normal_op

→TensorImplPtr

int normal_op(const int & shape, double mean, double std, Dtype dt, Device device, Generator * gen)

source edit

Python wrappers

Allocates a tensor filled with Gaussian samples $\mathcal{N}(\mathrm{mean}, \mathrm{std}^2)$ .

Math

$x_i \sim \mathcal{N}(\mathrm{mean}, \mathrm{std}^2)$

See Also

randn_op — shortcut for $\mathcal{N}(0, 1)$ .

Parameters

shapeconst Shape&

Shape of the output tensor.

meandouble

Location parameter

\mu

stddouble

Scale parameter

\sigma

. Must be non-negative; a value of 0 produces a constant tensor equal to mean.

dtDtype

Element dtype. Floating-point only.

deviceDevice

Memory domain.

genGenerator*= None

Bit-stream state.

Returns

TensorImplPtr

Newly allocated leaf tensor (requires_grad = false).

Raises

LucidError

If std < 0.