Module: tf.dtypes

Public API for tf.dtypes namespace.

Classes

class DType: Represents the type of the elements in a Tensor.

Functions

as_dtype(...): Converts the given type_value to a DType.

cast(...): Casts a tensor to a new type.

complex(...): Converts two real numbers to a complex number.

saturate_cast(...): Performs a safe saturating cast of value to dtype.

QUANTIZED_DTYPES 

{
 tf.qint16,
 tf.qint16_ref,
 tf.qint32,
 tf.qint32_ref,
 tf.qint8,
 tf.qint8_ref,
 tf.quint16,
 tf.quint16_ref,
 tf.quint8,
 tf.quint8_ref
}

bfloat16 Instance of tf.dtypes.DType

16-bit bfloat (brain floating point).

bool Instance of tf.dtypes.DType

Boolean.

complex128 Instance of tf.dtypes.DType

128-bit complex.

complex64 Instance of tf.dtypes.DType

64-bit complex.

double Instance of tf.dtypes.DType

64-bit (double precision) floating-point.

float16 Instance of tf.dtypes.DType

16-bit (half precision) floating-point.

float32 Instance of tf.dtypes.DType

32-bit (single precision) floating-point.

float64 Instance of tf.dtypes.DType

64-bit (double precision) floating-point.

half Instance of tf.dtypes.DType

16-bit (half precision) floating-point.

int16 Instance of tf.dtypes.DType

Signed 16-bit integer.

int32 Instance of tf.dtypes.DType

Signed 32-bit integer.

int64 Instance of tf.dtypes.DType

Signed 64-bit integer.

int8 Instance of tf.dtypes.DType

Signed 8-bit integer.

qint16 Instance of tf.dtypes.DType

Signed quantized 16-bit integer.

qint32 Instance of tf.dtypes.DType

signed quantized 32-bit integer.

qint8 Instance of tf.dtypes.DType

Signed quantized 8-bit integer.

quint16 Instance of tf.dtypes.DType

Unsigned quantized 16-bit integer.

quint8 Instance of tf.dtypes.DType

Unsigned quantized 8-bit integer.

resource Instance of tf.dtypes.DType

Handle to a mutable, dynamically allocated resource.

string Instance of tf.dtypes.DType

Variable-length string, represented as byte array.

uint16 Instance of tf.dtypes.DType

Unsigned 32-bit (dword) integer.

uint32 Instance of tf.dtypes.DType
uint64 Instance of tf.dtypes.DType

Unsigned 64-bit (qword) integer.

uint8 Instance of tf.dtypes.DType

Unsigned 8-bit (byte) integer.

variant Instance of tf.dtypes.DType

Data of arbitrary type (known at runtime).