|
--- |
|
language: es |
|
license: cc-by-sa-4.0 |
|
datasets: |
|
- wikipedia |
|
- cc100 |
|
widget: |
|
- text: "Yo soy" |
|
- text: "Me llamo" |
|
--- |
|
|
|
## GPT2 Spanish base model (Uncased) |
|
|
|
### Prerequisites |
|
|
|
transformers==4.19.2 |
|
|
|
### Model architecture |
|
|
|
This model uses GPT2 base setttings except vocabulary size. |
|
|
|
### Tokenizer |
|
|
|
Using BPE tokenizer with vocabulary size 50,000. |
|
|
|
### Training Data |
|
|
|
* [wiki40b/es](https://www.tensorflow.org/datasets/catalog/wiki40b#wiki40bes) (Spanish Wikipedia) |
|
* Subset of [CC-100/es](https://data.statmt.org/cc-100/) : Monolingual Datasets from Web Crawl Data |
|
|
|
### Usage |
|
|
|
```python |
|
from transformers import pipeline |
|
|
|
generator = pipeline('text-generation', model='ClassCat/gpt2-base-spanish') |
|
generator("Yo soy ", max_length=50, num_return_sequences=5) |
|
``` |
