Recently have trained a model on the DocVQA English dataset. Am able to run inference with the model's pdiparams file after running the export_model.py.

Now, want to train the model to understand a few business specific document types but don't want to start from PaddlePaddle/ernie-layoutx-base-uncased. Can anyone provide guidance on how I can re-train a new model based off of an old model?