import os

os.system("pip install transformers==4.37.0")
os.system("pip install torch==2.0.1")
os.system("pip install accelerate")

import streamlit as st
from transformers import AutoModelForCausalLM, AutoTokenizer

# Set device
device = "cpu"

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen1.5-0.5B-Chat",
    torch_dtype="auto",
).to(device)
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen1.5-0.5B-Chat")

# Create a chatbot interface
st.title("Chatbot")
st.write("Ask me anything!")

# Initialize messages
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
]

# Display chat history
for message in messages:
    if message["role"] == "system":
        st.write(f"*System*: {message['content']}")
    elif message["role"] == "user":
        st.write(f"*You*: {message['content']}")
    elif message["role"] == "assistant":
        st.write(f"*Assistant*: {message['content']}")

# Get user input
user_input = st.text_input("Your message")
print("received!")

# Generate response
if user_input:
    messages.append({"role": "user", "content": user_input})
    print("good!")
    text = tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True
    )
    model_inputs = tokenizer([text], return_tensors="pt").to(device)
    print("good!")
    generated_ids = model.generate(
        model_inputs.input_ids,
        max_new_tokens=512
    )
    generated_ids = [
        output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
    ]
    print("good!")
    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
    print("good!")
    messages.append({"role": "assistant", "content": response})

    # Display response
    st.write(f"*Assistant*: {response}")