import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
import gradio as gr

MODEL_NAME = "manycore-research/SpatialLM-Llama-1B"

# Carrega tokenizer e modelo
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_NAME,
    torch_dtype=torch.float16,
    device_map="auto"  # Usa GPU se disponível
)
model.eval()

# Função de geração
def generate_response(prompt, temperature=0.7, top_p=0.95, max_new_tokens=200):
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            do_sample=True,
            temperature=temperature,
            top_p=top_p,
            max_new_tokens=max_new_tokens,
            pad_token_id=tokenizer.eos_token_id
        )
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return response

# Interface Gradio
interface = gr.Interface(
    fn=generate_response,
    inputs=[
        gr.Textbox(label="Prompt", placeholder="Digite algo como 'Luan invadiu a base da Hegemonia...'"),
        gr.Slider(minimum=0.1, maximum=1.5, value=0.7, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, label="Top-p"),
        gr.Slider(minimum=10, maximum=512, value=200, label="Max Tokens")
    ],
    outputs=gr.Textbox(label="Resposta do Modelo"),
    title="SpatialLM - Llama 1B",
    description="Modelo SpatialLM LLaMA 1B rodando com GPU no Hugging Face Spaces. Ideal pra geração de texto contextualizado e linguagem espacial. Use prompts criativos."
)

interface.launch()