use tract_core::ops::cast::wire_cast;
use crate::infer::*;
use crate::internal::*;
#[derive(Debug, Clone, new, Default, Hash, PartialEq, Eq)]
pub struct ScatterElements {
axis: i64,
reduction: tract_core::ops::array::ScatterReduction,
}
impl Expansion for ScatterElements {
fn name(&self) -> StaticName {
"ScatterElements".into()
}
fn rules<'r, 'p: 'r, 's: 'r>(
&'s self,
s: &mut Solver<'r>,
inputs: &'p [TensorProxy],
outputs: &'p [TensorProxy],
) -> InferenceResult {
check_input_arity(inputs, 3)?;
check_output_arity(outputs, 1)?;
s.given_2(&inputs[0].datum_type, &inputs[2].datum_type, move |s, input, updates| {
let super_type: DatumType = DatumType::super_type_for([input, updates])
.with_context(|| format!("No supertype found for {input:?} and {updates:?}"))?;
s.equals(&outputs[0].datum_type, super_type)
})?;
s.equals(&inputs[0].rank, &inputs[1].rank)?;
s.equals(&inputs[1].shape, &inputs[2].shape)?;
s.equals(&outputs[0].shape, &inputs[0].shape)?;
Ok(())
}
fn wire(
&self,
prefix: &str,
model: &mut TypedModel,
inputs: &[OutletId],
) -> TractResult<TVec<OutletId>> {
let input_rank = model.outlet_fact(inputs[0])?.rank();
let axis = if self.axis < 0 { self.axis + input_rank as i64 } else { self.axis } as usize;
let super_type = if let Some(super_type) = DatumType::super_type_for([
model.outlet_fact(inputs[0])?.datum_type,
model.outlet_fact(inputs[2])?.datum_type,
]) {
super_type
} else {
bail!("Can not type op");
};
let casted = wire_cast(prefix, model, &[inputs[0], inputs[2]], super_type)?;
model.wire_node(
prefix,
tract_core::ops::array::ScatterElements { axis, reduction: self.reduction },
&[casted[0], inputs[1], casted[1]],
)
}
}