search

feizc / DiS

Scalable Diffusion Models with State Space Backbone
Other
145 stars 8 forks source link

How many GPUs used for the training #13

Open rambo-coder opened 6 months ago

rambo-coder commented 6 months ago

Thanks for releasing the code and your wonderful work. Could you please let know how many gpus you used for the training?

Best

FanqingM commented 6 months ago

The paper says it use bs = 1024 in imagenet 256, but I user global bs = 128, and 4 A100 80G to train DiS-H/2 , it OOD.... I wonder which GPU you use?