Question about the training device on text2image

CompVis / latent-diffusion

High-Resolution Image Synthesis with Latent Diffusion Models

MIT License

11.48k stars 1.5k forks source link

Question about the training device on text2image #105

Open zhaowt61 opened 2 years ago

zhaowt61 commented 2 years ago

Thanks for the excellent work! I wonder about the training device on text2image. The paper says it is trained on a single A100, but it seems the settings in table 15 should take more than 640GB of memory. Looking forward to your reply!

zhangqizky commented 2 years ago

真的很奇怪，按照作者的配置，训LSUN churche，根本无法训出来，把学习率减小才可以