How to fine-tune Docling Heron model?

I noticed the repo doesn't provide a general training script, so wondering where to go from here, and what the format my dataset should be in

Start by reviewing the repository's existing Docling Heron model and training-related material; the issue does not name specific files or entry points. Define what a general fine-tuning script should accept and document the expected dataset format, with completion demonstrated by a usable training path for the model.

Written by the indexing model from the issue text.