POST /v1/finetuning/finetuned-models
Creates a new fine-tuned model. The model will be trained on the dataset specified in the request body. The training process may take some time, and the model will be available once the training is complete.
Servers
- https://api.cohere.com
Request headers
| Name | Type | Required | Description |
|---|---|---|---|
Content-Type |
String | Yes |
The media type of the request body.
Default value: "application/json" |
X-Client-Name |
String | No |
The name of the project that is making the request. |
Request body fields
| Name | Type | Required | Description |
|---|---|---|---|
id |
String | No |
read-only. FinetunedModel ID. |
name |
String | Yes |
FinetunedModel name (e.g. |
creator_id |
String | No |
read-only. User ID of the creator. |
settings |
Object | Yes |
FinetunedModel settings such as dataset, hyperparameters... |
settings.multi_label |
Boolean | No |
read-only. Whether the model is single-label or multi-label (only for classification). |
settings.dataset_id |
String | Yes |
The data used for training and evaluating the fine-tuned model. |
settings.hyperparameters |
Object | No |
Fine-tuning hyper-parameters. |
settings.hyperparameters.train_epochs |
Integer | No |
The number of epochs to train for. |
settings.hyperparameters.lora_rank |
Integer | No |
Specifies the rank for low-rank matrices. Lower ranks reduce parameters but may limit model flexibility. |
settings.hyperparameters.lora_alpha |
Integer | No |
Controls the scaling factor for LoRA updates. Higher values make the updates more impactful. |
settings.hyperparameters.early_stopping_threshold |
Number | No |
How much the loss must improve to prevent early stopping. |
settings.hyperparameters.learning_rate |
Number | No |
The learning rate to be used during training. |
settings.hyperparameters.lora_target_modules |
String | No |
The combination of LoRA modules to target. Valid values:
Default value: "LORA_TARGET_MODULES_UNSPECIFIED" |
settings.hyperparameters.early_stopping_patience |
Integer | No |
Stops training if the loss metric does not improve beyond the value of
|
settings.hyperparameters.train_batch_size |
Integer | No |
The batch size is the number of training examples included in a single training pass. |
settings.base_model |
Object | Yes |
The base model to fine-tune. |
settings.base_model.name |
String | No |
The name of the base model. |
settings.base_model.base_type |
String | Yes |
The type of the base model. Valid values:
Default value: "BASE_TYPE_UNSPECIFIED" |
settings.base_model.strategy |
String | No |
Deprecated: The fine-tuning strategy. Valid values:
Default value: "STRATEGY_UNSPECIFIED" |
settings.base_model.version |
String | No |
read-only. The version of the base model. |
settings.wandb |
Object | No |
The Weights & Biases configuration (Chat fine-tuning only). |
settings.wandb.api_key |
String | Yes |
The WandB API key to be used during training. |
settings.wandb.entity |
String | No |
The WandB entity name to be used during training. |
settings.wandb.project |
String | Yes |
The WandB project name to be used during training. |
status |
String | No |
read-only. Current stage in the life-cycle of the fine-tuned model. Valid values:
Default value: "STATUS_UNSPECIFIED" |
completed_at |
String | No |
read-only. Timestamp for the completed fine-tuning. |
last_used |
String | No |
read-only. Deprecated: Timestamp for the latest request to this fine-tuned model. |
organization_id |
String | No |
read-only. Organization ID. |
created_at |
String | No |
read-only. Creation timestamp. |
updated_at |
String | No |
read-only. Latest update timestamp. |
How to start integrating
- Add HTTP Task to your workflow definition.
- Search for the API you want to integrate with and click on the name.
- This loads the API reference documentation and prepares the Http request settings.
- Click Test request to test run your request to the API and see the API's response.