app.py

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer

# Carregar o tokenizer e o modelo
tokenizer = AutoTokenizer.from_pretrained("Locutusque/gpt2-xl-conversational")
model = AutoModelForCausalLM.from_pretrained("Locutusque/gpt2-xl-conversational")

# Função para gerar respostas
def generate_response(prompt, max_length=100, temperature=0.7, top_k=50, top_p=0.95):
    inputs = tokenizer(prompt, return_tensors="pt")
    outputs = model.generate(
        inputs.input_ids,
        max_length=max_length,
        temperature=temperature,
        top_k=top_k,
        top_p=top_p,
        num_return_sequences=1
    )
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return response

# Interface do Gradio
iface = gr.Interface(fn=generate_response, inputs="text", outputs="text", title="GPT-2 Conversational")

if __name__ == "__main__":
    iface.launch()