org.platanios.tensorflow.api.ops.training.distribute.strategies
Distribution strategy.
ID of the tower that is being defined, which is a number in [0, numTowers - 1]
.
Returns the device this tower is to be executed on, as a string.
Distribution strategy.
Distribution strategy.
ID of the tower that is being defined, which is a number in [0, numTowers - 1]
.
Represents an in-tower context (as opposed to a cross-tower context).
This context is only present during a
forEachTower()
call (except during amergeRun()
call), and in such a scope it will be implicitly available.Distribution strategy.
ID of the tower that is being defined, which is a number in
[0, numTowers - 1]
.