randn_op

→TensorImplPtr

int randn_op(const int & shape, Dtype dt, Device device, Generator * gen)

Allocates a tensor filled with standard-normal samples $\mathcal{N}(0, 1)$ .

Convenience wrapper around normal_op with mean = 0 / std = 1.

Math

$x_i \sim \mathcal{N}(0, 1)$

See Also

normal_op — arbitrary mean / std. rand_op — uniform variant.

shapeconst Shape&

Shape of the output tensor.

dtDtype

Element dtype. Floating-point only.

deviceDevice

Memory domain.

genGenerator*= None

Bit-stream state; falls back to default_generator.

TensorImplPtr

Newly allocated leaf tensor (requires_grad = false).