|
--- |
|
language: es |
|
thumbnail: https://i.imgur.com/uxAvBfh.png |
|
tags: |
|
- Spanish |
|
- Electra |
|
|
|
datasets: |
|
-large_spanish_corpus |
|
|
|
--- |
|
|
|
## ELECTRICIDAD: The Spanish Electra [Imgur](https://imgur.com/uxAvBfh) |
|
|
|
**Electricidad-base-discriminator** (uncased) is a ```base``` Electra like model (discriminator in this case) trained on a [Large Spanish Corpus](https://github.com/josecannete/spanish-corpora) (aka BETO's corpus) |
|
|
|
As mentioned in the original [paper](https://openreview.net/pdf?id=r1xMH1BtvB): |
|
**ELECTRA** is a new method for self-supervised language representation learning. It can be used to pre-train transformer networks using relatively little compute. ELECTRA models are trained to distinguish "real" input tokens vs "fake" input tokens generated by another neural network, similar to the discriminator of a [GAN](https://arxiv.org/pdf/1406.2661.pdf). At small scale, ELECTRA achieves strong results even when trained on a single GPU. At large scale, ELECTRA achieves state-of-the-art results on the [SQuAD 2.0](https://rajpurkar.github.io/SQuAD-explorer/) dataset. |
|
|
|
For a detailed description and experimental results, please refer the paper [ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators](https://openreview.net/pdf?id=r1xMH1BtvB). |
|
|
|
|
|
## Model details ⚙ |
|
|
|
|Name| # Value| |
|
|-----|--------| |
|
|Layers| 12 | |
|
|Hidden | 768 | |
|
|Params| 110M | |
|
|
|
## Evaluation metrics (for discriminator) 🧾 |
|
|
|
|Metric | # Score | |
|
|-------|---------| |
|
|Accuracy| 0.985| |
|
|Precision| 0.726| |
|
|AUC | 0.922| |
|
|
|
|
|
|
|
## Fast example of usage 🚀 |
|
|
|
```python |
|
from transformers import ElectraForPreTraining, ElectraTokenizerFast |
|
import torch |
|
|
|
discriminator = ElectraForPreTraining.from_pretrained("mrm8488/electricidad-base-discriminator") |
|
tokenizer = ElectraTokenizerFast.from_pretrained("mrm8488/electricidad-base-discriminator") |
|
|
|
sentence = "El rápido zorro marrón salta sobre el perro perezoso" |
|
fake_sentence = "El rápido zorro marrón amar sobre el perro perezoso" |
|
|
|
fake_tokens = tokenizer.tokenize(fake_sentence) |
|
fake_inputs = tokenizer.encode(fake_sentence, return_tensors="pt") |
|
discriminator_outputs = discriminator(fake_inputs) |
|
predictions = torch.round((torch.sign(discriminator_outputs[0]) + 1) / 2) |
|
|
|
[print("%7s" % token, end="") for token in fake_tokens] |
|
|
|
[print("%7s" % prediction, end="") for prediction in predictions.tolist()] |
|
|
|
# Output: |
|
''' |
|
el rapido zorro marro ##n amar sobre el perro pere ##zoso 0.0 0.0 0.0 0.0 0.0 0.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0[None, None, None, None, None, None, None, None, None, None, None, None, None |
|
''' |
|
``` |
|
As you can see there are **1s** in the places where the model detected a fake token. So, it works! 🎉 |
|
|
|
|
|
### Some models fine-tuned on a downstream task 🛠️ |
|
|
|
[Question Answering](https://huggingface.co/mrm8488/electricidad-base-finetuned-squadv1-es) |
|
|
|
[POS](https://huggingface.co/mrm8488/electricidad-base-finetuned-pos) |
|
|
|
[NER](https://huggingface.co/mrm8488/electricidad-base-finetuned-ner) |
|
|
|
|
|
### Spanish LM model comparison 📊 |
|
| Dataset | Metric | RoBERTa-b | RoBERTa-l | BETO | mBERT | BERTIN | Electricidad-b | |
|
|-------------|----------|-----------|-----------|--------|--------|--------|---------| |
|
| UD-POS | F1 | 0.9907 | 0.9901 | 0.9900 | 0.9886 | 0.9904 | 0.9818 | |
|
| Conll-NER | F1 | 0.8851 | 0.8772 | 0.8759 | 0.8691 | 0.8627 | 0.7954 | |
|
| Capitel-POS | F1 | 0.9846 | 0.9851 | 0.9836 | 0.9839 | 0.9826 | 0.9816 | |
|
| Capitel-NER | F1 | 0.8959 | 0.8998 | 0.8771 | 0.8810 | 0.8741 | 0.8035 | |
|
| STS | Combined | 0.8423 | 0.8420 | 0.8216 | 0.8249 | 0.7822 | 0.8065 | |
|
| MLDoc | Accuracy | 0.9595 | 0.9600 | 0.9650 | 0.9560 | 0.9673 | 0.9490 | |
|
| PAWS-X | F1 | 0.9035 | 0.9000 | 0.8915 | 0.9020 | 0.8820 | **0.9045** | |
|
| XNLI | Accuracy | 0.8016 | 0.7958 | 0.8130 | 0.7876 | 0.7864 | 0.7878 | |
|
|
|
|
|
|
|
## Acknowledgments |
|
|
|
I thank [🤗/transformers team](https://github.com/huggingface/transformers) for allowing me to train the model (specially to [Julien Chaumond](https://twitter.com/julien_c)). |
|
|
|
|
|
## Citation |
|
If you want to cite this model you can use this: |
|
|
|
```bibtex |
|
@misc{mromero2020electricidad-base-discriminator, |
|
title={Spanish Electra by Manuel Romero}, |
|
author={Romero, Manuel}, |
|
publisher={Hugging Face}, |
|
journal={Hugging Face Hub}, |
|
howpublished={\url{https://huggingface.co/mrm8488/electricidad-base-discriminator/}}, |
|
year={2020} |
|
} |
|
``` |
|
|
|
|
|
|
|
> Created by [Manuel Romero/@mrm8488](https://twitter.com/mrm8488) |
|
|
|
> Made with <span style="color: #e25555;">♥</span> in Spain |
|
|
