@Operator(group="train") public final class ConditionalAccumulator extends PrimitiveOp implements Operand<String>
The accumulator accepts gradients marked with local_step greater or equal to the most recent global_step known to the accumulator. The average can be extracted from the accumulator, provided sufficient gradients have been accumulated. Extracting the average automatically resets the aggregate to 0, and increments the global_step recorded by the accumulator.
Modifier and Type | Class and Description |
---|---|
static class |
ConditionalAccumulator.Options
Optional attributes for
ConditionalAccumulator |
operation
Modifier and Type | Method and Description |
---|---|
Output<String> |
asOutput()
Returns the symbolic handle of a tensor.
|
static ConditionalAccumulator.Options |
container(String container) |
static <T> ConditionalAccumulator |
create(Scope scope,
Class<T> dtype,
Shape shape,
ConditionalAccumulator.Options... options)
Factory method to create a class wrapping a new ConditionalAccumulator operation.
|
Output<String> |
handle()
The handle to the accumulator.
|
static ConditionalAccumulator.Options |
reductionType(String reductionType) |
static ConditionalAccumulator.Options |
sharedName(String sharedName) |
equals, hashCode, op, toString
public static <T> ConditionalAccumulator create(Scope scope, Class<T> dtype, Shape shape, ConditionalAccumulator.Options... options)
scope
- current scopedtype
- The type of the value being accumulated.shape
- The shape of the values, can be [], in which case shape is unknown.options
- carries optional attributes valuespublic static ConditionalAccumulator.Options container(String container)
container
- If non-empty, this accumulator is placed in the given container.
Otherwise, a default container is used.public static ConditionalAccumulator.Options sharedName(String sharedName)
sharedName
- If non-empty, this accumulator will be shared under the
given name across multiple sessions.public static ConditionalAccumulator.Options reductionType(String reductionType)
reductionType
- public Output<String> asOutput()
Operand
Inputs to TensorFlow operations are outputs of another TensorFlow operation. This method is used to obtain a symbolic handle that represents the computation of the input.
asOutput
in interface Operand<String>
OperationBuilder.addInput(Output)
Copyright © 2022. All rights reserved.