README.md

---
language: es
tags:
 - Spanish
 - Electra
 - Bio
 - Medical

datasets:
 - cowese
---

## 🦠 BIOMEDtra 🏥

**BIOMEDtra** (small) is an Electra like model (discriminator in this case) trained on [Spanish Biomedical Crawled Corpus](https://zenodo.org/record/5510033#.Yhdk1ZHMLJx).

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).


## Training details

The model was trained using the Electra base code for 3 days on 1 GPU (Tesla V100 16GB).

## Dataset details

The largest Spanish biomedical and heath corpus to date gathered from a massive Spanish health domain crawler over more than 3,000 URLs were downloaded and preprocessed. The collected data have been preprocessed to produce the **CoWeSe** (Corpus Web Salud Español) resource, a large-scale and high-quality corpus intended for biomedical and health NLP in Spanish.

## Model details ⚙

|Param| # Value|
|-----|--------|
|Layers| 12 |
|Hidden | 256 |
|Params| 14M |

## Evaluation metrics (for discriminator) 🧾

|Metric | # Score |
|-------|---------|
|Accuracy| 0.9561|
|Precision| 0.808|
|Recall | 0.531 |
|AUC | 0.949|


## Benchmarks 🔨

WIP 🚧

## How to use the discriminator in `transformers`
```py
from transformers import ElectraForPreTraining, ElectraTokenizerFast
import torch

discriminator = ElectraForPreTraining.from_pretrained("mrm8488/biomedtra-small-es")
tokenizer = ElectraTokenizerFast.from_pretrained("mrm8488/biomedtra-small-es")

sentence = "Los españoles tienden a sufir déficit de vitamina c"
fake_sentence = "Los españoles tienden a déficit sufrir de vitamina c"

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()]
```

## Acknowledgments

TBA

## Citation
If you want to cite this model you can use this:

```bibtex
@misc{mromero2022biomedtra,
  title={Spanish BioMedical Electra (small)},
  author={Romero, Manuel},
  publisher={Hugging Face},
  journal={Hugging Face Hub},
  howpublished={\url{https://huggingface.co/mrm8488/biomedtra-small-es},
  year={2022}
}
```


> Created by [Manuel Romero/@mrm8488](https://twitter.com/mrm8488)

> Made with <span style="color: #e25555;">&hearts;</span> in Spain