View source on GitHub

Writes a scalar summary if possible.

Unlike tf.contrib.summary.generic this op may change the dtype depending on the writer, for both practical and efficiency concerns.

name An arbitrary name for this summary.
tensor A tf.Tensor Must be one of the following types: float32, float64, int32, int64, uint8, int16, int8, uint16, half, uint32, uint64.
family Optional, the summary's family.
step The int64 monotonic step variable, which defaults to tf.compat.v1.train.get_global_step.

The created tf.Operation or a tf.no_op if summary writing has not been enabled for this context.