import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Qwen/Qwen2.5-Coder-1.5B-Instruct"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    device_map="auto",
    torch_dtype="auto"
)

def chat_fn(message, history):
    inputs = tokenizer.apply_chat_template(
        history + [{"role": "user", "content": message}],
        return_tensors="pt"
    ).to(model.device)

    outputs = model.generate(inputs, max_new_tokens=350)
    reply = tokenizer.decode(outputs[0], skip_special_tokens=True)

    return reply

iface = gr.ChatInterface(fn=chat_fn, title="AI Coder Chatbot")
iface.launch()