Gothic AI - Chatbot for Novel Character

This project aims to build a chatbot that emulates a character from a novel using a fine-tuned GPT-2 model. The chatbot will generate responses based on the character's personality as depicted in the novel.

Table of Contents

• Installation
• Setup
• Training the Model
• Running Inference
• Directory Structure
• Dependencies

Installation

Step 1: Clone the Repository

Clone the repository to your local machine:

git clone https://github.com/Gothic-AI/gothic_model.git
cd gothic_model

Step 2: Set up a virtual environment

Create a virtual environment

python3 -m venv venv
source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

Step 3: Install the required dependencies

pip install -r requirements.txt

Model Setup Instructions

Prepare Your Training Data

Before running the model, you need to prepare the environment by providing the necessary data files. Your training data should be organized as follows:

1. Plain Text Version of the Novel:
   Place the plain text version of the novel in the following path:

   data/test_plain.txt
   

2. Q&A Version of the Novel:
   Place the Q&A version of the novel in the following path:

   data/test_q&a.txt
   

Configuration

You need to modify the configuration file based on your experiment setup.

1. If you are using the Plain Text version of the novel, modify the config/config_plain.yaml file.

2. If you are using the Q&A version of the novel, modify the config/config_q&a.yaml file.

In the configuration file, you can adjust various preferences such as training parameters, learning rates, and other model settings according to your specific needs.

---

Once your data and configuration are ready, you can proceed with running the model!

Training the Model

To train the GPT-2 model on your dataset, follow these steps:

Prepare the data:

Make sure your data is in the data/ folder and is in .txt format (e.g., test_plain.txt, test_q&a.txt).

Run the training script: To train the model on plain text data, run:

python scripts/train_model.py --config config/config_plain.yaml

To train the model on Q&A data, run:

python scripts/train_model.py --config config/config_q&a.yaml

This command will start the fine-tuning process based on the configuration specified in the selected YAML file.

The training process may take a while depending on your hardware and the size of your data.

Model Output:

Once the training completes, the model will be saved in the models/ directory (as specified in the respective config/*.yaml file).

Generating Text:

Once the model is trained, you can use it to generate text (simulating responses from the Gothic chatbot).

Run the text generation script: To generate text, use the following command:

python inference/generate_text.py

You can modify the generate_text.py script to provide different prompts for your chatbot, or change the model and output sequence length as required.

Modify the prompt: In inference/generate_text.py, you can modify the sequence1 variable to set different questions or inputs for the chatbot. Folder Structure

Here's a brief overview of the folder structure:

gothic_model/
├── config/
│   ├── config_plain.yaml        # Configuration file for training with plain text data
│   └── config_q&a.yaml          # Configuration file for training with Q&A data
├── data/
│   ├── test_plain.txt           # Your training data (plain text version of the novel)
│   └── test_q&a.txt             # Your training data (Q&A version of the novel)
├── inference/
│   ├── generate_text.py         # Script to generate text based on the trained model
│   └── utils.py                 # Helper functions for loading model and tokenizer
├── models/                      # Directory where the trained models are saved
├── scripts/
│   ├── data.py                  # Data handling and processing
│   └── train_model.py           # Main script for training the model
├── readme.md                    # This readme file
├── requirements.txt             # List of dependencies
└── .gitignore                   # Ignore unnecessary files in Git

Troubleshooting

If you encounter any issues, here are a few common troubleshooting tips:

1. File Not Found Errors: Make sure all files referenced in the config/*.yaml files exist in the proper directories. Verify the paths and filenames.
2. Memory Issues: If your model is too large or your system runs out of memory, try reducing the batch size or number of epochs in the configuration file.
3. Tokenization Issues: If you encounter issues with tokenization, ensure your data is properly formatted and free of any non-text elements that could interfere with the tokenizer.
4. CUDA/GPUs: If you're using a GPU and facing issues related to CUDA, make sure that your system has the correct CUDA version installed. Ensure the proper GPU drivers are configured.