/chat_GPT_dall_e_implementation

A small implementation to consume the OpenAI API for Chat GPT and Dall-E image generator

Primary LanguageJupyter Notebook

Chat GPT and Dall-E implementation

This is a small implementation to consume Chat GPT and Dall-E image generator through the OpenAI API.

To use you need Python and the OpenAI library on your machine. In order tu install the OpenAI lib, run the following command:

pip install openai

Basic configuration

To use this tool you need an OpenAI API key. You can generate your key here. After that, put your key in the specified place in the code:

MY_KEY = "xxxxxxxxxxxxxxxxxxxxxx"

Be aware that the usage of this API may incur in billing. If you're creating your account, don't worry: New users receive sufficient credit to run hundreds of tests.

Usage

After setting the API key you just need to run one of the main files:

  • main.ipynb in your preferred notebook platform or
  • main.py in your IDE of choice or even in the terminal.

The tool has a menu that you can navigate to perform the actions:

  • 1 - Chat GPT
  • 2 - Dall-E

Each option has its own configurable values:

Chat GPT config

  • model: Model to be used in text generation. Eg: text-davinci-003, text-curie-001, etc. Default value -> text-davinci-003. Details of each one can be found here;
  • n: Number of responses to be generated. Default value -> 1;
  • max_tokens: Maximum number of tokens used for prompt and responses. Notice that punctuation and prompt also count. Max value vary according to the model you're using. Max value for text-davinci-003 is 4000, and for others is 2048. Default value -> 4000;
  • temperature: The temperature decimal level varies between 0 - generates a more conservative response - and 1 - which generates a more 'creative' response. Default value -> 0.7;
  • best_of: Number of responses to be displayed from the n generated answers you set before, Default value -> 1;

Dall-E config

  • size: Set the output image size. Must be one of 256x256 512x512 or 1024x1024 Default value -> 512x512;

Here you will find a in-depth explanation of those parameters

Notice that by running the tool from the .ipynb file, the generated image will be shown inside the notebook. If you run in the .py file will need to open the provided image URL in your browser manually.