Struct DetectDecodeOp 

pub struct DetectDecodeOp<D: FloatBytes + Send + Sync + 'static> {
    pub anchor_x: Vec<f32>,
    pub anchor_y: Vec<f32>,
    pub strides: Vec<f32>,
    pub block_a: i32,
    /* private fields */
}

Graph-level representation of the detect_decode op.

Stores precomputed anchor grid and stride data used to build the DetectDecodeRuntimeOp at lowering time via CustomOp::lower().

Fields

anchor_x: Vec<f32>

Anchor point x-coordinates, one per anchor.

anchor_y: Vec<f32>

Anchor point y-coordinates, one per anchor.

strides: Vec<f32>

Per-anchor stride (8/16/32 for the respective FPN scale).

block_a: i32

Kernel launch block size along the anchor dimension.

Implementations

impl<D: FloatBytes + Send + Sync + 'static> DetectDecodeOp<D>

pub fn new( anchor_x: Vec<f32>, anchor_y: Vec<f32>, strides: Vec<f32>, block_a: i32, ) -> Self

Creates the op with a precomputed anchor grid and launch block size.

Trait Implementations

impl<D: FloatBytes + Send + Sync + 'static> CustomOp for DetectDecodeOp<D>

fn name(&self) -> &str

Identifier used in error messages and debug output.

fn infer_output_shape(&self, input_shapes: &[&Shape]) -> Shape

Compute the output shape given the shapes of all input tensors in order.

fn as_any(&self) -> &dyn Any

Expose self as &dyn Any so the custom lowering can downcast to the concrete op type. Implement as fn as_any(&self) -> &dyn Any { self }.

fn lower(&self) -> Option<(String, String, String, Arc<dyn RuntimeOp>)>

Return kernel lowering info so TritonLowering can compile this op without a project-specific middleware. Return None to keep the existing middleware / error behaviour.

fn lower_backward_source(&self) -> String

Return the backward kernel source for this op (used in training mode). Return an empty string if this op has no backward pass.