streamlit_app.py

import streamlit as st
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
import torch

st.set_page_config(page_title="OpenChat 3.5 - HF Space", page_icon="🤖")
st.title("🤖 YahyaGPT")

@st.cache_resource
def load_model():
    tokenizer = AutoTokenizer.from_pretrained("openchat/openchat-3.5-1210")
    model = AutoModelForCausalLM.from_pretrained(
        "openchat/openchat-3.5-1210",
        torch_dtype=torch.float16,
        device_map="auto"
    )
    return tokenizer, model

tokenizer, model = load_model()

if "messages" not in st.session_state:
    st.session_state.messages = []

for msg in st.session_state.messages:
    st.chat_message(msg["role"]).write(msg["content"])


prompt = st.chat_input("Ask Yahya anything...")

if prompt:
    st.chat_message("user").write(prompt)
    st.session_state.messages.append({"role": "user", "content": prompt})

    full_prompt = "You are a helpful assistant.\n"
    for msg in st.session_state.messages:
        role = msg["role"]
        content = msg["content"]
        if role == "user":
            full_prompt += f"User: {content}\n"
        else:
            full_prompt += f"Assistant: {content}\n"
    full_prompt += "Assistant:"

    inputs = tokenizer(full_prompt, return_tensors="pt").to(model.device)

    output = model.generate(
        **inputs,
        max_new_tokens=300,
        do_sample=True,
        temperature=0.7,
        top_p=0.9,
        repetition_penalty=1.1,
        pad_token_id=tokenizer.eos_token_id
    )

    response = tokenizer.decode(output[0], skip_special_tokens=True)
    reply = response.split("Assistant:")[-1].strip()

    st.chat_message("assistant").write(reply)
    st.session_state.messages.append({"role": "assistant", "content": reply})