Closed duduscript closed 6 years ago
Hi,
I train transformer model in 2 GPUs and I translate WMT data in one GPU.
I find the computation is 100% in CPU rather than GPU, I need to compare between marian version and my old tensorflow version in one GPU.
How to run it in one GPU while I trained the model in 2 GPUs?
It seems that works when I change the GPU setting both in command params and yml file
Hi,
I train transformer model in 2 GPUs and I translate WMT data in one GPU.
I find the computation is 100% in CPU rather than GPU, I need to compare between marian version and my old tensorflow version in one GPU.
How to run it in one GPU while I trained the model in 2 GPUs?