bertimbau-base-finetuned-brazilian_court_decisions_bt16_ep15

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

• Loss: 1.523958
• Accuracy: 0.772277

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Epoch	Training Loss	Validation Loss	Accuracy
1	No log	0.852318	0.603960
2	No log	0.728222	0.660891
3	0.781100	0.662818	0.742574
4	0.781100	0.687966	0.742574
5	0.399400	0.727256	0.762376
6	0.399400	0.843507	0.762376
7	0.399400	0.936927	0.759901
8	0.182400	1.065885	0.769802
9	0.182400	1.154641	0.754950
10	0.082200	1.375061	0.745050
11	0.082200	1.377540	0.757426
12	0.082200	1.465057	0.759901
13	0.033800	1.497934	0.762376
14	0.033800	1.504722	0.769802
15	0.017900	1.523958	0.772277

Framework versions

• Transformers 4.24.0
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1