Fine-tuning microsoft/trocr_small_stage1 for OCR

This repository provides a Jupyter Notebook for fine-tuning the trocr_small_stage1 model for OCR using the Hugging Face Transformers library.

Overview

The notebook provides step-by-step instructions to set up the environment, preprocess data, train the model, and evaluate its performance. It leverages popular libraries like:

• transformers
• datasets
• torch
• torchvision
• streamlit
• uvicorn
• accelerate
• sentencepiece

Setup

To get started with the notebook, make sure you have the necessary dependencies installed. You can install them using the following commands:

```bash pip install -r libraries.txt ```

Usage

1. Clone the repository: ```bash git clone https://github.com/youzaina001/fine_tuned_trocr_small_stage1_OCR.git cd fine_tuned_trocr_small_stage1_OCR ```

2. Open the Jupyter notebook: ```bash jupyter notebook model_training_notebook.ipynb ```

3. Follow the steps in the notebook to set up the environment, load data, train your model, and evaluate its performance.

Key Sections

• Environment Setup: Instructions for installing necessary libraries and configuring the environment.
• Data Preprocessing: Loading and preprocessing the dataset for model training.
• Model Training: Defining the model architecture, training, and fine-tuning.
• Evaluation: Evaluating the model performance using various metrics.
• Deployment: Options for deploying the trained model using tools like FastAPI or Streamlit.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Author

• Yassine Ouzaina