cuda.core.TensorMapDescriptorOptions

class cuda.core.TensorMapDescriptorOptions(

box_dim: tuple[int, ...],

element_strides: tuple[int, ...] | None = None,

data_type: object = None,

interleave: TensorMapInterleave = TensorMapInterleave.NONE,

swizzle: TensorMapSwizzle = TensorMapSwizzle.NONE,

l2_promotion: TensorMapL2Promotion = TensorMapL2Promotion.NONE,

oob_fill: TensorMapOOBFill = TensorMapOOBFill.NONE,

)

Options for cuda.core.StridedMemoryView.as_tensor_map().

box_dim

Tile size for each tensor dimension, expressed in elements.

Type:

tuple[int, …]

element_strides

Per-dimension element traversal strides.

Type:

tuple[int, …], optional

data_type

Explicit dtype override. Prefer NumPy or ml_dtypes dtype objects; TensorMapDataType remains accepted for compatibility.

Type:

object, optional

interleave

Interleave layout. Default NONE.

Type:

TensorMapInterleave, optional

swizzle

Swizzle mode. Default NONE.

Type:

TensorMapSwizzle, optional

l2_promotion

L2 promotion mode. Default NONE.

Type:

TensorMapL2Promotion, optional

oob_fill

Out-of-bounds fill mode. Default NONE.

Type:

TensorMapOOBFill, optional