bert-ner-conll2003

This model is a fine-tuned version of bert-base-uncased on the conll2003 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0631
• Precision: 0.9414
• Recall: 0.9493
• F1: 0.9454
• Accuracy: 0.9866

Model description

from transformers import AutoTokenizer, AutoModelForTokenClassification
from transformers import pipeline

tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
model = AutoModelForTokenClassification.from_pretrained("PassbyGrocer/bert-ner-conll2003")

nlp = pipeline("ner", model=model, tokenizer=tokenizer)
example = "My name is Wolfgang and I live in Berlin."

ner_results = nlp(example)
print(ner_results)

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use 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: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.0844	1.0	878	0.0693	0.9029	0.9201	0.9114	0.9806
0.0216	2.0	1756	0.0559	0.9340	0.9444	0.9391	0.9854
0.0206	3.0	2634	0.0569	0.9436	0.9447	0.9442	0.9863
0.0141	4.0	3512	0.0634	0.9369	0.9488	0.9428	0.9860
0.0176	5.0	4390	0.0631	0.9414	0.9493	0.9454	0.9866

Framework versions

• Transformers 4.46.1
• Pytorch 1.13.1+cu116
• Datasets 3.1.0
• Tokenizers 0.20.1