AIKIA_0.6_OFFENSIVE_BERT_MULTILINGUAL

This model is a fine-tuned version of bert-base-multilingual-uncased on the AIKIA Greek dataset with 0.6 threshold for the Offensive label. It achieves the following results on the evaluation set:

• Loss: 1.5183
• Macro F1: 0.6983
• Micro F1: 0.7565
• Accuracy: 0.7565

Results on test set:

• Accuracy: 0.7784313725490196

• F1 score: 0.7575514222182127

• Precision: 0.7545776520467404

• Recall : 0.7612935323383084

• Matthews Correlation Coefficient: 0.5158274671154518

• Precision of each class: [0.84188912 0.66726619]

• Recall of each class: [0.8159204 0.70666667]

• F1 score of each class: [0.82870136 0.68640148]

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Macro F1	Micro F1	Accuracy
No log	1.0	338	0.5701	0.7024	0.7122	0.7122
0.5808	2.0	676	0.5355	0.7338	0.7565	0.7565
0.4115	3.0	1014	0.6522	0.7206	0.7528	0.7528
0.4115	4.0	1352	0.8994	0.6287	0.7306	0.7306
0.2334	5.0	1690	0.9278	0.7000	0.7417	0.7417
0.1489	6.0	2028	1.3355	0.7048	0.7491	0.7491
0.1489	7.0	2366	1.5183	0.6983	0.7565	0.7565

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.1
• Tokenizers 0.13.3