oracle-n 0.1.0

A custom model for sentiment analysis

Oracle-n Model

Oracle-n is an AI model based on the BERT architecture, designed for text and sentiment analysis. Using the foundational strengths of BERT, Oracle-n is designed to to optimize performance for specific needs. This repository includes the model code, tokenizer, and training scripts.

Features

• Customized BERT Configuration: Tailored configurations to fit text and sentiment analysis tasks.
• Oracle-n Tokenizer: Custom tokenizer designed for preprocessing text data efficiently.
• Sentiment Analysis: Trained on the IMDb dataset to perform sentiment analysis tasks.

Directory Structure

• aclImdb/: Directory containing the IMDb dataset files in Parquet format.
• dataset.py: Script for handling the dataset loading and preprocessing.
• logs/: Directory for TensorBoard logs during training.
• oracle-n-model/: Directory containing the saved model.
• oracle-n-tokenizer/: Directory containing the saved tokenizer.
• scripts/: Directory containing additional scripts for training and evaluation.
• .gitignore: Git ignore file to exclude unnecessary files from the repository.
• requirements.txt: File listing the dependencies required for the project.

Setup and Installation

1. Clone the repository:

   git clone https://github.com/hilarl/oracle-n.git
   cd oracle-n
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Download the dataset: Ensure you have the IMDb dataset files in the aclImdb/ directory. If needed, you can download them from the IMDb dataset page and convert them to Parquet format.

Usage

Training the Model

1. Prepare the Dataset: Ensure the dataset files are in the aclImdb/ directory in Parquet format.

2. Run the Training Script:

   python scripts/train_model.py
   

3. Monitor Training with TensorBoard:

   tensorboard --logdir logs
   

Evaluating the Model

1. Run the Evaluation Script:

   python scripts/evaluate_model.py
   

Customizing the Model

1. Modify the Configuration: Edit the dataset.py script to change the model configuration parameters such as hidden size, number of layers, and attention heads.

2. Add Your Own Tokenizer: Customize the tokenizer by editing the oracle-n-tokenizer directory.

Contribution

Contributions are welcome! Please fork the repository and submit a pull request.

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

Acknowledgments

• BERT: This model is based on the BERT architecture developed by Google.
• Hugging Face: Leveraging the Hugging Face Transformers library for model development.

Contact

For any questions or suggestions, please open an issue on GitHub or contact us at [e-mail].

---

This README provides an overview of the Oracle-n model, its features, and how to set up and use it. Feel free to customize it further based on your specific requirements and details.