import gradio as gr
from huggingface_hub import InferenceClient
from optimum.intel import OVModelForCausalLM
from transformers import AutoTokenizer, pipeline

# 載入模型和標記器
model_id = "hsuwill000/Qwen2.5-3B-Instruct-openvino"
model = OVModelForCausalLM.from_pretrained(model_id, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_id)

# 建立生成管道
#pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

def respond(prompt , history):
    # 將當前訊息與歷史訊息合併
    #input_text = message if not history else history[-1]["content"] + " " + message
    #input_text = message+",(450字內回覆)"
    
    messages = [
    {"role": "system", "content": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
    {"role": "user", "content": prompt }
    ]
    text = tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True
    )
    model_inputs = tokenizer([text], return_tensors="pt").to(model.device)    
    generated_ids = model.generate(
        **model_inputs,
        max_new_tokens=512
    )    
    
    # 獲取模型的回應
    #response = pipe(input_text, max_length=512, truncation=True, num_return_sequences=1)
    #reply = response[0]['generated_text']
    generated_ids = [
        output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
    ]
    response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]    
    
    # 返回新的消息格式
    print(f"Messages: {messages}")
    print(f"Reply: {response}")
    return response
    
# 設定 Gradio 的聊天界面
demo = gr.ChatInterface(fn=respond, title="Qwen2.5-0.5B-Instruct-openvino-4bit", description="Qwen2.5-0.5B-Instruct-openvino-4bit", type='messages')

if __name__ == "__main__":
    demo.launch()