View source on GitHub |
Write a scalar summary.
tf.summary.scalar(
name, data, step=None, description=None
)
See also tf.summary.image
, tf.summary.histogram
, tf.summary.SummaryWriter
.
Writes simple numeric values for later analysis in TensorBoard. Writes go to
the current default summary writer. Each summary point is associated with an
integral step
value. This enables the incremental logging of time series
data. A common usage of this API is to log loss during training to produce
a loss curve.
For example:
test_summary_writer = tf.summary.create_file_writer('test/logdir')
with test_summary_writer.as_default():
tf.summary.scalar('loss', 0.345, step=1)
tf.summary.scalar('loss', 0.234, step=2)
tf.summary.scalar('loss', 0.123, step=3)
Multiple independent time series may be logged by giving each series a unique
name
value.
See Get started with TensorBoard
for more examples of effective usage of tf.summary.scalar
.
In general, this API expects that data points are logged iwth a monotonically increasing step value. Duplicate points for a single step or points logged out of order by step are not guaranteed to display as desired in TensorBoard.
Returns | |
---|---|
True on success, or false if no summary was written because no default summary writer was available. |
Raises | |
---|---|
ValueError
|
if a default writer exists, but no step was provided and
tf.summary.experimental.get_step() is None.
|