig type: object BasetenNamedCheckpointConfig: properties: checkpoint_name: description: Name of the checkpoint to load from title: Checkpoint Name type: string project_name: anyOf: - type: string - type: 'null' default: null description: Name of the project to load the checkpoint from title: Project Name job_id: anyOf: - type: string - type: 'null' default: null description: ID of the job to load the checkpoint from title: Job Id typ: const: baseten_named_checkpoint default: baseten_named_checkpoint title: Typ type: string required: - checkpoint_name title: BasetenNamedCheckpointConfig type: object securitySchemes: ApiKeyAuth: type: apiKey in: header name: Authorization description: >- You must specify the scheme 'Api-Key' in the Authorization header. For example, `Authorization: Api-Key ` ````