clasificador-ser-estar-window-5-bert-base

This model is a fine-tuned version of google-bert/bert-base-multilingual-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9606
• F1 Score: 0.8491
• Recall: 0.8903
• Precision: 0.8115
• Roc Auc: 0.9028

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Score	Recall	Precision	Roc Auc
No log	1.0	25	0.6774	0.7465	1.0	0.5955	0.5843
No log	2.0	50	0.5260	0.7636	0.7089	0.8276	0.8132
No log	3.0	75	0.4120	0.8393	0.9367	0.7603	0.8954
No log	4.0	100	0.4838	0.8363	0.8945	0.7852	0.8914
No log	5.0	125	0.4970	0.8267	0.8354	0.8182	0.8933
No log	6.0	150	0.5585	0.8381	0.8734	0.8054	0.8932
No log	7.0	175	0.5102	0.84	0.8861	0.7985	0.8960
No log	8.0	200	0.7016	0.8316	0.8439	0.8197	0.8892
No log	9.0	225	0.8863	0.8313	0.8523	0.8112	0.8793
No log	10.0	250	0.8687	0.8273	0.7679	0.8966	0.8876
No log	11.0	275	0.8506	0.8457	0.8903	0.8053	0.8926
No log	12.0	300	0.9160	0.8393	0.7932	0.8910	0.8951
No log	13.0	325	1.0401	0.8352	0.7806	0.8981	0.8918
No log	14.0	350	1.0209	0.8361	0.8608	0.8127	0.8913
No log	15.0	375	0.9580	0.8438	0.7975	0.8957	0.9036
No log	16.0	400	0.9606	0.8491	0.8903	0.8115	0.9028
No log	17.0	425	1.0079	0.8413	0.8945	0.7940	0.9000
No log	18.0	450	1.0042	0.8463	0.8945	0.8030	0.9015
No log	19.0	475	0.9886	0.8474	0.8903	0.8084	0.9006
0.2037	20.0	500	1.0047	0.8457	0.8903	0.8053	0.9012

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0