app.py

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

MODEL_ID = "google/gemma-3-1b-it"

SYSTEM_PROMPT = """You are THMEXAAI, a helpful AI assistant.

Your name is THMEXAAI, powered by a SLM (Small Language Model).

You were pretrained on 140+ languages (including low-resource languages).

Out-of-the-box, you can communicate in 35+ languages and are optimized for instruction following.

Supported languages include (but are not limited to):
English, Spanish, French, German, Italian, Portuguese, Dutch, Swedish, Norwegian, Danish, Finnish, Polish, Czech, Slovak, Hungarian, Romanian, Bulgarian, Greek, Turkish, Arabic, Hebrew, Russian, Ukrainian, Hindi, Bengali, Urdu, Tamil, Telugu, Indonesian, Malay, Vietnamese, Thai, Chinese, Japanese and Korean.

Always:
- Be helpful
- Be accurate
- Follow user instructions
- Answer in the user's language whenever possible
- Explain clearly
- Be concise unless more detail is requested

You are THMEXAAI.
"""

print("Loading tokenizer...")
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)

print("Loading model...")
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    torch_dtype=torch.float32,
    device_map="cpu",
    low_cpu_mem_usage=True,
)

model.eval()

def chat(message, history):
    messages = [
        {
            "role": "system",
            "content": SYSTEM_PROMPT,
        }
    ]

    for user_msg, assistant_msg in history:
        messages.append(
            {
                "role": "user",
                "content": user_msg,
            }
        )
        messages.append(
            {
                "role": "assistant",
                "content": assistant_msg,
            }
        )

    messages.append(
        {
            "role": "user",
            "content": message,
        }
    )

    inputs = tokenizer.apply_chat_template(
        messages,
        tokenize=True,
        add_generation_prompt=True,
        return_tensors="pt",
    )

    with torch.no_grad():
        outputs = model.generate(
            inputs,
            max_new_tokens=512,
            do_sample=True,
            temperature=0.7,
            top_p=0.95,
            repetition_penalty=1.05,
            pad_token_id=tokenizer.eos_token_id,
            eos_token_id=tokenizer.eos_token_id,
        )

    generated = outputs[0][inputs.shape[-1]:]

    response = tokenizer.decode(
        generated,
        skip_special_tokens=True,
    )

    return response

demo = gr.ChatInterface(
    fn=chat,
    title="THMEXAAI",
    description="Powered by Gemma 3 1B IT",
    examples=[
        "Hello!",
        "Who are you?",
        "Explain machine learning",
        "Hola, ¿puedes hablar español?",
        "日本語で自己紹介してください"
    ],
)

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