where is the training notebook?
Thank you!
"For more in detail information, you can take a look at the training notebook."
+1
+1
+1
Hey everyone!
At the moment we are still discussing whether we will release the notebooks.
Believe there wont be a training notebook - just a finetuning one . The model is finetuned on infosec using bloomberg/KBIR .
https://huggingface.co/bloomberg/KBIR.
Bloomberg developers mentioned : As mentioned above since KBIR is built on top of the RoBERTa architecture, it is compatible with any AutoModel setting that RoBERTa is also compatible with.
Sorry for the spam - feel free to delete the discussion. For anyone else looking to Finetune, the KBIR model can be finetuned using standard huggingface/Roberta API.
The training notebook contains the code to fine-tune a KBIR model. So "training" notebook is not the right word.
Sorry for the confusion. Thank you
@shershah11
for pointing this out!
MIDAS is also busy developing a library. Please note that it's still a work in progress.
Github Repository: https://github.com/midas-research/dlkp
If you are more interested in the training part, I suggest you look at this: https://github.com/bloomberg/kbir_keybart. This contains the experimental code of KBIR and KeyBART paper.
I hope this clarifies it a bit more! :)