app.py

import os
import subprocess

subprocess.check_call([os.sys.executable, "-m", "pip", "install", "-r", "requirements.txt"])

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

MODEL_ID = "le-llm/gemma-3-12b-it-reasoning"

# Load model & tokenizer
device = "cuda" if torch.cuda.is_available() else "cpu"
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    torch_dtype=torch.float16 if device == "cuda" else torch.float32
).to(device)

SYSTEM_PROMPT = "You are a friendly Chatbot."

def respond(
    message,
    history: list[dict[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Build conversation in chat template format
    messages = [{"role": "system", "content": system_message}] + history + [
        {"role": "user", "content": message}
    ]

    input_text = tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True  # ensures model knows it's assistant's turn
    )
    inputs = tokenizer(input_text, return_tensors="pt").to(device)

    output_ids = model.generate(
        **inputs,
        max_new_tokens=max_tokens,
        temperature=temperature,
        top_p=top_p,
        do_sample=True,
        eos_token_id=tokenizer.eos_token_id,  # stop at EOS
    )

    # Only return the newly generated assistant message
    response = tokenizer.decode(
        output_ids[0][inputs["input_ids"].shape[1]:],
        skip_special_tokens=True
    )
    return response

chatbot = gr.ChatInterface(
    respond,
    type="messages",
    additional_inputs=[
        gr.Textbox(value=SYSTEM_PROMPT, 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)",
        ),
    ],
)

chatbot.launch()