Skip to main content

module aixplain.modules.finetune

Global Variables​

  • cost
  • hyperparameters
  • status

class Finetune​

FineTune is a powerful tool for fine-tuning machine learning models and using your own datasets for specific tasks.

Attributes:

  • name (Text): Name of the FineTune.
  • dataset_list (List[Dataset]): List of Datasets to be used for fine-tuning.
  • model (Model): Model to be fine-tuned.
  • cost (Cost): Cost of the FineTune.
  • id (Text): ID of the FineTune.
  • description (Text): Description of the FineTune.
  • supplier (Text): Supplier of the FineTune.
  • version (Text): Version of the FineTune.
  • train_percentage (float): Percentage of training samples.
  • dev_percentage (float): Percentage of development samples.
  • prompt_template (Text): Fine-tuning prompt_template.
  • hyperparameters (Hyperparameters): Hyperparameters for fine-tuning.
  • additional_info (dict): Additional information to be saved with the FineTune.
  • backend_url (str): URL of the backend.
  • api_key (str): The TEAM API key used for authentication.

method __init__​

__init__(
    name: str,
    dataset_list: List[Dataset],
    model: Model,
    cost: FinetuneCost,
    id: Optional[str] = '',
    description: Optional[str] = '',
    supplier: Optional[str] = 'aiXplain',
    version: Optional[str] = '1.0',
    train_percentage: Optional[float] = 100,
    dev_percentage: Optional[float] = 0,
    prompt_template: Optional[str] = None,
    hyperparameters: Optional[Hyperparameters] = None,
    **additional_info
) → None

Create a FineTune with the necessary information.

Args:

  • name (Text): Name of the FineTune.
  • dataset_list (List[Dataset]): List of Datasets to be used for fine-tuning.
  • model (Model): Model to be fine-tuned.
  • cost (Cost): Cost of the FineTune.
  • id (Text, optional): ID of the FineTune. Defaults to "".
  • description (Text, optional): Description of the FineTune. Defaults to "".
  • supplier (Text, optional): Supplier of the FineTune. Defaults to "aiXplain".
  • version (Text, optional): Version of the FineTune. Defaults to "1.0".
  • train_percentage (float, optional): Percentage of training samples. Defaults to 100.
  • dev_percentage (float, optional): Percentage of development samples. Defaults to 0.
  • prompt_template (Text, optional): Fine-tuning prompt_template. Should reference columns in the dataset using format <<COLUMN_NAME>>. Defaults to None.
  • hyperparameters (Hyperparameters, optional): Hyperparameters for fine-tuning. Defaults to None.
  • **additional_info: Additional information to be saved with the FineTune.

method start​

start() → Model

Start the Finetune job.

Returns:

  • Model: The model object representing the Finetune job.