use ndarray::*;
use tract_core::internal::*;
#[derive(Debug, Clone, new)]
pub struct GatherNd {}
pub fn gather_nd(_pb: &crate::tfpb::node_def::NodeDef) -> TractResult<Box<Op>> {
Ok(Box::new(GatherNd::new()))
}
impl GatherNd {
fn eval_t<T: Datum + Copy>(
&self,
data: &Arc<Tensor>,
indices: &ArrayViewD<i32>,
) -> TractResult<TVec<Arc<Tensor>>> {
let data = data.to_array_view::<T>()?;
let mut shape: TVec<usize> = indices.shape().into();
let n = shape.pop().unwrap();
shape.extend(data.shape()[n..].iter().cloned());
let mut array = unsafe { ArrayD::<T>::uninitialized(&*shape) };
for prefix in ndarray::indices(&indices.shape()[0..indices.ndim() - 1]) {
let mut dst = array.view_mut();
let mut coords = indices.view();
for &x in prefix.slice().iter() {
dst.index_axis_inplace(Axis(0), x);
coords.index_axis_inplace(Axis(0), x);
}
let mut src = data.view();
for &x in coords.iter() {
src.index_axis_inplace(Axis(0), x as _);
}
dst.assign(&src);
}
Ok(tvec![array.into_arc_tensor()])
}
}
impl Op for GatherNd {
fn name(&self) -> Cow<str> {
"tf.GatherNd".into()
}
}
impl StatelessOp for GatherNd {
fn eval(&self, mut inputs: TVec<Arc<Tensor>>) -> TractResult<TVec<Arc<Tensor>>> {
let (data, indices) = args_2!(inputs);
let indices = indices.cast_to::<i32>()?;
let indices = indices.to_array_view::<i32>()?;
dispatch_copy!(Self::eval_t(data.datum_type())(self, &data, &indices))
}
}
impl InferenceRulesOp for GatherNd {
fn rules<'r, 'p: 'r, 's: 'r>(
&'s self,
s: &mut Solver<'r>,
inputs: &'p [TensorProxy],
outputs: &'p [TensorProxy],
) -> InferenceResult {
check_input_arity(&inputs, 2)?;
check_output_arity(&outputs, 1)?;
s.equals(&outputs[0].datum_type, &inputs[0].datum_type)?;
s.given(&inputs[1].rank, move |s, indices_rank| {
let indices_rank = indices_rank as usize;
for i in 0..(indices_rank - 1) {
s.equals(&outputs[0].shape[i], &inputs[1].shape[i])?;
}
s.given_2(
&inputs[1].shape[indices_rank - 1],
&inputs[1].rank,
move |s, n, input_rank| {
if let Ok(n) = n.to_integer() {
for i in 0..(input_rank - n) as usize {
s.equals(&outputs[0].shape[indices_rank - 1 + i], &inputs[1].shape[i])?;
}
}
Ok(())
},
)
})
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn simple_indexing() {
let g = GatherNd::new();
assert_eq!(
g.eval(tvec!(rctensor2(&[[1, 2], [3, 4]]), rctensor2(&[[0, 0], [1, 1]]))).unwrap(),
tvec!(rctensor1(&[1, 4]))
);
}
#[test]
fn slice_indexing() {
let g = GatherNd::new();
assert_eq!(
g.eval(tvec!(rctensor2(&[[1, 2], [3, 4]]), rctensor2(&[[1], [0]]))).unwrap(),
tvec!(rctensor2(&[[3, 4], [1, 2]]))
);
}
#[test]
fn tensor_3d_1() {
let g = GatherNd::new();
let t = rctensor3(&[[[10, 20], [30, 40]], [[11, 21], [31, 41]]]);
assert_eq!(
g.eval(tvec!(t.clone(), rctensor2(&[[1]]))).unwrap(),
tvec!(rctensor3(&[[[11, 21], [31, 41]]]))
);
}
#[test]
fn tensor_3d_2() {
let g = GatherNd::new();
let t = rctensor3(&[[[10, 20], [30, 40]], [[11, 21], [31, 41]]]);
assert_eq!(
g.eval(tvec!(t.clone(), rctensor2(&[[0, 1], [1, 0]]))).unwrap(),
tvec!(rctensor2(&[[30, 40], [11, 21]]))
);
}
#[test]
fn tensor_3d_3() {
let g = GatherNd::new();
let t = rctensor3(&[[[10, 20], [30, 40]], [[11, 21], [31, 41]]]);
assert_eq!(
g.eval(tvec!(t.clone(), rctensor2(&[[0, 0, 1], [1, 0, 1]]))).unwrap(),
tvec!(rctensor1(&[20, 21]))
);
}
}