When the word "distributed" is used in the same sentence with "tensorflow" everyone is thinking for "training"
My question has to do with building tensorflow in a distributed environment. I want to build tensoflow from sources to create a custom version of mine; but it takes ages and if I decide to compile with CUDA support it takes even more
Is there any documentation on how to distributed build tensorflow?
thnx