bert-large-cased-finetuned-ner-lenerBr

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

• Loss: nan
• Precision: 0.8046
• Recall: 0.8298
• F1: 0.8170
• Accuracy: 0.9645

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.9974	244	nan	0.6627	0.7490	0.7032	0.9402
No log	1.9990	489	nan	0.7002	0.8005	0.7470	0.9503
0.1592	2.9964	733	nan	0.7482	0.8080	0.7769	0.9545
0.1592	3.9980	978	nan	0.7749	0.8166	0.7952	0.9614
0.0279	4.9995	1223	nan	0.7845	0.7973	0.7909	0.9634
0.0279	5.9969	1467	nan	0.7840	0.8203	0.8017	0.9622
0.0122	6.9985	1712	nan	0.7989	0.8224	0.8105	0.9638
0.0122	8.0	1957	nan	0.7977	0.8286	0.8129	0.9634
0.007	8.9974	2201	nan	0.7947	0.8265	0.8103	0.9643
0.007	9.9745	2440	nan	0.8046	0.8298	0.8170	0.9645

Framework versions

• Transformers 4.41.1
• Pytorch 2.1.2
• Datasets 2.19.1
• Tokenizers 0.19.1