Edit model card

twitter-emotion-deberta-v3-base

This model is a fine-tuned version of DeBERTa-v3. It achieves the following results on the evaluation set:

  • Loss: 0.1474
  • Accuracy: 0.937

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-05
  • train_batch_size: 80
  • eval_batch_size: 80
  • lr_scheduler_type: linear
  • num_epochs: 6.0

Framework versions

  • Transformers 4.12.5
  • Pytorch 1.10.0+cu113
  • Datasets 1.15.1
  • Tokenizers 0.10.3
Downloads last month
44

Dataset used to train Emanuel/twitter-emotion-deberta-v3-base

Spaces using Emanuel/twitter-emotion-deberta-v3-base 2

Evaluation results