Cache tensorflow build dependencies, to accelerate repeated tf configures, or run on a plane - View it on GitHub