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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223

use crate::{shapes::Dtype, tensor::*, tensor_ops::Device};

use super::*;

/// Splits input into multiple heads. `T` should be a tuple,
/// where every element of the tuple accepts the same input type.
///
/// This provides a utility for multi headed structures where
/// the tape needs to be moved around a number of times.
///
/// Each head's operations will be stored in its output's tape, while the operations stored in the
/// input tape will be saved in the first output's tape.
///
/// # Generics
/// - `T` the module to split the input into.
///
/// # Examples
/// ```rust
/// # use dfdx::prelude::*;
/// # let dev: Cpu = Default::default();
/// type Model = SplitInto<(Linear<5, 3>, Linear<5, 7>)>;
/// let model = dev.build_module::<Model, f32>();
/// let _: (Tensor<Rank1<3>, f32, _>, Tensor<Rank1<7>, f32, _>) = model.forward(dev.zeros::<Rank1<5>>());
/// ```
#[derive(Debug, Default, Clone)]
pub struct SplitInto<T>(pub T);

impl<T: BuildOnDevice<D, E>, D: Device<E>, E: Dtype> BuildOnDevice<D, E> for SplitInto<T> {
    type Built = SplitInto<T::Built>;
}

impl<E: Dtype, D: Device<E>, T: TensorCollection<E, D>> TensorCollection<E, D> for SplitInto<T> {
    type To<E2: Dtype, D2: Device<E2>> = SplitInto<T::To<E2, D2>>;

    fn iter_tensors<V: ModuleVisitor<Self, E, D>>(
        visitor: &mut V,
    ) -> Result<Option<Self::To<V::E2, V::D2>>, V::Err> {
        visitor.visit_fields(Self::module("0", |s| &s.0, |s| &mut s.0), SplitInto)
    }
}

macro_rules! tuple_impls {
    ($head:ident [$($tails:ident),+]) => {
        impl<
            Input: WithEmptyTape,
            $head: Module<Input>,
            $($tails : Module<Input, Error = $head::Error>,)+
        > Module<Input> for SplitInto<($head, $($tails,)+)> {
            type Output = (
                $head::Output,
                $($tails::Output),+
            );
            type Error = $head::Error;

            #[allow(non_snake_case)]
            fn try_forward(&self, x: Input) -> Result<Self::Output, $head::Error> {
                let ($head, $($tails,)+) = &self.0;
                let ($($tails,)+) = ($($tails.try_forward(x.with_empty_tape())?,)+);
                let $head = $head.try_forward(x)?;

                Ok(($head, $($tails,)+))
            }
        }

        impl<
            Input: WithEmptyTape,
            $head: ModuleMut<Input>,
            $($tails : ModuleMut<Input, Error = $head::Error>,)+
        > ModuleMut<Input> for SplitInto<($head, $($tails,)+)> {
            type Output = (
                $head::Output,
                $($tails::Output),+
            );
            type Error = $head::Error;

            #[allow(non_snake_case)]
            fn try_forward_mut(&mut self, x: Input) -> Result<Self::Output, $head::Error> {
                let ($head, $($tails,)+) = &mut self.0;
                let ($($tails,)+) = ($($tails.try_forward_mut(x.with_empty_tape())?,)+);
                let $head = $head.try_forward_mut(x)?;

                Ok(($head, $($tails,)+))
            }
        }
    }
}

tuple_impls!(A[B]);
tuple_impls!(A [B, C]);
tuple_impls!(A [B, C, D]);
tuple_impls!(A [B, C, D, E]);
tuple_impls!(A [B, C, D, E, F]);

#[cfg(test)]
mod tests {
    #![allow(clippy::type_complexity)]

    use super::*;
    use crate::{nn::builders::Linear, tests::*};
    use crate::{shapes::*, tensor_ops::*};

    #[test]
    fn test_unused() {
        let dev: TestDevice = Default::default();
        type Model = SplitInto<(Linear<1, 1>, Linear<1, 1>)>;
        let m = dev.build_module::<Model, TestDtype>();
        let (left, right) = m.forward(dev.sample_normal::<Rank1<1>>().leaky_trace());
        let r = right.retaped::<NoneTape>();
        let gr = right.mean().backward();
        let l = left.retaped::<NoneTape>();
        let gl = left.mean().backward();
        assert_ne!(gl.get(&l).array(), [TestDtype::zero(); 1]);
        assert_ne!(gr.get(&r).array(), [TestDtype::zero(); 1]);
    }

    #[test]
    fn test_split_into_2() {
        let dev: TestDevice = Default::default();
        type Model = SplitInto<(Linear<5, 1>, Linear<5, 2>)>;
        let m = dev.build_module::<Model, TestDtype>();
        let _: (
            Tensor<Rank1<1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<2>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank1<5>>().leaky_traced());
        let _: (
            Tensor<Rank2<3, 1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 2>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank2<3, 5>>().leaky_traced());
    }

    #[test]
    fn test_split_into_3() {
        let dev: TestDevice = Default::default();
        type Model = SplitInto<(Linear<5, 1>, Linear<5, 2>, Linear<5, 3>)>;
        let m = dev.build_module::<Model, TestDtype>();
        let _: (
            Tensor<Rank1<1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<2>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<3>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank1<5>>().leaky_traced());
        let _: (
            Tensor<Rank2<3, 1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 2>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 3>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank2<3, 5>>().leaky_traced());
    }

    #[test]
    fn test_split_into_4() {
        let dev: TestDevice = Default::default();
        type Model = SplitInto<(Linear<5, 1>, Linear<5, 2>, Linear<5, 3>, Linear<5, 4>)>;
        let m = dev.build_module::<Model, TestDtype>();
        let _: (
            Tensor<Rank1<1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<2>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<3>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<4>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank1<5>>().leaky_traced());
        let _: (
            Tensor<Rank2<3, 1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 2>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 3>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 4>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank2<3, 5>>().leaky_traced());
    }

    #[test]
    fn test_split_into_5() {
        let dev: TestDevice = Default::default();
        type Model = SplitInto<(
            Linear<5, 1>,
            Linear<5, 2>,
            Linear<5, 3>,
            Linear<5, 4>,
            Linear<5, 5>,
        )>;
        let m = dev.build_module::<Model, TestDtype>();
        let _: (
            Tensor<Rank1<1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<2>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<3>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<4>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<5>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank1<5>>().leaky_traced());
        let _: (
            Tensor<Rank2<3, 1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 2>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 3>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 4>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 5>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank2<3, 5>>().leaky_traced());
    }

    #[test]
    fn test_split_into_6() {
        let dev: TestDevice = Default::default();
        type Model = SplitInto<(
            Linear<5, 1>,
            Linear<5, 2>,
            Linear<5, 3>,
            Linear<5, 4>,
            Linear<5, 5>,
            Linear<5, 6>,
        )>;
        let m = dev.build_module::<Model, TestDtype>();
        let _: (
            Tensor<Rank1<1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<2>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<3>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<4>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<5>, _, _, OwnedTape<_, _>>,
            Tensor<Rank1<6>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank1<5>>().leaky_traced());
        let _: (
            Tensor<Rank2<3, 1>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 2>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 3>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 4>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 5>, _, _, OwnedTape<_, _>>,
            Tensor<Rank2<3, 6>, _, _, OwnedTape<_, _>>,
        ) = m.forward(dev.zeros::<Rank2<3, 5>>().leaky_traced());
    }
}