copy_to_numpy_tensor

def copy_to_numpy_tensor[dtype: DType, origin: ImmOrigin, *shape_types: CoordLike](data: Span[Scalar[dtype], origin], shape: Coord[shape_types]) -> PythonObject

Builds an N-D NumPy array from a Mojo Span of scalars and a shape.

The span supplies the elements in C order (last axis varying fastest) and shape supplies the extents, so shape.product() must equal len(data). The data is copied into a new, independent NumPy array, so the result remains valid after data is later mutated or freed.

Example:

from std.python.numpy import copy_to_numpy_tensor
from std.utils.coord import Coord, Idx

var values = List[Float64](capacity=6)
for i in range(6):
    values.append(Float64(i))

var arr = copy_to_numpy_tensor(values, Coord(Idx[2], Idx[3]))  # 2x3 array

Dimensions may be compile-time (Idx[N]) or runtime (Int) in any mix.

Constraints:

dtype must be one of the fixed-width numeric dtypes supported by NumPy: int8-int64, uint8-uint64, float16, float32, or float64. shape must be flat (no nested Coord) and must not contain All.

Parameters:

• ​dtype (DType): The element dtype of the span (inferred).
• ​origin (ImmOrigin): The origin of the span (inferred).
• ​*shape_types (CoordLike): The per-dimension types of shape (inferred).

Args:

• ​data (Span[Scalar[dtype], origin]): The scalars to copy into a NumPy array, in C order.
• ​shape (Coord[shape_types]): The extents of the result, one element per axis.

Returns:

PythonObject: A NumPy ndarray of dtype dtype and shape shape.

Raises:

If shape.product() does not equal len(data), if NumPy is unavailable, or if the underlying NumPy calls fail.