result_type

→DType

result_type(tensor: Tensor | Scalar, other: Tensor | Scalar)

source edit

Compute the dtype that a binary operation on the inputs would produce.

The dtype a + b (and every other binary op) computes in: the same rule, from the same place, as the operators themselves.

Parameters

tensorTensor | Scalar

First operand — a tensor or a Python number.

otherTensor | Scalar

Second operand. Same convention as tensor.

Returns

DType

The dtype that an arithmetic operation on tensor and other would produce.

Notes

Operands fall into three tiers: tensors with at least one dimension, 0-d tensors, and Python numbers. Each tier's dtypes meet on the promote_types lattice, and a lower tier can lift the result's category (bool $<$ int $<$ float $<$ complex) but never its width within a category:

\text{result\_type}(a, b) = \operatorname{combine}\bigl(\text{dim},\; \operatorname{combine}(\text{zero-dim},\; \text{number})\bigr)

A Python bool counts as bool, an int as int64, a float as the default dtype and a complex as its complex counterpart. Where the reference would answer complex32 (float16 beside a complex operand), Lucid answers complex64.

See also promote_types (operates directly on dtype objects).

Examples

>>> import lucid
>>> a = lucid.tensor([1, 2], dtype=lucid.int32)
>>> b = lucid.tensor([1., 2.], dtype=lucid.float32)
>>> lucid.result_type(a, b)
lucid.float32
>>> lucid.result_type(a, 2.5)
lucid.float32
>>> lucid.result_type(lucid.ones(2, dtype=lucid.float16), lucid.tensor(0.5))
lucid.float16

>>> import lucid >>> a = lucid.tensor([1, 2], dtype=lucid.int32) >>> b = lucid.tensor([1., 2.], dtype=lucid.float32) >>> lucid.result_type(a, b) lucid.float32 >>> lucid.result_type(a, 2.5) lucid.float32 >>> lucid.result_type(lucid.ones(2, dtype=lucid.float16), lucid.tensor(0.5)) lucid.float16