I'm working on a project where I need to fine-tune pair-classification-roberta-snli model offered by AllenNLP. I have prepared my custom dataset in the snli format but couldn't manage to find a way to retrain the model. Currently, I am following this approach to train bert-base model for textual entailment. But how to fine-tune AllenNLP's pair-classification-roberta-snli model?
Asked
Active
Viewed 545 times
0
-
You can start with this training config, but change the "model" part to use the "from_archive" type: https://github.com/allenai/allennlp-models/blob/39557dd8a85ffedd195033ef760c1ac53caaccf9/training_config/pair_classification/snli_roberta.jsonnet#L24-L45. For example, `"model": {"type": "from_archive", "archive_file": "gs://allennlp-public-models/snli-roberta.2021-03-11.tar.gz"}` – petew Oct 22 '21 at 22:18
-
This was a great help :D – Sujan Dutta Jan 22 '22 at 16:16