Edit model card

irony_en_United_Kingdom

This model is a fine-tuned version of roberta-base on part of the EPIC dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0023
  • Accuracy: 0.6833
  • Precision: 0.4764
  • Recall: 0.7339
  • F1: 0.5778

Model description

The model is trained considering the annotation of annotators from the United Kingdom only. The annotations from these annotators are aggregated using majority voting and then used to train the model.

Training and evaluation data

The model has been trained on the annotation from annotators from the United Kingdom from the EPIC dataset. The data has been randomly split in a train and a validation set.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-06
  • train_batch_size: 16
  • eval_batch_size: 64
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 20
  • early stopping (patience: 2)

Training results

Training Loss Epoch Step Validation Loss Accuracy Precision Recall F1
0.0044 1.0 79 0.0042 0.5310 0.3612 0.7661 0.4910
0.0043 2.0 158 0.0040 0.6595 0.4417 0.5806 0.5017
0.0039 3.0 237 0.0034 0.6310 0.4188 0.6452 0.5079
0.0033 4.0 316 0.0027 0.7286 0.5352 0.6129 0.5714
0.0022 5.0 395 0.0024 0.5833 0.4066 0.8952 0.5592
0.0015 6.0 474 0.0022 0.7357 0.5474 0.6048 0.5747
0.001 7.0 553 0.0022 0.7262 0.5302 0.6371 0.5788
0.0005 8.0 632 0.0023 0.6833 0.4764 0.7339 0.5778

Framework versions

  • Transformers 4.34.1
  • Pytorch 2.0.1+cu117
  • Datasets 2.14.5
  • Tokenizers 0.14.1
Downloads last month
5

Finetuned from