Fine-tuning the GPT-2 model for generating descriptions of establishments based on name, classes and rating

This project demonstrates the fine-tuning and application of a GPT-2 based model for generating text specific to product names and categories. The project utilizes the powerful GPT-2 architecture to produce contextually relevant descriptions that align with provided product characteristics such as name, category, and rating.

Articles

• Russian

Dataset

The dataset used for fine-tuning the model is the "geo-reviews-dataset-2023" from Hugging Face, which includes a variety of product names, categories, and ratings. This dataset provides a rich set of inputs for training the model to generate meaningful and context-aware text descriptions.

• Dataset Source: geo-reviews-dataset-2023 on Hugging Face

Model

The base model used for fine-tuning is the "ai-forever/rugpt3small_based_on_gpt2" from Hugging Face. This model offers a robust starting point for further training and adaptation to specific text generation tasks.

• Model Source: ai-forever/rugpt3small_based_on_gpt2 on Hugging Face

Requirements

To install the necessary libraries for running the project, please use the provided requirements.txt file. This file contains all the dependencies required to execute the project successfully.

pip install -r requirements.txt

Usage

To run the project, follow these steps:

1. Clone the repository to your local machine.
2. Ensure that Python and Pip are installed.
3. Install the required packages using the requirements.txt file.
4. Run the script to fine-tune the model with the dataset provided.
5. Use the trained model to generate text based on new product information.

Contributing

Contributions to this project are welcome. Please feel free to fork the repository, make changes, and submit a pull request. We appreciate your inputs to improve the project.

Citation

@misc{pisarenko2024fine-tuning,
  author = {Pisarenko, Anton},
  title = {Fine-tuning the {GPT-2} model for generating descriptions of establishments based on type and rating},
  year = {2024},
  howpublished = {\url{https://github.com/AntonSHBK/fine_tune_gpt_model}},
}

Contacts

• Telegram

License

This project is licensed under the MIT License - see the LICENSE.md file for details.