metadata

license: mit
base_model: roberta-base
tags:
  - generated_from_trainer
metrics:
  - accuracy
  - precision
  - recall
  - f1
model-index:
  - name: irony_pt_Brazil
    results: []

irony_pt_Brazil

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

Loss: 0.0033
Accuracy: 0.6463
Precision: 0.44
Recall: 0.5739
F1: 0.4981

Model description

The model is trained considering the annotation of annotators from Brazil only, on instances in Portuguese (PT and BZ linguistic varieties). 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 Brazil from the MultiPICo dataset (instances in Portuguese). The data has been randomly split into 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

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.0043	1.0	71	0.0042	0.5745	0.3448	0.4348	0.3846
0.0043	2.0	142	0.0041	0.5266	0.3562	0.6783	0.4671
0.0039	3.0	213	0.0039	0.5266	0.3562	0.6783	0.4671
0.004	4.0	284	0.0038	0.6170	0.4110	0.5826	0.4820
0.0036	5.0	355	0.0035	0.6516	0.4452	0.5652	0.4981
0.0035	6.0	426	0.0036	0.4973	0.3630	0.8522	0.5091
0.0031	7.0	497	0.0033	0.5904	0.4156	0.8348	0.5549
0.0027	8.0	568	0.0033	0.6543	0.4460	0.5391	0.4882
0.0027	9.0	639	0.0031	0.6144	0.4257	0.7478	0.5426
0.0023	10.0	710	0.0029	0.6303	0.4388	0.7478	0.5531
0.0021	11.0	781	0.0031	0.6383	0.4348	0.6087	0.5072
0.0018	12.0	852	0.0033	0.6463	0.44	0.5739	0.4981

Framework versions

Transformers 4.34.1
Pytorch 2.0.1+cu117
Datasets 2.14.5
Tokenizers 0.14.1