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
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147

use crate::internal::*;

#[derive(Debug, Clone, new)]
pub struct RmDims {
    pub axes: Vec<usize>,
}

impl RmDims {
    fn compute_shape<D: DimLike>(&self, input: &[D]) -> TVec<D> {
        input
            .iter()
            .enumerate()
            .filter(|(ix, _d)| !self.axes.contains(ix))
            .map(|(_ix, d)| d.clone())
            .collect()
    }

    /// Evaluates the operation given the input tensors.
    fn eval_t<T: Datum>(&self, input: Arc<Tensor>) -> TractResult<TVec<Arc<Tensor>>> {
        let shape = self.compute_shape(input.shape());
        Ok(tvec![input.into_tensor().into_array::<T>()?.into_shape(&*shape)?.into_arc_tensor()])
    }
}

impl Op for RmDims {
    fn name(&self) -> Cow<str> {
        "RmDims".into()
    }

    fn info(&self) -> TractResult<Vec<String>> {
        Ok(vec![format!("axes: {:?}", self.axes)])
    }

    fn declutter(
        &self,
        model: &TypedModel,
        node: &TypedNode,
    ) -> TractResult<Option<TypedModelPatch>> {
        'axis: for &rm_axis in &self.axes {
            let mut current = node;
            let mut axis = rm_axis;
            while let Some(prec) = model.single_prec(current.id)? {
                if let Some(add_dims) = prec.op_as::<super::AddDims>() {
                    if add_dims.axes.contains(&axis) {
                        let mut patch = TypedModelPatch::default();
                        let mut wire: OutletId = patch.tap_model(model, prec.inputs[0])?.into();
                        if add_dims.axes.len() > 1 {
                            let mut add_dims = add_dims.clone();
                            add_dims.axes.retain(|&a| a != axis);
                            wire = patch.wire_node(&*prec.name, add_dims, [wire].as_ref())?[0];
                        }
                        let mut next = model.single_succ(prec.id)?.unwrap();
                        while next.id != node.id {
                            let op = next
                                .op
                                .dispose_dummy_axis(model, next, axis)?
                                .unwrap_or_else(|| next.op.clone());
                            wire = patch.wire_node(&*next.name, op, [wire].as_ref())?[0];
                            axis = next
                                .op
                                .axes_info(model, next)?
                                .unary_track_axis_down(axis, true)
                                .unwrap();
                            next = model.single_succ(next.id)?.unwrap();
                        }
                        if self.axes.len() > 1 {
                            let mut rm_dims = self.clone();
                            rm_dims.axes.retain(|&a| a != rm_axis);
                            wire = patch.wire_node(&*node.name, rm_dims, [wire].as_ref())?[0];
                        }
                        patch.shunt_outside(OutletId::new(node.id, 0), wire)?;
                        return Ok(Some(patch));
                    }
                }
                let invariants = prec.op.axes_info(model, prec)?;
                if let Some(up_axis) = invariants.unary_track_axis_up(axis, true) {
                    current = prec;
                    axis = up_axis;
                } else {
                    continue 'axis;
                }
            }
        }
        Ok(None)
    }

    canonic!();
    op_as_typed_op!();
}

impl StatelessOp for RmDims {
    /// Evaluates the operation given the input tensors.
    fn eval(&self, mut inputs: TVec<Arc<Tensor>>) -> TractResult<TVec<Arc<Tensor>>> {
        let input = args_1!(inputs);
        dispatch_datum!(Self::eval_t(input.datum_type())(self, input))
    }
}

impl InferenceRulesOp for RmDims {
    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 i32)?;
        for axis in &self.axes {
            s.equals(&inputs[0].shape[*axis], 1.to_dim())?;
        }
        s.given(&inputs[0].shape, move |s, shape| {
            let output_shape = self.compute_shape(&shape);
            s.equals(&outputs[0].shape, output_shape)
        })
    }

    inference_op_as_op!();
    to_typed!();
}

impl TypedOp for RmDims {
    typed_op_as_op!();

    fn output_facts(&self, inputs: &[&TypedTensorInfo]) -> TractResult<TVec<TypedTensorInfo>> {
        Ok(tvec!(TypedTensorInfo::dt_shape(
            inputs[0].datum_type,
            self.compute_shape(&*inputs[0].shape.to_tvec()).as_ref(),
        )?))
    }

    fn pulsify(
        &self,
        _source: &NormalizedModel,
        node: &NormalizedNode,
        target: &mut PulsedModel,
        mapping: &HashMap<OutletId, OutletId>,
        _pulse: usize,
    ) -> TractResult<TVec<OutletId>> {
        let input = mapping[&node.inputs[0]];
        let mut fact = target.outlet_fact(input)?.clone();
        fact.shape = self.compute_shape(&fact.shape);
        fact.axis -= self.axes.iter().filter(|&ax| *ax <= fact.axis).count();
        let id = target.chain_after(input, &*node.name, self.clone(), tvec!(fact))?;
        Ok(tvec!(OutletId::new(id, 0)))
    }
}