DOC : Example for fine_tuning BERT models

[Thibeb]

Example of fine tuning two BERT models on the QUAERO, E3C and CASM2 corpus

• The 3 corpus have been already converted to medkit text documents format, pre-processed and saved to a jsonl format.
• Fine-tuning was performed on google collab using a T4 GPU, training time approximately took 10 minutes for each BERT models (note that models were fine tuned for 10 epochs).
• The models fine tuned in this example are available on the [medkit hugging face organization](https://pages.github.com/)

[ghisvail]

@Thibeb There are some hardcoded paths left starting with:

/content/drive/MyDrive/datasets/{c}/

Any idea how I can find those files?

[Thibeb]

@Thibeb There are some hardcoded paths left starting with:

/content/drive/MyDrive/datasets/{c}/

Any idea how I can find those files?

Yes I can give you a link to those files on my personal Drive, it will be much simpler than trying to recreate them using the scripts that I provided.

Here's the link : https://drive.google.com/drive/folders/1tjMtBUkr5Qatfbj7z6GBOM-qZM6rPLTQ?usp=drive_link

[ghisvail]

Here's the link : https://drive.google.com/drive/folders/1tjMtBUkr5Qatfbj7z6GBOM-qZM6rPLTQ?usp=drive_link

I just sent you a request for access.