KaiquanMah's picture
Update README.md
daadcab verified
metadata
license: mit
datasets:
  - PolyAI/banking77
language:
  - en
base_model:
  - google-bert/bert-base-uncased
metrics:
  - f1

BERT finetuned on Banking 77 Open Intent Classification Dataset

This is a BERT model finetuned on the PolyAI/banking77 dataset.

Training Configuration

  • PRETRAINED_MODEL_NAME = "bert-base-uncased"
  • BATCH_SIZE = 128
  • LR_PRETRAIN = 2e-5
  • EPOCHS_PRETRAIN = 20
  • DEVICE = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
  • MAX_LEN = 128, to truncate long sequences down to 128 tokens, or pad short sequences up to 128 tokens