View source on GitHub |
Like tf.math.reduce_logsumexp.
tfp.experimental.math.reduce_logsumexp(
a, axis=None, keepdims=False, name='reduce_logsumexp'
)
This has no functional difference from the regular version, except that it's implemented inline here, allowing monkey-patching of the special functions it uses (e.g. exp).
Args | |
---|---|
a
|
A tensor. |
axis
|
Dimensions to reduce. If None , reduces all dimensions.
|
keepdims
|
If True , retains the reduced dimensions with length 1.
|
name
|
Name for the op. |
Returns | |
---|---|
y
|
The reduced tensor. |