How To Create Natural Language Semantic Search For Arbitrary Objects With Deep Learning
Hamel Husain

This is a great article. When trying to reproduce the results, for example, from ‘2 — Train Function Summarizer With Keras + TF’, it looks like the TF backend is not scripted to leverage all the CPUs available in the computer. For example, the fit ETA in this module is 7 hr per epoch no matter if I use my Mac laptop or an Ubuntu server with 32 CPUs. What’d be the right place to tweak TF to run faster if possible? Thanks.