tf.UnconnectedGradients

TensorFlow 1 version

View source on GitHub

Controls how gradient computation behaves when y does not depend on x.

View aliases

Compat aliases for migration

See Migration guide for more details.

tf.compat.v1.UnconnectedGradients

The gradient of y with respect to x can be zero in two different ways: there could be no differentiable path in the graph connecting x to y (and so we can statically prove that the gradient is zero) or it could be that runtime values of tensors in a particular execution lead to a gradient of zero (say, if a relu unit happens to not be activated). To allow you to distinguish between these two cases you can choose what value gets returned for the gradient when there is no path in the graph from x to y:

NONE: Indicates that [None] will be returned if there is no path from x to y
ZERO: Indicates that a zero tensor will be returned in the shape of x.

Class Variables
NONE	`tf.UnconnectedGradients`
ZERO	`tf.UnconnectedGradients`

TensorFlow

tf

tf.audio

tf.autograph

tf.bitwise

tf.compat

tf.config

tf.data

tf.debugging

tf.distribute

tf.dtypes

tf.errors

tf.estimator

tf.experimental

tf.feature_column

tf.graph_util

tf.image

tf.initializers

tf.io

tf.keras

tf.linalg

tf.lite

tf.lookup

tf.losses

tf.math

tf.metrics

tf.nest

tf.nn

tf.optimizers

tf.quantization

tf.queue

tf.ragged

tf.random

tf.raw_ops

tf.saved_model

tf.sets

tf.signal

tf.sparse

tf.strings

tf.summary

tf.sysconfig

tf.test

tf.tpu

tf.train

tf.version

tf.xla

tf / UnconnectedGradients

View aliases

Class Variables