facebookresearch/TransCoder

how to run translate.py without GPU?

abhishekgoldypathak opened this issue · 1 comments

Is it possible to run using only CPU? I have tried to run it by changing all instances of

device='cuda:0'

to device='cpu'

or setting CUDA_VISIBLE_DEVICES="" etc

but nevertheless not able to git rid of all the errors regardless of what I try. What's the easy way to run it on a CPU only machine?

Did you also change all the instances where the .cuda() is called on a tensor? For instance in the to_cuda(*args) method.