import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
import gradio as gr

# Load the DeepSeek model and tokenizer
model_name = "deepseek-ai/deepseek-llm-7b-base"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)
model.eval()

# Chat function
def chat(message, history=[]):
    history_text = "".join([f"User: {u}\nAssistant: {a}\n" for u, a in history])
    prompt = history_text + f"User: {message}\nAssistant:"
    
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(
        **inputs,
        max_new_tokens=200,
        temperature=0.7,
        do_sample=True,
        top_p=0.9,
        repetition_penalty=1.1,
        eos_token_id=tokenizer.eos_token_id,
        pad_token_id=tokenizer.eos_token_id
    )

    output_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
    reply = output_text.split("Assistant:")[-1].strip()
    history.append((message, reply))
    return reply, history

# Gradio UI
iface = gr.ChatInterface(
    fn=chat,
    title="DeepSeek Chatbot",
    description="Chatbot using DeepSeek 7B LLM",
    theme="default"
)

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