| | --- |
| | language: es |
| | license: cc-by-sa-4.0 |
| | datasets: |
| | - wikipedia |
| | - cc100 |
| | widget: |
| | - text: "¿Hablas español?" |
| | - text: "Es clima es" |
| | - text: "Las negociaciones están paradas, pero" |
| | --- |
| | |
| | ## 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) |
| | ``` |
