Multi-GPUs problem

[BongBong87]

Does this model can train and test in Multi-GPUs? Thank you

[Peilun-Li]

The code is single-GPU only at that time. For the model, at training time we found batch size 1 would end up with the best performance (perhaps restricted by dataset size); at test time it could be paralleled/distributed for faster inference,

[BongBong87]

Thank you so much for quick reply. In case If I change the bath size to 4 or 8. Which part I need to modify for multi-GPUs. Thanks again.

[Peilun-Li]

You may have to restructure model.py to place generators/discriminators in different GPUs. https://www.tensorflow.org/guide/using_gpu .

[BongBong87]

Okay. Thank you so much for your sharing