class

CeilBackward

CeilBackward

source edit

Python wrappers

pylucid.ops.ceilfunction

Autograd node for element-wise ceiling: $y = \lceil x \rceil$ .

Piecewise-constant; derivative is zero almost everywhere and undefined at integer boundaries. kHasGradient = false.

Math

$y_i = \lceil x_i \rceil$ (smallest integer not less than $x_i$ ).

Attributes

kSavesInputbool

false.

kHasGradientbool

false.

schema_v1OpSchema

Op name "ceil" with AmpPolicy::KeepInput.

Notes

Integer dtypes are a no-op. Negative values round toward zero (e.g. ceil(-1.2) = -1).

Static methods

static

dispatch

int dispatch(backend::IBackend & be, const int & a, const int & s, Dtype dt)

source edit

Forward — calls IBackend::ceil to compute $y = \lceil x \rceil$ .

Methods

grad_formula

→Storage

int grad_formula(const int & g)

source edit

Zero-gradient sentinel; never invoked since kHasGradient = false.

Parameters

gconst Storage&

Incoming gradient (ignored).

Returns

Storage

Empty CpuStorage placeholder.

class

CeilBackward

CeilBackward

source edit

Python wrappers

pylucid.ops.ceilfunction

Autograd node for element-wise ceiling: $y = \lceil x \rceil$ .

Piecewise-constant; derivative is zero almost everywhere and undefined at integer boundaries. kHasGradient = false.

Math

$y_i = \lceil x_i \rceil$ (smallest integer not less than $x_i$ ).

Attributes

kSavesInputbool

false.

kHasGradientbool

false.

schema_v1OpSchema

Op name "ceil" with AmpPolicy::KeepInput.

Notes

Integer dtypes are a no-op. Negative values round toward zero (e.g. ceil(-1.2) = -1).

Static methods

static

dispatch

int dispatch(backend::IBackend & be, const int & a, const int & s, Dtype dt)

source edit

Forward — calls IBackend::ceil to compute $y = \lceil x \rceil$ .

Methods

grad_formula

→Storage

int grad_formula(const int & g)

source edit

Zero-gradient sentinel; never invoked since kHasGradient = false.

Parameters

gconst Storage&

Incoming gradient (ignored).

Returns

Storage

Empty CpuStorage placeholder.