Donut LLM Tools

Donut LLM Tools is a suite of two programs that helps you create dataset from wikipedia data and your own LLM with the created or already existing dataset.

Latest Version 24.05.29

How to install, import and use DonutLLM?

Installation

• Before installing Donut LLM Tools, make sure you have Python 3.9 or later till Python 3.11. Note : Python 3.12 is not supported by PyTorch, hence Donut LLM Tools will also have trouble running in it.

• The dependencies for installing Donut LLM Tools are:

  1. DonutLLMCore (pip install DonutLLMCore)
  2. torch (Required by DonutLLMCore)
  3. torchvision (Required by CreateLLM, PyTorch)
  4. torchaudio (Required by CreateLLM, PyTorch)
  5. Wikipedia

  Note : You need to check PyTorch website to find more about installation on devices with only CPU or with GPU. Donut LLM Tools support CPU, however it is very slow to train a model in CPU.

• After ensuring the installation of the above mentioned dependencies, do pip install donut-llm-tools.

• You have now installed Donut LLM Tools in your device.

Importing and Using

from donutllmtools import Tools

Tools.DatasetCreator() # For creating wikipedia based datasets.

Tools.LLMCreator() # For creating/loading AI model from the above created dataset or a custom dataset.

The above code snippet is used to create a dataset and also train/load an AI model

from donutllmtools import HelpAndInfo

HelpAndInfo.help() # To view help.

HelpAndInfo.about() # To view details of the program.

HelpAndInfo.create_llm() # Also create LLM from HelpAndInfo class.

HelpAndInfo.create_dataset() # ALso create dataset from HelpAndInfo class.

HelpAndInfo.main() # To get a menu based interface for dataset creation or model load/creation.

The above code uses HelpAndInfo class which has the same functions from Tools class as well as help functions.