![]() |
Class RaggedTensorSpec
Type specification for a tf.RaggedTensor
.
Aliases:
__init__
__init__(
shape=None,
dtype=tf.dtypes.float32,
ragged_rank=None,
row_splits_dtype=tf.dtypes.int64
)
Constructs a type specification for a tf.RaggedTensor
.
Args:
shape
: The shape of the RaggedTensor, orNone
to allow any shape. If a shape is specified, then all ragged dimensions must have sizeNone
.dtype
:tf.DType
of values in the RaggedTensor.ragged_rank
: Python integer, the ragged rank of the RaggedTensor to be described. Defaults toshape.ndims - 1
.row_splits_dtype
:dtype
for the RaggedTensor'srow_splits
tensor. One oftf.int32
ortf.int64
.
Properties
value_type
The Python type for values that are compatible with this TypeSpec.
Methods
tf.RaggedTensorSpec.__eq__
__eq__(other)
Return self==value.
tf.RaggedTensorSpec.__ne__
__ne__(other)
Return self!=value.
tf.RaggedTensorSpec.from_value
@classmethod
from_value(
cls,
value
)
tf.RaggedTensorSpec.is_compatible_with
is_compatible_with(spec_or_value)
Returns true if spec_or_value
is compatible with this TypeSpec.
tf.RaggedTensorSpec.most_specific_compatible_type
most_specific_compatible_type(other)
Returns the most specific TypeSpec compatible with self
and other
.
Args:
other
: ATypeSpec
.
Raises:
ValueError
: If there is no TypeSpec that is compatible with bothself
andother
.