randint_op

→TensorImplPtr

int randint_op(const int & shape, int low, int high, Dtype dt, Device device, Generator * gen)

source edit

Python wrappers

pylucid.creation.randintfunction

Allocates an integer tensor filled with samples drawn uniformly from the integer interval $[\mathrm{low}, \mathrm{high})$ .

Math

\mathrm{high}-1\}$$

Parameters

shapeconst Shape&

Shape of the output tensor.

lowint64_t

Inclusive lower bound.

highint64_t

Exclusive upper bound; must be strictly greater than low.

dtDtype

Integer dtype (I32, I64, U8 …).

deviceDevice

Memory domain.

genGenerator*= None

Bit-stream state.

Returns

TensorImplPtr

Newly allocated leaf integer tensor (requires_grad = false).

Notes

The integer dtype must be wide enough to represent every value in $[\mathrm{low}, \mathrm{high})$ losslessly; the underlying Storage helper does not mask or wrap on overflow.