Comment by ramoz

4 years ago

x2 Gunicorn workers, MKL mapped to half physical cores for each... for some reason the model (tensorflow) performs better on half vs 1 worker

0 comments