inference
inference copied to clipboard
Why not support multi-gpu?
I checked the reference implementation and found that the code didn't support multi-gpu.
So why not support multi-gpu?