ELECTRICIDAD: The Spanish Electra Imgur

Electricidad-base-discriminator (uncased) is a base Electra like model (discriminator in this case) trained on a Large Spanish Corpus (aka BETO's corpus)

As mentioned in the original paper: 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. 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 dataset.

For a detailed description and experimental results, please refer the paper ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators.

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 🚀

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

POS

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 for allowing me to train the model (specially to Julien Chaumond).

Created by Manuel Romero/@mrm8488

Made with in Spain

Downloads last month
249
Hosted inference API

Unable to determine this model’s pipeline type. Check the docs .