You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Nov 28, 2025. It is now read-only.
Repository navigation
This repository was archived by the owner on Nov 28, 2025. It is now read-only.
About spark-tensorflow-distributor---I have the same problem as you. When I use a spark cluster with two or more nodes, the algorithm needs the same time as if I have only one node. I don't understand why. Please, someone answer this doubt #176
I have the same problem as you. When I use a spark cluster with two or more nodes, the algorithm needs the same time as if I have only one node. I don't understand why. Please, someone answer this doubt
changed the title [-]I have the same problem as you. When I use a spark cluster with two or more nodes, the algorithm needs the same time as if I have only one node. I don't understand why. Please, someone answer this doubt[/-][+]About spark-tensorflow-distributor---I have the same problem as you. When I use a spark cluster with two or more nodes, the algorithm needs the same time as if I have only one node. I don't understand why. Please, someone answer this doubt[/+]on Dec 19, 2020
Activity
[-]I have the same problem as you. When I use a spark cluster with two or more nodes, the algorithm needs the same time as if I have only one node. I don't understand why. Please, someone answer this doubt[/-][+]About spark-tensorflow-distributor---I have the same problem as you. When I use a spark cluster with two or more nodes, the algorithm needs the same time as if I have only one node. I don't understand why. Please, someone answer this doubt[/+]