app.py

import gradio as gr
from huggingface_hub import login
from dotenv import load_dotenv
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, TextIteratorStreamer
from threading import Thread

MODEL = "m-a-p/OpenCodeInterpreter-DS-33B"

system_message = "You are a computer programmer that can translate python code to C++ in order to improve performance"

def user_prompt_for(python):
    return f"Rewrite this python code to C++. You must search for the maximum performance. \
    Format your response in Markdown. This is the Code: \
    \n\n\
    {python}"

def messages_for(python):
    return [
        {"role": "system", "content": system_message},
        {"role": "user", "content": user_prompt_for(python)}
    ]

tokenizer = AutoTokenizer.from_pretrained(MODEL)
model = AutoModelForCausalLM.from_pretrained(MODEL)
streamer = TextIteratorStreamer(tokenizer)

cplusplus = None
def translate(python):
    inputs = tokenizer(messages_for(python), return_tensors="pt")
    generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=20)
    thread = Thread(target=model.generate, kwargs=generation_kwargs)
    thread.start()
    cplusplus = ""
    for chunk in streamer:
        cplusplus += chunk
        yield cplusplus

demo = gr.Interface(fn=translate, inputs="code", outputs="markdown")
demo.launch()