Create a fine-tuning job

Creates a fine-tuning job which begins the process of creating a new model from a given dataset.

Response includes details of the enqueued job including job status and the name of the fine-tuned models once complete.

Learn more about fine-tuning

POST /fine_tuning/jobs

Try it

Create a fine-tuning job

cURL

curl --request POST \
  --url https://api.openai.com/v1/fine_tuning/jobs \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model": "gpt-4o-mini",
  "training_file": "file-abc123",
  "hyperparameters": {
    "batch_size": "auto",
    "learning_rate_multiplier": "auto",
    "n_epochs": "auto"
  },
  "suffix": null,
  "validation_file": "file-abc123",
  "integrations": [
    {
      "type": "wandb",
      "wandb": {
        "project": "my-wandb-project",
        "name": "<string>",
        "entity": "<string>",
        "tags": [
          "custom-tag"
        ]
      }
    }
  ],
  "seed": 42,
  "method": {
    "type": "supervised",
    "supervised": {
      "hyperparameters": {
        "batch_size": "auto",
        "learning_rate_multiplier": "auto",
        "n_epochs": "auto"
      }
    },
    "dpo": {
      "hyperparameters": {
        "beta": "auto",
        "batch_size": "auto",
        "learning_rate_multiplier": "auto",
        "n_epochs": "auto"
      }
    }
  },
  "metadata": {}
}
'

200

{
  "id": "<string>",
  "created_at": 123,
  "error": {
    "code": "<string>",
    "message": "<string>",
    "param": "<string>"
  },
  "fine_tuned_model": "<string>",
  "finished_at": 123,
  "hyperparameters": {
    "batch_size": "auto",
    "learning_rate_multiplier": "auto",
    "n_epochs": "auto"
  },
  "model": "<string>",
  "object": "fine_tuning.job",
  "organization_id": "<string>",
  "result_files": [
    "file-abc123"
  ],
  "status": "validating_files",
  "trained_tokens": 123,
  "training_file": "<string>",
  "validation_file": "<string>",
  "seed": 123,
  "integrations": [
    {
      "type": "wandb",
      "wandb": {
        "project": "my-wandb-project",
        "name": "<string>",
        "entity": "<string>",
        "tags": [
          "custom-tag"
        ]
      }
    }
  ],
  "estimated_finish": 123,
  "method": {
    "type": "supervised",
    "supervised": {
      "hyperparameters": {
        "batch_size": "auto",
        "learning_rate_multiplier": "auto",
        "n_epochs": "auto"
      }
    },
    "dpo": {
      "hyperparameters": {
        "beta": "auto",
        "batch_size": "auto",
        "learning_rate_multiplier": "auto",
        "n_epochs": "auto"
      }
    }
  },
  "metadata": {}
}

Authorizations

Authorization · string · header · required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json

model · required

The name of the model to fine-tune. You can select one of the supported models.

Example:

"gpt-4o-mini"

training_file · string · required

The ID of an uploaded file that contains training data.

See upload file for how to upload a file.

Your dataset must be formatted as a JSONL file. Additionally, you must upload your file with the purpose fine-tune.

The contents of the file should differ depending on if the model uses the chat, completions format, or if the fine-tuning method uses the preference format.

See the fine-tuning guide for more details.

Example:

"file-abc123"

hyperparameters · object · deprecated

The hyperparameters used for the fine-tuning job. This value is now deprecated in favor of method, and should be passed in under the method parameter.

Show child attributes

hyperparameters. batch_size · default: auto

Number of examples in each batch. A larger batch size means that model parameters are updated less frequently, but with lower variance.

Available options: auto

hyperparameters. learning_rate_multiplier · default: auto

Scaling factor for the learning rate. A smaller learning rate may be useful to avoid overfitting.

Available options: auto

hyperparameters. n_epochs · default: auto

The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset.

Available options: auto

suffix · string | null

A string of up to 64 characters that will be added to your fine-tuned model name.

For example, a suffix of "custom-model-name" would produce a model name like ft:gpt-4o-mini:openai:custom-model-name:7p4lURel.

Required string length: 1 - 64

validation_file · string | null

The ID of an uploaded file that contains validation data.

If you provide this file, the data is used to generate validation metrics periodically during fine-tuning. These metrics can be viewed in the fine-tuning results file. The same data should not be present in both train and validation files.

Your dataset must be formatted as a JSONL file. You must upload your file with the purpose fine-tune.

See the fine-tuning guide for more details.

Example:

"file-abc123"

integrations · object[] | null

A list of integrations to enable for your fine-tuning job.

Show child attributes

integrations. type · enum<string> · required

The type of integration to enable. Currently, only "wandb" (Weights and Biases) is supported.

Available options: wandb

integrations. wandb · object · required

The settings for your integration with Weights and Biases. This payload specifies the project that metrics will be sent to. Optionally, you can set an explicit display name for your run, add tags to your run, and set a default entity (team, username, etc) to be associated with your run.

Show child attributes

integrations.wandb. project · string · required

The name of the project that the new run will be created under.

Example:

"my-wandb-project"

integrations.wandb. name · string | null

A display name to set for the run. If not set, we will use the Job ID as the name.

integrations.wandb. entity · string | null

The entity to use for the run. This allows you to set the team or username of the WandB user that you would like associated with the run. If not set, the default entity for the registered WandB API key is used.

integrations.wandb. tags · string[]

A list of tags to be attached to the newly created run. These tags are passed through directly to WandB. Some default tags are generated by OpenAI: "openai/finetune", "openai/{base-model}", "openai/{ftjob-abcdef}".

seed · integer | null

The seed controls the reproducibility of the job. Passing in the same seed and job parameters should produce the same results, but may differ in rare cases. If a seed is not specified, one will be generated for you.

Required range: 0 <= x <= 2147483647

Example:

42

method · object

The method used for fine-tuning.

Show child attributes

method. type · enum<string>

The type of method. Is either supervised or dpo.

Available options: supervised, dpo

method. supervised · object

Configuration for the supervised fine-tuning method.

Show child attributes

method.supervised. hyperparameters · object

The hyperparameters used for the fine-tuning job.

Show child attributes

method.supervised.hyperparameters. batch_size · default: auto

Number of examples in each batch. A larger batch size means that model parameters are updated less frequently, but with lower variance.

Available options: auto

method.supervised.hyperparameters. learning_rate_multiplier · default: auto

Scaling factor for the learning rate. A smaller learning rate may be useful to avoid overfitting.

Available options: auto

method.supervised.hyperparameters. n_epochs · default: auto

The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset.

Available options: auto

method. dpo · object

Configuration for the DPO fine-tuning method.

Show child attributes

method.dpo. hyperparameters · object

The hyperparameters used for the fine-tuning job.

Show child attributes

method.dpo.hyperparameters. beta · default: auto

The beta value for the DPO method. A higher beta value will increase the weight of the penalty between the policy and reference model.

Available options: auto

method.dpo.hyperparameters. batch_size · default: auto

Number of examples in each batch. A larger batch size means that model parameters are updated less frequently, but with lower variance.

Available options: auto

method.dpo.hyperparameters. learning_rate_multiplier · default: auto

Scaling factor for the learning rate. A smaller learning rate may be useful to avoid overfitting.

Available options: auto

method.dpo.hyperparameters. n_epochs · default: auto

The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset.

Available options: auto

metadata · object

Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format, and querying for objects via API or the dashboard.

Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters.

Show child attributes

metadata. {key} · string

Response

200 - application/json

OK

The fine_tuning.job object represents a fine-tuning job that has been created through the API.

id · string · required

The object identifier, which can be referenced in the API endpoints.

created_at · integer · required

The Unix timestamp (in seconds) for when the fine-tuning job was created.

error · object · required

For fine-tuning jobs that have failed, this will contain more information on the cause of the failure.

Show child attributes

error. code · string · required

A machine-readable error code.

error. message · string · required

A human-readable error message.

error. param · string | null · required

The parameter that was invalid, usually training_file or validation_file. This field will be null if the failure was not parameter-specific.

fine_tuned_model · string | null · required

The name of the fine-tuned model that is being created. The value will be null if the fine-tuning job is still running.

finished_at · integer | null · required

The Unix timestamp (in seconds) for when the fine-tuning job was finished. The value will be null if the fine-tuning job is still running.

hyperparameters · object · required

The hyperparameters used for the fine-tuning job. This value will only be returned when running supervised jobs.

Show child attributes

hyperparameters. batch_size · default: auto

Number of examples in each batch. A larger batch size means that model parameters are updated less frequently, but with lower variance.

Available options: auto

hyperparameters. learning_rate_multiplier · default: auto

Scaling factor for the learning rate. A smaller learning rate may be useful to avoid overfitting.

Available options: auto

hyperparameters. n_epochs · default: auto

The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset.

Available options: auto

model · string · required

The base model that is being fine-tuned.

object · enum<string> · required

The object type, which is always "fine_tuning.job".

Available options: fine_tuning.job

organization_id · string · required

The organization that owns the fine-tuning job.

result_files · string[] · required

The compiled results file ID(s) for the fine-tuning job. You can retrieve the results with the Files API.

status · enum<string> · required

The current status of the fine-tuning job, which can be either validating_files, queued, running, succeeded, failed, or cancelled.

Available options: validating_files, queued, running, succeeded, failed, cancelled

trained_tokens · integer | null · required

The total number of billable tokens processed by this fine-tuning job. The value will be null if the fine-tuning job is still running.

training_file · string · required

The file ID used for training. You can retrieve the training data with the Files API.

validation_file · string | null · required

The file ID used for validation. You can retrieve the validation results with the Files API.

seed · integer · required

The seed used for the fine-tuning job.

integrations · Fine-Tuning Job Integration · object[] | null

A list of integrations to enable for this fine-tuning job.

Maximum array length: 5

Show child attributes

integrations. type · enum<string> · required

The type of the integration being enabled for the fine-tuning job

Available options: wandb

integrations. wandb · object · required

The settings for your integration with Weights and Biases. This payload specifies the project that metrics will be sent to. Optionally, you can set an explicit display name for your run, add tags to your run, and set a default entity (team, username, etc) to be associated with your run.

Show child attributes

integrations.wandb. project · string · required

The name of the project that the new run will be created under.

Example:

"my-wandb-project"

integrations.wandb. name · string | null

A display name to set for the run. If not set, we will use the Job ID as the name.

integrations.wandb. entity · string | null

The entity to use for the run. This allows you to set the team or username of the WandB user that you would like associated with the run. If not set, the default entity for the registered WandB API key is used.

integrations.wandb. tags · string[]

A list of tags to be attached to the newly created run. These tags are passed through directly to WandB. Some default tags are generated by OpenAI: "openai/finetune", "openai/{base-model}", "openai/{ftjob-abcdef}".

estimated_finish · integer | null

The Unix timestamp (in seconds) for when the fine-tuning job is estimated to finish. The value will be null if the fine-tuning job is not running.

method · object

The method used for fine-tuning.

Show child attributes

method. type · enum<string>

The type of method. Is either supervised or dpo.

Available options: supervised, dpo

method. supervised · object

Configuration for the supervised fine-tuning method.

Show child attributes

method.supervised. hyperparameters · object

The hyperparameters used for the fine-tuning job.

Show child attributes

method.supervised.hyperparameters. batch_size · default: auto

Number of examples in each batch. A larger batch size means that model parameters are updated less frequently, but with lower variance.

Available options: auto

method.supervised.hyperparameters. learning_rate_multiplier · default: auto

Scaling factor for the learning rate. A smaller learning rate may be useful to avoid overfitting.

Available options: auto

method.supervised.hyperparameters. n_epochs · default: auto

The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset.

Available options: auto

method. dpo · object

Configuration for the DPO fine-tuning method.

Show child attributes

method.dpo. hyperparameters · object

The hyperparameters used for the fine-tuning job.

Show child attributes

method.dpo.hyperparameters. beta · default: auto

The beta value for the DPO method. A higher beta value will increase the weight of the penalty between the policy and reference model.

Available options: auto

method.dpo.hyperparameters. batch_size · default: auto

Number of examples in each batch. A larger batch size means that model parameters are updated less frequently, but with lower variance.

Available options: auto

method.dpo.hyperparameters. learning_rate_multiplier · default: auto

Scaling factor for the learning rate. A smaller learning rate may be useful to avoid overfitting.

Available options: auto

method.dpo.hyperparameters. n_epochs · default: auto

The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset.

Available options: auto

metadata · object

Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format, and querying for objects via API or the dashboard.

Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters.

Show child attributes

metadata. {key} · string