1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
use crate::infer::*;
use crate::internal::*;
use tract_itertools::Itertools;
#[derive(Debug, Clone, new, Hash)]
pub struct RmDims {
pub axes: Vec<isize>,
}
impl_dyn_hash!(RmDims);
impl RmDims {
fn compute_shape<D: DimLike>(&self, input: &[D]) -> TVec<D> {
let axes = self
.axes
.iter()
.map(|&a| if a < 0 { a + input.len() as isize } else { a } as usize)
.collect::<Vec<_>>();
input
.iter()
.enumerate()
.filter(|(ix, _d)| !axes.contains(ix))
.map(|(_ix, d)| d.clone())
.collect()
}
}
impl Expansion for RmDims {
fn name(&self) -> Cow<str> {
"RmDims".into()
}
op_hir!();
fn rules<'r, 'p: 'r, 's: 'r>(
&'s self,
s: &mut Solver<'r>,
inputs: &'p [TensorProxy],
outputs: &'p [TensorProxy],
) -> InferenceResult {
check_output_arity(&outputs, 1)?;
s.equals(&outputs[0].datum_type, &inputs[0].datum_type)?;
s.equals(&outputs[0].rank, (&inputs[0].rank).bex() - self.axes.len() as i64)?;
s.given(&inputs[0].rank, move |s, rank| {
for axis in &self.axes {
let axis = if *axis < 0 { axis + rank as isize } else { *axis } as usize;
s.equals(&inputs[0].shape[axis], 1.to_dim())?;
}
Ok(())
})?;
s.given(&inputs[0].shape, move |s, shape| {
let output_shape = self.compute_shape(&shape);
s.equals(&outputs[0].shape, output_shape)
})
}
fn wire(
&self,
prefix: &str,
target: &mut TypedModel,
inputs: &[OutletId],
) -> TractResult<TVec<OutletId>> {
let mut wire = inputs[0];
let rank = target.outlet_fact(inputs[0])?.rank();
let axes = self
.axes
.iter()
.map(|&a| if a < 0 { a + rank as isize } else { a } as usize)
.sorted()
.rev();
for axis in axes {
wire =
target.wire_node(format!("{}.axis-{}", prefix, axis), AxisOp::Rm(axis), &[wire])?
[0];
}
Ok(tvec!(wire))
}
}