Training Requirement for GPU

[黄梓楷]

Excellent work and thanks for sharing your code.
I'm a novice at diffusion models, and I'm concerned about theGPU resource this kind of models need.
I wonder can all the training be done on a single RTX3090.

[Jooyoung Choi]

Hello,
We trained our unconditional models with a single GPU with 48GB VRAM. It took about 1~2 weeks for FFHQ and AFHQ dataset. However, you might need more resources to train on larger datasets (ImageNet, LSUN). Diffusion Models Beat GANs on Image Synthesis spend almost 1000 GPU days to train on those datasets.

[黄梓楷]

Thank you very much for the quick reply.