|
|
|
|
|
import gradio as gr |
|
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM |
|
|
|
tokenizer = AutoTokenizer.from_pretrained("t5-small") |
|
model = AutoModelForSeq2SeqLM.from_pretrained("t5-small") |
|
|
|
def translate_text(text): |
|
inputs = tokenizer.encode("translate English to French: " + text, return_tensors="pt") |
|
outputs = model.generate(inputs, max_length=128, num_beams=4, early_stopping=True) |
|
translated_text = tokenizer.decode(outputs[0], skip_special_tokens=True) |
|
return translated_text |
|
|
|
output_1 = gr.Textbox(label="Speech to Text") |
|
output_2 = gr.Textbox(label="Speech Translation") |
|
|
|
generator = gr.Interface.load("huggingface/facebook/wav2vec2-base-960h", |
|
inputs="microphone", |
|
outputs=output_1, |
|
title="Speech-to-text", |
|
) |
|
|
|
translator = gr.Interface(fn=translate_text, |
|
inputs=output_1, |
|
outputs=output_2, |
|
title="English to French Translator", |
|
description="Translate English speech to French text using the T5-small model.", |
|
) |
|
|
|
gr.Series(generator, translator).launch() |
|
|
