ModernBERT-base-2-contract-sections-classification-v4-10-max

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

• Loss: 0.4801
• Accuracy Evaluate: 0.917
• Precision Evaluate: 0.9221
• Recall Evaluate: 0.9209
• F1 Evaluate: 0.9210
• Accuracy Sklearn: 0.917
• Precision Sklearn: 0.9173
• Recall Sklearn: 0.917
• F1 Sklearn: 0.9166
• Acuracia Rotulo Objeto: 0.9649
• Acuracia Rotulo Obrigacoes: 0.8838
• Acuracia Rotulo Valor: 0.8166
• Acuracia Rotulo Vigencia: 0.9816
• Acuracia Rotulo Rescisao: 0.9474
• Acuracia Rotulo Foro: 0.9385
• Acuracia Rotulo Reajuste: 0.9039
• Acuracia Rotulo Fiscalizacao: 0.8423
• Acuracia Rotulo Publicacao: 0.9951
• Acuracia Rotulo Pagamento: 0.8913
• Acuracia Rotulo Casos Omissos: 0.9015
• Acuracia Rotulo Sancoes: 0.9266
• Acuracia Rotulo Dotacao Orcamentaria: 0.9780

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: 1e-06
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy Evaluate	Precision Evaluate	Recall Evaluate	F1 Evaluate	Accuracy Sklearn	Precision Sklearn	Recall Sklearn	F1 Sklearn	Acuracia Rotulo Objeto	Acuracia Rotulo Obrigacoes	Acuracia Rotulo Valor	Acuracia Rotulo Vigencia	Acuracia Rotulo Rescisao	Acuracia Rotulo Foro	Acuracia Rotulo Reajuste	Acuracia Rotulo Fiscalizacao	Acuracia Rotulo Publicacao	Acuracia Rotulo Pagamento	Acuracia Rotulo Casos Omissos	Acuracia Rotulo Sancoes	Acuracia Rotulo Dotacao Orcamentaria
2.3971	1.0	2000	0.8286	0.7678	0.8292	0.7831	0.7914	0.7678	0.8077	0.7678	0.7695	0.9504	0.6734	0.5244	0.6115	0.8975	0.8808	0.6512	0.7382	0.9113	0.7754	0.8227	0.7982	0.9451
1.3917	2.0	4000	0.6639	0.8558	0.8777	0.8708	0.8702	0.8558	0.8673	0.8558	0.8557	0.9587	0.7357	0.6619	0.8976	0.9030	0.9231	0.8719	0.7918	0.9655	0.8587	0.8867	0.8991	0.9670
0.887	3.0	6000	0.5830	0.874	0.8789	0.8885	0.8804	0.874	0.8812	0.874	0.8743	0.9442	0.7761	0.7421	0.9081	0.8615	0.9346	0.9004	0.8265	0.9852	0.8841	0.9015	0.9083	0.9780
0.9516	4.0	8000	0.5632	0.8885	0.8987	0.8992	0.8979	0.8885	0.8899	0.8885	0.8881	0.9194	0.8081	0.7736	0.9554	0.9474	0.9385	0.8719	0.8297	0.9803	0.8877	0.8818	0.9174	0.9780
0.7614	5.0	10000	0.5290	0.8998	0.9083	0.9102	0.9084	0.8998	0.9021	0.8998	0.8998	0.9628	0.8064	0.8052	0.9370	0.9418	0.9846	0.8897	0.8328	0.9951	0.8804	0.9015	0.9174	0.9780
0.6291	6.0	12000	0.5590	0.8978	0.9029	0.9089	0.9044	0.8978	0.9009	0.8978	0.8975	0.9421	0.7946	0.7880	0.9764	0.9391	0.9385	0.9075	0.8454	0.9951	0.8913	0.9064	0.9083	0.9835
0.4869	7.0	14000	0.4875	0.9103	0.9140	0.9156	0.9143	0.9103	0.9102	0.9103	0.9096	0.9587	0.8737	0.7966	0.9790	0.9446	0.9346	0.9039	0.8139	0.9951	0.8913	0.9064	0.9266	0.9780
0.5691	8.0	16000	0.4929	0.9083	0.9119	0.9162	0.9134	0.9083	0.9093	0.9083	0.9079	0.9607	0.8232	0.8195	0.9843	0.9446	0.9385	0.8932	0.8580	0.9951	0.8913	0.9064	0.9174	0.9780
0.3831	9.0	18000	0.4892	0.9153	0.9184	0.9198	0.9186	0.9153	0.9157	0.9153	0.9149	0.9649	0.8737	0.8138	0.9843	0.9446	0.9385	0.8968	0.8486	0.9951	0.8913	0.9015	0.9266	0.9780
0.3	10.0	20000	0.4801	0.917	0.9221	0.9209	0.9210	0.917	0.9173	0.917	0.9166	0.9649	0.8838	0.8166	0.9816	0.9474	0.9385	0.9039	0.8423	0.9951	0.8913	0.9015	0.9266	0.9780

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.0
• Tokenizers 0.21.0