Wireless Wingman

The Wireless Wingman is designed to take a simple prompt and to expand on that. With dataset scraped from The Algorithm repository, our model does not only trained with pre-written algorithm but also mimic the coding style written by professional programmers. The pre-trained GPT-2 variants are offered by the awesome 🤗 transformers library and tenserflow.

Features

• Generate code written in Python, Java, JavaScript, C, C++ and C#

Files and Directories:

The architecture of files and directories are as follows:

• app/
  • model/
    • EMITTED
  • static/
    • css/
    • img/
    • js/
    • favicon.ico
  • templates/
    • C-Popup.html
    • Cpp-Popup.html
    • CS-Popup.html
    • index.html
    • Java-Popip.html
    • JS-Popup.html
    • Python-Popup.html
    • writer_home.html
    • write-your-code-with-AI.html
  • main.py
  • popup_main.py
  • requirements.txt
  • utils.py
• .gitignore
• Dockerfile
• Readme.md
• config.py
• entrypoint.sh
• host_config
• nginx_host

Quick Start to Start Local Testing

$cd app
$python -m main

Example Code Generation Process

pip install aitextgen pip install transformer

# import stuff for our web server
from flask import Flask, request, redirect, url_for, render_template, session
from utils import get_base_url
# import stuff for our models
from aitextgen import aitextgen
from transformers import PreTrainedTokenizerFast, GPT2DoubleHeadsModel, GPT2TokenizerFast, GPT2Tokenizer


def load_model(model_path):
    model = GPT2DoubleHeadsModel.from_pretrained(model_path)
    return model


def load_tokenizer(tokenizer_path):
    tokenizer = GPT2Tokenizer.from_pretrained(tokenizer_path)
    return tokenizer

#For generating language from any existing model
def GenAI_JS(): 

    model = load_model("model/js")
    tokenizer = load_tokenizer("model/js")
    if prompt is not None:
        generated = model.generate(
            ids,
            do_sample=True,
            max_length=150,
            pad_token_id=model.config.eos_token_id,
            top_k=50,
            top_p=0.95,
            return_as_list=True
        )
    generated = tokenizer.decode(generated[0], skip_special_tokens=True)

Generative examples

-------------Example 1--------------------------------------
Context code: void Dijkstra
Generated: 

void Dijkstra(double B, double B2, double B2,
                              int limit1 = 10) {
        this->m = B[0].size();
        this->prev = B[0].top().second;
    }
}

int main() {
    std::srand(std::time(nullptr));
    test();
    return 0;
}

#include <cmath>
#include <iostream>

--------------Example 2-------------------------------------
```C
Context code: int bubbleSort(arr)
Generated:
int bubbleSort(arr)
{
    for (int i = 0; i < array->size; i++)
    {
        assert(array->array[i] <= array->array[i - 1]);
    }

    free(arr);
}


int main()
{
    test();
    return 0;
}