|
--- |
|
library_name: transformers |
|
license: mit |
|
base_model: bert-base-multilingual-cased |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: fine_tune_bert_output |
|
results: [] |
|
datasets: |
|
- unimelb-nlp/wikiann |
|
language: |
|
- es |
|
metrics: |
|
- recall |
|
- precision |
|
- f1 |
|
pipeline_tag: token-classification |
|
--- |
|
|
|
 |
|
|
|
--- |
|
# **spanish_bert_based_ner** |
|
--- |
|
|
|
# fine_tune_bert_output |
|
|
|
This model is a fine-tuned version of [bert-base-multilingual-cased](https://huggingface.co/bert-base-multilingual-cased) on an [wikiann](https://huggingface.co/datasets/unimelb-nlp/wikiann) dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.3320 |
|
- Overall Precision: 0.9051 |
|
- Overall Recall: 0.9121 |
|
- Overall F1: 0.9086 |
|
- Overall Accuracy: 0.9577 |
|
- Loc F1: 0.9190 |
|
- Org F1: 0.8663 |
|
- Per F1: 0.9367 |
|
|
|
## Labels |
|
The following table represents the labels used by the model along with their corresponding indices: |
|
|
|
| Index | Label | |
|
|-------|---------| |
|
| 0 | O | |
|
| 1 | B-PER | |
|
| 2 | I-PER | |
|
| 3 | B-ORG | |
|
| 4 | I-ORG | |
|
| 5 | B-LOC | |
|
| 6 | I-LOC | |
|
|
|
### Label Descriptions |
|
- **O**: Outside of a named entity. |
|
- **B-PER**: Beginning of a person's name. |
|
- **I-PER**: Inside a person's name. |
|
- **B-ORG**: Beginning of an organization's name. |
|
- **I-ORG**: Inside an organization's name. |
|
- **B-LOC**: Beginning of a location name. |
|
- **I-LOC**: Inside a location name. |
|
|
|
## Inference Example |
|
```python |
|
from transformers import pipeline |
|
|
|
# Load the model |
|
ner_pipeline = pipeline("ner", model="syubraj/spanish_bert_based_ner") |
|
|
|
# Example text |
|
text = "Elon Musk vive en Estados Unidos y es dueño de Space X, Tesla y Starlink" |
|
|
|
# Perform inference |
|
entities = ner_pipeline(text) |
|
for ent in entities: |
|
print(f"Word: {ent['word']} | Label: {ent['entity']}") |
|
|
|
``` |
|
|
|
## 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: 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 | Overall Precision | Overall Recall | Overall F1 | Overall Accuracy | Loc F1 | Org F1 | Per F1 | |
|
|:-------------:|:-----:|:-----:|:---------------:|:-----------------:|:--------------:|:----------:|:----------------:|:------:|:------:|:------:| |
|
| 0.2713 | 0.8 | 1000 | 0.2236 | 0.8498 | 0.8672 | 0.8584 | 0.9401 | 0.8834 | 0.8019 | 0.8790 | |
|
| 0.1537 | 1.6 | 2000 | 0.1909 | 0.8772 | 0.8943 | 0.8857 | 0.9495 | 0.9002 | 0.8369 | 0.9164 | |
|
| 0.1152 | 2.4 | 3000 | 0.2095 | 0.8848 | 0.8981 | 0.8914 | 0.9523 | 0.9039 | 0.8432 | 0.9220 | |
|
| 0.0889 | 3.2 | 4000 | 0.2223 | 0.8978 | 0.8998 | 0.8988 | 0.9546 | 0.9080 | 0.8569 | 0.9290 | |
|
| 0.0701 | 4.0 | 5000 | 0.2152 | 0.8937 | 0.9042 | 0.8989 | 0.9544 | 0.9113 | 0.8565 | 0.9246 | |
|
| 0.0457 | 4.8 | 6000 | 0.2365 | 0.9017 | 0.9069 | 0.9043 | 0.9563 | 0.9164 | 0.8616 | 0.9310 | |
|
| 0.0364 | 5.6 | 7000 | 0.2622 | 0.9037 | 0.9086 | 0.9061 | 0.9578 | 0.9148 | 0.8639 | 0.9365 | |
|
| 0.026 | 6.4 | 8000 | 0.2916 | 0.9037 | 0.9159 | 0.9097 | 0.9585 | 0.9183 | 0.8712 | 0.9366 | |
|
| 0.0215 | 7.2 | 9000 | 0.2985 | 0.9022 | 0.9128 | 0.9074 | 0.9565 | 0.9178 | 0.8676 | 0.9323 | |
|
| 0.0134 | 8.0 | 10000 | 0.3071 | 0.904 | 0.9131 | 0.9085 | 0.9574 | 0.9198 | 0.8671 | 0.9344 | |
|
| 0.0091 | 8.8 | 11000 | 0.3335 | 0.9056 | 0.9115 | 0.9085 | 0.9573 | 0.9175 | 0.8670 | 0.9373 | |
|
| 0.0074 | 9.6 | 12000 | 0.3320 | 0.9051 | 0.9121 | 0.9086 | 0.9577 | 0.9190 | 0.8663 | 0.9367 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.44.2 |
|
- Pytorch 2.4.1+cu121 |
|
- Datasets 3.2.0 |
|
- Tokenizers 0.19.1 |
