Warning: This API is deprecated and will be removed in a future version of TensorFlow after the replacement is stable.

TPUExecuteAndUpdateVariables

public final class TPUExecuteAndUpdateVariables

Op that executes a program with optional in-place variable updates.

It (optionally) reads device variables, loads and executes a TPU program on a TPU device, and then (optionally) in-place updates variables using the program outputs, as specified in attributes device_var_reads_indices (program input indices from directly reading variables) and device_var_updates_indices (program output indices used to update variables, -1 means no-update/read-only). Such program outputs are consumed by these variables will not appear in the op output. For the internal use of the distributed TPU compiler.

Public Methods

static TPUExecuteAndUpdateVariables
create ( Scope scope, Iterable< Operand <?>> args, Operand <String> key, List<Class<?>> Tresults, List<Long> deviceVarReadsIndices, List<Long> deviceVarUpdatesIndices)
Factory method to create a class wrapping a new TPUExecuteAndUpdateVariables operation.
Iterator< Operand <Object>>
List< Output <?>>

Inherited Methods

Public Methods

public static TPUExecuteAndUpdateVariables create ( Scope scope, Iterable< Operand <?>> args, Operand <String> key, List<Class<?>> Tresults, List<Long> deviceVarReadsIndices, List<Long> deviceVarUpdatesIndices)

Factory method to create a class wrapping a new TPUExecuteAndUpdateVariables operation.

Parameters
scope current scope
Returns
  • a new instance of TPUExecuteAndUpdateVariables

public Iterator< Operand <Object>> iterator ()

public List< Output <?>> results ()