Returns a cluster with do_eval option turned on/off. E.g.:: def FProp(...): with SetEval(mode=True): # Turns off dropout, noise, etc. y = self.foo.FProp(..., x) z = self.bar.FProp(..., y) # Returns to previous state (e.g., training). y = self.foo.FPr
(mode)
source not stored for this graph (policy: none)
nothing calls this directly
no test coverage detected
searching dependent graphs…