search

openai / glide-text2im

GLIDE: a diffusion-based text-conditional image synthesis model
MIT License
3.54k stars 500 forks source link

Training Code #7

Closed AminRezaei0x443 closed 2 years ago

AminRezaei0x443 commented 2 years ago

Really Great Work! I wonder if you'll make training scripts available or not. If yes, are there any plans about when you'll make it available? Thanks

unixpickle commented 2 years ago

We currently do not plan to make the training scripts available, although you can base training off of the implementation in guided-diffusion. The main part that is changed is data loading, which you will need to setup yourself with your own text-image dataset. We currently don't plan to release our dataset or data loader for this project.

meatflavourdev commented 2 years ago

That's fine and your right as the intellectual property creators, but if you're not going to be open, don't call yourselves "Open."

INF800 commented 2 years ago

We currently do not plan to make the training scripts available, although you can base training off of the implementation in guided-diffusion. The main part that is changed is data loading, which you will need to setup yourself with your own text-image dataset. We currently don't plan to release our dataset or data loader for this project.

Hey @unixpickle,

Are the data loading strategies well defined in paper so that I might be able to reproduce it?

Best, Rakesh

alicedingyueming commented 1 year ago

We currently do not plan to make the training scripts available, although you can base training off of the implementation in guided-diffusion. The main part that is changed is data loading, which you will need to setup yourself with your own text-image dataset. We currently don't plan to release our dataset or data loader for this project.

May I ask for the training code for inpainting? Or can you provide me some hints for implementing a training script for inpainting?