4 ms·
x2 Gunicorn workers, MKL mapped to half physical cores for each... for some reason the model (tensorflow) performs better on half vs 1 worker
by ramoz 5y 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