T2T-BinFormer Pre-trained Model

Model Link

You can download the pre-trained model from Google Drive: T2T-BinFormer Model.

Pre-processing Data

To preprocess the data, run the following command with the appropriate parameters:

python process_dibco2.py --data_path /path/ --split_size 256 --testing_dataset 2018 --validation_dataset 2016

Training the Model

To train the model with specified parameters, use the command below:

python train2.py --data_path /path/ --batch_size 32 --vit_model_size base --vit_patch_size 16 --epochs 1000 --split_size 256 --validation_dataset 2016

Results

Acknowledgements

Special thanks to Risab Biswas for the initial setup and model configuration provided in his GitHub repository.

Note

This model training is a preliminary attempt at entering the field of computer science. The process and results might contain errors due to my novice level in this field.