TensorFlow 1 version

Converts each entry in the given tensor to strings.

    input, precision=-1, scientific=False, shortest=False, width=-1, fill='',

Supports many numeric types and boolean.

For Unicode, see the tutorial.


tf.strings.as_string([3, 2]) 
<tf.Tensor: shape=(2,), dtype=string, numpy=array([b'3', b'2'], dtype=object)> 
tf.strings.as_string([3.1415926, 2.71828], precision=2).numpy() 
array([b'3.14', b'2.72'], dtype=object) 


  • input: A Tensor. Must be one of the following types: int8, int16, int32, int64, complex64, complex128, float32, float64, bool.
  • precision: An optional int. Defaults to -1. The post-decimal precision to use for floating point numbers. Only used if precision > -1.
  • scientific: An optional bool. Defaults to False. Use scientific notation for floating point numbers.
  • shortest: An optional bool. Defaults to False. Use shortest representation (either scientific or standard) for floating point numbers.
  • width: An optional int. Defaults to -1. Pad pre-decimal numbers to this width. Applies to both floating point and integer numbers. Only used if width > -1.
  • fill: An optional string. Defaults to "". The value to pad if width > -1. If empty, pads with spaces. Another typical value is '0'. String cannot be longer than 1 character.
  • name: A name for the operation (optional).


A Tensor of type string.