app.py

import streamlit as st

@st.cache_resource(show_spinner="Loading Model & Tokenizer")
def load_model():
    # This is cached and will not run again and again.
    from transformers import AutoTokenizer, AutoModelForCausalLM
    import torch
    from peft import PeftModel

    base_model = AutoModelForCausalLM.from_pretrained(
        "unsloth/Qwen2.5-0.5B", device_map="cpu", torch_dtype=torch.bfloat16)

    m = PeftModel.from_pretrained(base_model, "mosama/Qwen2.5-0.5B-Pretraining-ar-eng-urd-LoRA-Adapters")

    merged_model = m.merge_and_unload()

    tokenizer = AutoTokenizer.from_pretrained("mosama/Qwen2.5-0.5B-Pretrained-ar-end-urd-500")
    st.success('Model & Tokenizer Loaded Successfully!', icon="✅")
    return merged_model, tokenizer

st.title("Qwen2.5-0.5B Arabic, English & Urdu Continuous Pretrained")

model, tokenizer = load_model()

# Initialize chat history
if "messages" not in st.session_state:
    st.session_state.messages = []

for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.markdown(message["content"])

if not st.session_state.messages:
    with st.chat_message("assistant", avatar="assistant"):
        st.write("Hello 👋 I am an AI bot powered by Qwen 2.5 0.5B model.")
    st.session_state.messages.append({"role": "assistant", "content": "Hello 👋 I am an AI bot powered by Qwen 2.5 0.5B model."})

    st.session_state.state_chat_input = False

if prompt := st.chat_input("Say Something", key="input_1", disabled=st.session_state.state_chat_input):
    # Display user message in chat message container
    with st.chat_message("user"):
        st.markdown(prompt)
    # Add user message to chat history
    st.session_state.messages.append({"role": "user", "content": prompt})

if prompt or st.session_state.state_chat_input:
    if st.session_state.state_chat_input:
        with st.spinner(text="Generating response..."):
            model_inputs = tokenizer(st.session_state.messages[-1]['content'], return_tensors="pt").to(model.device)
            print(model_inputs)
            generated_ids = model.generate(
                **model_inputs,
                max_new_tokens=50,
                repetition_penalty=1.2,
                temperature=0.5,
                do_sample=True,
                top_p=0.9,
                top_k=20
            )
            print("Generated Response!")
            response = tokenizer.decode(generated_ids[0], skip_special_tokens=True)

        # Display assistant response in chat message container
        with st.chat_message("assistant"):
            st.markdown(response)
        # Add assistant response to chat history
        st.session_state.messages.append({"role": "assistant", "content": response})
        st.session_state.state_chat_input = False
        st.rerun()
    else:
        st.session_state.state_chat_input = True
        st.rerun()