beto-base-cased-finetuned-conll2002

This model is a fine-tuned version of NazaGara/NER-fine-tuned-BETO on the conll2002 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1281
• Precision: 0.8608
• Recall: 0.8610
• F1: 0.8609
• Accuracy: 0.9787

Model description

The model described here is a fine-tuned version of the BETO (BERT-based Spanish language model) for Named Entity Recognition (NER) tasks, trained on the CoNLL-2002 dataset. BETO is a pre-trained language model specifically designed for the Spanish language, based on the BERT architecture. By fine-tuning BETO on the CoNLL-2002 dataset, the model has been adapted to recognize and classify named entities such as persons, organizations, locations, and other miscellaneous entities within Spanish text. The fine-tuning process involves adjusting the pre-trained model weights to better fit the specific task of NER, thereby improving its performance and accuracy on Spanish text.

Intended uses & limitations

More information needed

Training and evaluation data

The training was performed using a GPU with 22.5 GB of RAM, 53 GB of system RAM, and 200 GB of disk space. This setup ensured efficient handling of the large dataset and the computational demands of fine-tuning the model.

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: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.019	1.0	1041	0.1095	0.8581	0.8628	0.8604	0.9790
0.018	2.0	2082	0.1121	0.8461	0.8589	0.8525	0.9783
0.0133	3.0	3123	0.1281	0.8608	0.8610	0.8609	0.9787

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1