Which GPU was used for training?

[leksshil]

From read me, I realized that V100. Do I understand correctly that Nvidia Tesla V100 Pcie 32Gb Graphics Accelerator PN: 699-2g500-0202-400?

[alphacoder01]

The authors have stated in the paper:

We train our image editing model for 10,000 steps on 8× 40GB NVIDIA A100 GPUs over 25.5 hours

[leksshil]

Can you tell me if the tesla V100 pcie 32Gb is suitable for training?

[andreemic]

You should be fine with a batch size of 8-16. If you run out of memory decrease it and up the "accumulate_grad_batches" in configs/train.yaml

[leksshil]

Thanks.