Further fine-tuned the language model with custom corpus

Hi, the repository is amazing and the architecture is fantastic stuff.
Along with this, I have below query:
I have my own **text file** which contains sentences per line, how can I use this my custom corpus to further fine-tuned the trained language model, as I want to perform **domain-specific text similarity task**.
Could you suggest any way to fine-tune the model with the **unsupervised** way as I do not have a labeled dataset?

The issue names no file, test, or entry point. First clarify the expected unsupervised fine-tuning workflow for a text file and domain-specific similarity task; completion would require an agreed implementation or documented supported approach.

Written by the indexing model from the issue text.