@Operator public final class ResourceSparseApplyKerasMomentum extends PrimitiveOp
Set use_nesterov = True if you want to use Nesterov momentum.
That is for rows we have grad for, we update var and accum as follows:
accum = accum * momentum - lr * grad var += accum
Modifier and Type | Class and Description |
---|---|
static class |
ResourceSparseApplyKerasMomentum.Options
Optional attributes for
ResourceSparseApplyKerasMomentum |
operation
Modifier and Type | Method and Description |
---|---|
static <T,U extends Number> |
create(Scope scope,
Operand<?> var,
Operand<?> accum,
Operand<T> lr,
Operand<T> grad,
Operand<U> indices,
Operand<T> momentum,
ResourceSparseApplyKerasMomentum.Options... options)
Factory method to create a class wrapping a new ResourceSparseApplyKerasMomentum operation.
|
static ResourceSparseApplyKerasMomentum.Options |
useLocking(Boolean useLocking) |
static ResourceSparseApplyKerasMomentum.Options |
useNesterov(Boolean useNesterov) |
equals, hashCode, op, toString
public static <T,U extends Number> ResourceSparseApplyKerasMomentum create(Scope scope, Operand<?> var, Operand<?> accum, Operand<T> lr, Operand<T> grad, Operand<U> indices, Operand<T> momentum, ResourceSparseApplyKerasMomentum.Options... options)
scope
- current scopevar
- Should be from a Variable().accum
- Should be from a Variable().lr
- Learning rate. Must be a scalar.grad
- The gradient.indices
- A vector of indices into the first dimension of var and accum.momentum
- Momentum. Must be a scalar.options
- carries optional attributes valuespublic static ResourceSparseApplyKerasMomentum.Options useLocking(Boolean useLocking)
useLocking
- If `True`, updating of the var and accum tensors will be protected
by a lock; otherwise the behavior is undefined, but may exhibit less
contention.public static ResourceSparseApplyKerasMomentum.Options useNesterov(Boolean useNesterov)
useNesterov
- If `True`, the tensor passed to compute grad will be
var + momentum * accum, so in the end, the var you get is actually
var + momentum * accum.Copyright © 2022. All rights reserved.