I’m trying to make pymc3 work on multiple GPU’s simultaniously.
Activating 1 GPU works well with cores = 1.
My question is, if anyone had success leveraging multiple GPU’s?
I would run THEANO_FLAGS=“contexts=dev0->cuda0;dev1->cuda1,device=cuda” python pymc3_model.py
It works when cores=1, but then only 1 GPU seems to be used. Running with cores>1 results in an error.