app.py

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

"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""

# Cargar el modelo y el tokenizer
model_name = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Construir el prompt con el formato correcto
    prompt = f"<|system|>\n{system_message}</s>\n"

    for val in history:
        if val[0]:
            prompt += f"<|user|>\n{val[0]}</s>\n"
        if val[1]:
            prompt += f"<|assistant|>\n{val[1]}</s>\n"

    prompt += f"<|user|>\n{message}</s>\n<|assistant|>\n"

    # Tokenizar el prompt
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

    # Generar la respuesta
    outputs = model.generate(
        **inputs,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id
    )

    # Decodificar la respuesta
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)

    # Extraer solo la parte de la respuesta del asistente
    response = response.split("<|assistant|>\n")[-1].strip()

    yield response


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/gradio/chatinterface
"""
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(
            value="You are a friendly Chatbot. Always reply in the language in which the user is writing to you.", 
            label="System message"
        ),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)


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