search

cientgu / VQ-Diffusion

MIT License
439 stars 43 forks source link

About training time #3

Closed yzxing87 closed 2 years ago

yzxing87 commented 2 years ago

Thanks for releasing the codes of this awesome work! May I know the training cost of the VQ-Diffusion-B model? How long does the training take when using 8 V100 GPUs?

cientgu commented 2 years ago

It depends on the complexity of data. For example, on CUB200, it needs about 5days, on MSCOCO, it needs about 14days (longer training may slightly improve the performance). And the VQ-Diffusion-F model can achieve better results in the same computation cost.

yzxing87 commented 2 years ago

Thanks for your clarification!