TrainExportToONNX

This project demonstrates a text classification pipeline using BERT, PyTorch, and ONNX. It includes data preprocessing, model training, evaluation, and exporting the trained model to ONNX format.

Features

• Text classification using BERT
• Data preprocessing and cleaning
• Model training and evaluation
• Exporting the trained model to ONNX format

Prerequisites

• Python 3.9 or later
• Install dependencies using the provided requirements.txt file:

py -m pip install -r requirements.txt

Files

• sample_script.py: Main script for training and evaluating the model.
• create_json.py: Script to generate a sample JSON dataset (L1Files.json).
• L1Files.json: Sample dataset for training and testing.
• requirements.txt: List of required Python packages.

Usage

1. Generate the Sample Dataset: Run the create_json.py script to create the L1Files.json file:

   py create_json.py

2. Train and Evaluate the Model: Run the sample_script.py script:

   py sample_script.py

   Follow the prompts to load a saved model or train a new one. The script will output the model's accuracy, classification report, and confusion matrix.

3. Export the Model to ONNX: If you choose to save the trained model, it will also be exported to ONNX format as bert_sequence_classification.onnx.

Notes

• The sample dataset (L1Files.json) is small and intended for demonstration purposes. Replace it with a larger, real-world dataset for better results.
• Ensure that the transformers and torch libraries are compatible with your Python version.

License

This project is licensed under the MIT License.