https://github.com/rom1504/dalle-service

Dalle service
https://github.com/rom1504/dalle-service

Last synced: about 2 months ago
JSON representation

Dalle service

• Host: GitHub
• URL: https://github.com/rom1504/dalle-service
• Owner: rom1504
• License: mit
• Created: 2021-06-06T21:56:40.000Z (over 3 years ago)
• Default Branch: master
• Last Pushed: 2021-11-27T14:29:02.000Z (almost 3 years ago)
• Last Synced: 2024-06-11T10:44:54.166Z (3 months ago)
• Language: JavaScript
• Homepage: https://rom1504.github.io/dalle-service/
• Size: 395 KB
• Stars: 50
• Watchers: 3
• Forks: 17
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome - rom1504/dalle-service - Dalle service (JavaScript)

README

        
# Dalle-service

This is a simple front + back making it easy to use [DALL-E models](https://github.com/lucidrains/DALLE-pytorch)

![image](https://user-images.githubusercontent.com/2346494/120942358-affaca80-c728-11eb-93c0-084e1c27435d.png)

Another option is to use [the gradio demo](https://colab.research.google.com/drive/1xFnY5kn9SadI_Jph37uxT1fJIMy_AdFK)

## Running the back on google colab

If you want to run the backend on google colab, you can run [this notebook](https://colab.research.google.com/github/rom1504/dalle-service/blob/master/dalle_back.ipynb)

You can then use the gh pages front with an url such as https://rom1504.github.io/dalle-service?back=https://XXXX.loca.lt

## Running the back myself

First follow [back](back) (you may choose to use https://ngrok.com/ to expose your locally running backend)

Then put your backend url in https://rom1504.github.io/dalle-service/

You can share an url such as https://rom1504.github.io/dalle-service?back=https://yourbackend.com

## Runnning the back and the front

If you want to run everything yourself, you can go to [back](back) then to [front](front)

## What Dalle models can I use ?

You can either train your model yourself with [DALL-E](https://github.com/lucidrains/DALLE-pytorch)

Or use a pretrained one from https://github.com/robvanvolt/DALLE-models/tree/main/models/taming_transformer 

* [laion subset cars](https://huggingface.co/spaces/rom1504/laion_car_subset/resolve/main/weird_car_model_continue.pt) a model trained on 10k cars extracted from [laion 400m](https://rom1504.github.io/clip-retrieval/)