search

WUSTL-CSPL / RIATIG

MIT License
20 stars 2 forks source link

How to utilize a limited api to access a mount of images for experiments? #3

Open datar001 opened 7 months ago

datar001 commented 7 months ago

Hi, this is a good work! But I have a problem in the experiment that needs to evaluate the performance of DALL E2 by generating many images. How did you solve the problem? by the official API? Thanks!

datar001 commented 7 months ago

And, How to generate images by Imagen model. I don't find the released pre-trained model for usage in both github and huggingface. Thanks !

MasterEndless commented 7 months ago

And, How to generate images by Imagen model. I don't find the released pre-trained model for usage in both github and huggingface. Thanks !

Hi, Thanks for your interest! You can find a pre-trained Imagen model here: https://github.com/cene555/Imagen-pytorch/blob/main/notebooks/Imagen_pytorch_inference_new.ipynb

MasterEndless commented 7 months ago

For DALLE2, we first train adversarial examples on LAION’s pre-trained model, then access the APIs to check if it was transferred successfully, you can find more details in "Target models" in the Experiments section.