Edit model card

twitter-roberta-base-mar2022-finetuned-emotion

This model is a fine-tuned version of cardiffnlp/twitter-roberta-base-mar2022 on the tweet_eval dataset. It achieves the following results on the evaluation set:

  • Loss: 0.5146
  • Accuracy: 0.8191
  • F1: 0.8171

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 32
  • eval_batch_size: 32
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 2

Training results

Training Loss Epoch Step Validation Loss Accuracy F1
0.8945 1.0 102 0.5831 0.7995 0.7887
0.5176 2.0 204 0.5266 0.8235 0.8200

Framework versions

  • Transformers 4.19.3
  • Pytorch 1.11.0+cu102
  • Datasets 2.2.2
  • Tokenizers 0.12.1
Downloads last month
13

Dataset used to train Tomas23/twitter-roberta-base-mar2022-finetuned-emotion

Evaluation results