Skip to main content
Version: 1.0

aixplain.modules.finetune

__author__​

Copyright 2022 The aiXplain SDK authors

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Author: Duraikrishna Selvaraju, Thiago Castro Ferreira, Shreyas Sharma and Lucas Pavanelli Date: June 14th 2023 Description: FineTune Class

Finetune Objects​

class Finetune(Asset)

[view_source]

A tool for fine-tuning machine learning models using custom datasets.

This class provides functionality to customize pre-trained models for specific tasks by fine-tuning them on user-provided datasets. It handles the configuration of training parameters, data splitting, and job execution.

Attributes:

  • name Text - Name of the fine-tuning job.
  • dataset_list List[Dataset] - List of datasets to use for fine-tuning.
  • model Model - The base model to be fine-tuned.
  • cost FinetuneCost - Cost information for the fine-tuning job.
  • id Text - ID of the fine-tuning job.
  • description Text - Detailed description of the fine-tuning purpose.
  • supplier Text - Provider/creator of the fine-tuned model.
  • version Text - Version identifier of the fine-tuning job.
  • train_percentage float - Percentage of data to use for training.
  • dev_percentage float - Percentage of data to use for validation.
  • dataset_list0 Text - Template for formatting training examples, using <<COLUMN_NAME>> to reference dataset columns.
  • dataset_list1 Hyperparameters - Configuration for the fine-tuning process.
  • dataset_list2 dict - Extra metadata for the fine-tuning job.
  • dataset_list3 str - URL endpoint for the backend API.
  • dataset_list4 str - Authentication key for API access.
  • dataset_list5 str - aiXplain-specific API key.

__init__​

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

[view_source]

Initialize a new Finetune instance.

Arguments:

  • name Text - Name of the fine-tuning job.
  • dataset_list List[Dataset] - List of datasets to use for fine-tuning.
  • model Model - The base model to be fine-tuned.
  • cost FinetuneCost - Cost information for the fine-tuning job.
  • id Text, optional - ID of the job. Defaults to "".
  • description Text, optional - Detailed description of the fine-tuning purpose. Defaults to "".
  • supplier Text, optional - Provider/creator of the fine-tuned model. Defaults to "aiXplain".
  • version Text, optional - Version identifier. Defaults to "1.0".
  • train_percentage float, optional - Percentage of data to use for training. Defaults to 100.
  • dev_percentage float, optional - Percentage of data to use for validation. Defaults to 0.
  • dataset_list0 Text, optional - Template for formatting training examples. Use <<COLUMN_NAME>> to reference dataset columns. Defaults to None.
  • dataset_list1 Hyperparameters, optional - Configuration for the fine-tuning process. Defaults to None.
  • dataset_list2 - Extra metadata for the fine-tuning job.

start​

def start() -> Model

[view_source]

Start the fine-tuning job on the backend.

This method submits the fine-tuning configuration to the backend and initiates the training process. It handles the creation of the training payload, including dataset splits and hyperparameters.

Returns:

  • Model - The model object representing the fine-tuning job. Returns None if the job submission fails.

Raises:

  • Exception - If there are errors in the API request or response handling.