File size: 4,637 Bytes
b1e7fea 322968f b1e7fea 322968f b1e7fea 322968f b1e7fea 554c817 b1e7fea bfa4dbb b1e7fea 322968f b1e7fea 1353d86 b1e7fea |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 |
---
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
|