This is a modified version of the ICTMCG/M3FEND repository for Multimodal Multitask Fake News Detection. The key modifications are the integration with PyTorch Lightning, a lightweight PyTorch wrapper that helps to organize PyTorch code, and TensorBoard, a tool for providing the measurements and visualizations needed during the machine learning workflow.
- Compatibility with TensorBoard for visualizing the performance of the model and identifying bottlenecks in the code.
- Utilizes the simplicity and flexibility of PyTorch Lightning to organize PyTorch code and easily run it on CPUs, GPUs or TPUs.
- Custom data loader for efficient data processing and preparation.
- Extensive configurability for hyperparameters and training setup.
- Python 3.8+
- PyTorch
- PyTorch Lightning
- HuggingFace Transformers
After setting up the project, you can run the model training with the following command:
python main.pyThe model parameters are saved in the ./params directory and the TensorBoard logs are saved in the ./logs/my_experiment/M3FEND directory. The model's performance on the test set is printed at the end of the script.
You can modify this script to suit your needs. Some common modifications might include:
- Adjusting the model's hyperparameters.
- Adding additional callbacks or changing the logging settings.
- Modifying the data loading and processing logic.
- Changing the model architecture or training strategy.
Remember to also update the ModelFactory and MyDataloader classes to suit your new configuration if necessary.
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.