JFrediani/Bertimbau-Large-Offensive

content

This model is a fine-tuned version of neuralmind/bert-large-portuguese-cased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4768
• Accuracy: 0.7739
• F1-score: 0.7823
• Recall: 0.9002
• Precision: 0.6917

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: 2.5e-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
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1-score	Recall	Precision
0.5016	0.3814	500	0.4686	0.7736	0.7865	0.9022	0.6971
0.4628	0.7628	1000	0.4437	0.7753	0.7769	0.8464	0.7180
0.4139	1.1442	1500	0.4633	0.7773	0.7573	0.7517	0.7630
0.3569	1.5256	2000	0.5019	0.7831	0.7930	0.8991	0.7093
0.357	1.9069	2500	0.4498	0.7839	0.7644	0.7585	0.7704
0.2612	2.2883	3000	0.6906	0.7665	0.7740	0.8650	0.7003
0.2292	2.6697	3500	0.6406	0.7624	0.7711	0.8656	0.6952
0.2345	3.0511	4000	0.8274	0.7687	0.7502	0.7511	0.7492
0.1527	3.4325	4500	0.8778	0.7602	0.7433	0.7511	0.7356
0.1613	3.8139	5000	0.8756	0.7564	0.7220	0.6842	0.7642
0.1188	4.1953	5500	1.2264	0.7567	0.7317	0.7176	0.7463
0.0992	4.5767	6000	1.2104	0.7636	0.7440	0.7430	0.7449
0.0938	4.9580	6500	1.1858	0.7616	0.7461	0.7579	0.7347

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1