Fine-tune DeBERTa model parameters for text classification

Jason-J-Choi / DeBERTa_TxtClassifier

Adversarial system using DeBERTa as base model, fine-tuned for text classification, and identifying robustness via TextFooler

6 stars 2 forks source link

Closed Jason-J-Choi closed 3 years ago

Jason-J-Choi commented 3 years ago

Requires #1

Jason-J-Choi commented 3 years ago

fine-tuning

[x] mr
[x] fake
[x] ag
[x] imdb
[x] yelp - due to the large size, each epoch is estimated at about 50 hours. Dataset was segmented and training slowly