app.py

import streamlit as st
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

st.set_page_config(page_title="Chat with Qwen2.5-Omni-7B", layout="centered")

st.title("Chat with Qwen2.5-Omni-7B")

# Model name
model_name = "Qwen/Qwen2.5-Omni-7B"

# Prompt input
system_prompt = st.text_area("System Prompt", "You are a helpful assistant.", height=100)
user_input = st.text_input("Your Message", "")

# Temp & token sliders
temperature = st.slider("Temperature", 0.0, 1.0, 0.7)
max_tokens = st.slider("Max Tokens", 16, 1024, 256)

# Optional: Hugging Face token field (left empty for user)
hf_token = st.text_input("Hugging Face Token (optional)", type="password")

# Load model pipeline
@st.cache_resource
def load_pipeline():
    return pipeline(
        "text-generation",
        model=model_name,
        tokenizer=model_name,
        use_auth_token=hf_token if hf_token else None,
        device_map="auto"
    )

if user_input:
    pipe = load_pipeline()
    prompt = f"{system_prompt}\nUser: {user_input}\nAssistant:"
    response = pipe(prompt, temperature=temperature, max_new_tokens=max_tokens)[0]['generated_text']
    st.markdown("**Response:**")
    st.write(response.replace(prompt, ""))