app.py

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


model_name = "INSAIT-Institute/MamayLM-Gemma-2-9B-IT-v0.1"  
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")

@spaces.GPU
def respond(message, chat_history, system_message, max_new_tokens, temperature, top_p):

    prompt = f"{system_message.strip()}\n"
    for user, bot in chat_history:
        prompt += f"User: {user}\nAssistant: {bot}\n"
    prompt += f"User: {message}\nAssistant:"

    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    output = model.generate(
        **inputs,
        max_new_tokens=int(max_new_tokens),
        pad_token_id=tokenizer.eos_token_id,
        do_sample=True,
        temperature=float(temperature),
        top_p=float(top_p),
        eos_token_id=tokenizer.eos_token_id,
    )
    decoded = tokenizer.decode(output[0], skip_special_tokens=True)

    response = decoded.split("Assistant:")[-1].strip().split("User:")[0].strip()
    return response

def clear_fn():
    return None

chat = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(value="Ти мовна модель, яка добре володіє українською.", 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)",
        ),
    ],
    examples=[
        ["Привіт!"],
        ["Хто такий Пес Патрон?"],
    ],
    title="💬 Chat with MamayLM",
    description="A multi-turn chat interface for MamayLM-v0.1-9B with configurable parameters.",
    theme="soft"
)

chat.launch()