| import gradio as gr | |
| from transformers import AutoModelForSeq2SeqLM, AutoTokenizer | |
| def translate(text): | |
| model_name = 'hackathon-pln-es/t5-small-finetuned-spanish-to-quechua' | |
| model = AutoModelForSeq2SeqLM.from_pretrained(model_name) | |
| tokenizer = AutoTokenizer.from_pretrained(model_name) | |
| input = tokenizer(text, return_tensors="pt") | |
| output = model.generate(input["input_ids"], max_length=40, num_beams=4, early_stopping=True) | |
| return tokenizer.decode(output[0], skip_special_tokens=True) | |
| title = "Spanish to Quechua translation 🦙" | |
| inputs = gr.inputs.Textbox(lines=1, label="Text in Spanish") | |
| outputs = [gr.outputs.Textbox(label="Translated text in Quechua")] | |
| description = "Here use the [t5-small-finetuned-spanish-to-quechua-model](https://huggingface.co/hackathon-pln-es/t5-small-finetuned-spanish-to-quechua) that was trained with [spanish-to-quechua dataset](https://huggingface.co/datasets/hackathon-pln-es/spanish-to-quechua)." | |
| article = ''' | |
| ## Challenges | |
| - Create a dataset, as there are different variants of Quechua. | |
| - Training of the model to optimize results using the least amount of computational resources. | |
| ## Team members | |
| - [Sara Benel](https://huggingface.co/sbenel) | |
| - [Jose Vílchez](https://huggingface.co/JCarlos) | |
| ''' | |
| examples=[ | |
| 'Dios ama a los hombres', | |
| 'A pesar de todo, soy feliz', | |
| '¿Qué harán allí?', | |
| 'Debes aprender a respetar', | |
| ] | |
| iface = gr.Interface(fn=translate, inputs=inputs, outputs=outputs, theme="grass", css="styles.css", examples=examples, title=title, description=description, article=article) | |
| iface.launch(enable_queue=True) |