Struct GridSamplePlan 

pub struct GridSamplePlan<T: Element> { /* private fields */ }

grid_sample plan.

Sample an NCHW input at arbitrary normalized (x, y) coordinates supplied via grid: [N, OH, OW, 2] (PyTorch F.grid_sample). Trailblazer config matches PyTorch defaults: mode='bilinear', padding_mode='zeros', align_corners=false.

When to use: forward 2-D grid sample. Pair with GridSampleBackwardPlan.

Dtypes: {f32, f64}.

Shape limits: rank-4 NCHW input; rank-4 grid [N, OH, OW, 2]; output [N, C, OH, OW].

Workspace: none.

Precision guarantee: deterministic, bit-stable.

Implementations

impl<T: Element> GridSamplePlan<T>

pub fn select( _stream: &Stream, desc: &GridSampleDescriptor, _pref: PlanPreference, ) -> Result<Self>

Pick a kernel.

pub fn can_implement(&self, args: &GridSampleArgs<'_, T>) -> Result<()>

Validate args.

pub fn workspace_size(&self) -> usize

Workspace (zero).

pub fn sku(&self) -> KernelSku

Identity.

pub fn precision_guarantee(&self) -> PrecisionGuarantee

Numerical guarantees.

pub fn run( &self, stream: &Stream, _workspace: Workspace<'_>, args: GridSampleArgs<'_, T>, ) -> Result<()>

Launch.