Train Cost

[mo230761]

I really like your work! I would like to ask if a 32G V100 can be trained. Thansk！

[hrz2000]

Our experiments use 8 × 80GB NVIDIA A100 GPUs, with a total batch size of 512, with the image at 256 × 256 resolution following InstructPix2Pix. This configuration can also be trained on 40GB GPUs, if you only have 32G GPU, you can consider reducing the batch size slightly.

[mo230761]

Thank you very much！