bertimbau-base-finetuned-brazilian_court_decisions_bt8_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: 1.814463
• Accuracy: 0.777228

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	0.780298	0.663366
2	0.827000	0.739960	0.705446
3	0.597100	0.775997	0.737624
4	0.413100	0.860354 0. 767327
5	0.296900	1.024743	0.767327
6	0.296900	1.274581	0.757426
7	0.193900	1.316006	0.787129
8	0.125700	1.486973	0.764851
9	0.102200	1.651981	0.777228
10	0.053300	1.640268	0.787129
11	0.053300	1.762075	0.782178
12	0.028500	1.796041	0.779703
13	0.016000	1.801749	0.784653
14	0.014100	1.784914	0.779703
15	0.012600	1.814463	0.777228

Framework versions

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