tf.keras.backend.cast_to_floatx

tf.keras.backend.cast_to_floatx(x)

Defined in tensorflow/python/keras/_impl/keras/backend.py.

Cast a Numpy array to the default Keras float type.

Arguments:

• x: Numpy array.

Returns:

The same Numpy array, cast to its new type.

Example:

>>> from keras import backend as K
>>> K.floatx()
'float32'
>>> arr = numpy.array([1.0, 2.0], dtype='float64')
>>> arr.dtype
dtype('float64')
>>> new_arr = K.cast_to_floatx(arr)
>>> new_arr
array([ 1.,  2.], dtype=float32)
>>> new_arr.dtype
dtype('float32')