dfdx 0.13.0

Ergonomic auto differentiation in Rust, with pytorch like apis.
Documentation 
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
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238

mod cpu_kernel;

#[cfg(feature = "cuda")]
mod cuda_kernel;

use crate::{shapes::*, tensor::*};

pub trait ReshapeKernel<E: Dtype>: Storage<E> {
    fn forward<Src: Shape, Dst: Shape>(
        &self,
        dst: &Dst,
        inp: &Tensor<Src, E, Self>,
    ) -> Result<Tensor<Dst, E, Self>, Self::Err>;
    fn backward<Src: Shape, Dst: Shape>(
        &self,
        dst: &Dst,
        inp: &Tensor<Src, E, Self>,
        grad_inp: &mut Self::Vec,
        grad_out: &Self::Vec,
    ) -> Result<(), Self::Err>;
}

/// Changes the shape of a tensor without re-ordering axes. If the tensor is contiguous
/// already, then no data movement will occur. If the tensor is not contiguous, the
/// result of this will be contiguous.
///
/// Compile time reshapes:
/// ```rust
/// # use dfdx::prelude::*;
/// # let dev: Cpu = Default::default();
/// let t: Tensor<Rank2<2, 4>, f32, _> = dev.zeros();
/// let t: Tensor<Rank1<8>, f32, _> = t.reshape();
/// ```
///
/// Compile time failure:
/// ```compile_fail
/// # use dfdx::prelude::*;
/// # let dev: Cpu = Default::default();
/// let t: Tensor<Rank2<2, 4>, f32, _> = dev.zeros();
/// let t: Tensor<Rank1<7>, f32, _> = t.reshape();
/// ```
///
/// Runtime reshapes:
/// ```rust
/// # use dfdx::prelude::*;
/// # let dev: Cpu = Default::default();
/// let t: Tensor<Rank2<2, 4>, f32, _> = dev.zeros();
/// let t: Tensor<(usize, ), f32, _> = t.reshape_like(&(8, ));
/// ```
pub trait ReshapeTo: HasErr + HasShape {
    /// Reshapes a tensor to a different compile time shape.
    fn reshape<Dst: ConstShape>(self) -> Self::WithShape<Dst>
    where
        Self::Shape: ConstShape,
    {
        <Self::Shape as AssertSameNumel<Dst>>::assert_same_numel();
        self.try_reshape().unwrap()
    }
    /// Reshapes a tensor to a different compile time shape.
    fn try_reshape<Dst: ConstShape>(self) -> Result<Self::WithShape<Dst>, Self::Err>
    where
        Self::Shape: ConstShape,
    {
        <Self::Shape as AssertSameNumel<Dst>>::assert_same_numel();
        self.try_reshape_like::<Dst>(&Default::default())
    }
    /// Reshapes a tensor to a different runtime shape.
    fn reshape_like<Dst: Shape>(self, dst: &Dst) -> Self::WithShape<Dst> {
        self.try_reshape_like(dst).unwrap()
    }
    /// Ensures the tensor's memory is contiguous.
    ///
    /// If the memory is already contiguous no copying is performed.
    fn contiguous(self) -> Self::WithShape<Self::Shape> {
        self.try_contiguous().unwrap()
    }
    /// See [`ReshapeTo::contiguous`]
    fn try_contiguous(self) -> Result<Self::WithShape<Self::Shape>, Self::Err> {
        let shape = *self.shape();
        self.try_reshape_like(&shape)
    }
    /// Reshapes a tensor to a different runtime shape.
    fn try_reshape_like<Dst: Shape>(self, dst: &Dst) -> Result<Self::WithShape<Dst>, Self::Err>;
}

impl<S: Shape, E: Dtype, D: ReshapeKernel<E>, T: Tape<E, D>> ReshapeTo for Tensor<S, E, D, T> {
    fn try_reshape_like<Dst: Shape>(self, dst: &Dst) -> Result<Self::WithShape<Dst>, Self::Err> {
        assert_eq!(self.shape().num_elements(), dst.num_elements());
        if self.shape.strides() == self.strides {
            Ok(Tensor {
                id: self.id,
                data: self.data,
                shape: *dst,
                strides: dst.strides(),
                device: self.device,
                tape: self.tape,
            })
        } else {
            let (inp, mut tape) = self.split_tape();
            let out = inp.device.forward(dst, &inp)?;
            let inp_ghost = inp.ghost();
            let out_ghost = out.ghost();
            let dst = *dst;
            tape.add_backward_op(move |grads| {
                grads.try_alloc_for(&inp_ghost)?;
                grads.try_alloc_for(&out_ghost)?;
                let (grad_inp, grad_out) = grads.mut_and_ref(&inp_ghost, &out_ghost);
                inp.device.backward(&dst, &inp, grad_inp, grad_out)
            });
            Ok(out.put_tape(tape))
        }
    }
}

#[cfg(test)]
mod tests {
    use crate::tensor::*;
    use crate::tensor_ops::*;
    use crate::tests::*;

    use super::*;

    #[test]
    #[should_panic]
    fn test_invalid_reshape() {
        let dev: TestDevice = Default::default();
        let t: Tensor<(usize,), TestDtype, _> = dev.zeros_like(&(5,));
        let _ = t.reshape_like(&(7,));
    }

    #[test]
    fn test_unit_non_contiguous_reshapes() {
        let dev: TestDevice = Default::default();
        let t: Tensor<Rank2<2, 3>, usize, _> = dev.zeros::<Rank0>().broadcast();
        let _: Tensor<Rank1<6>, usize, _> = t.reshape();
    }

    #[test]
    fn test_valid_reshapes() {
        let dev: TestDevice = Default::default();

        let t: Tensor<Rank0, TestDtype, _> = dev.zeros();
        let _: Tensor<Rank1<1>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank2<1, 1>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank3<1, 1, 1>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank4<1, 1, 1, 1>, TestDtype, _> = t.reshape();

        let t: Tensor<Rank1<16>, TestDtype, _> = dev.zeros();
        let _: Tensor<Rank1<16>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank2<2, 8>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank3<2, 2, 4>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank4<2, 2, 2, 2>, TestDtype, _> = t.reshape();

        let t: Tensor<Rank2<2, 8>, TestDtype, _> = dev.zeros();
        let _: Tensor<Rank1<16>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank2<8, 2>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank3<2, 2, 4>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank4<2, 2, 2, 2>, TestDtype, _> = t.reshape();

        let t: Tensor<Rank3<2, 2, 4>, TestDtype, _> = dev.zeros();
        let _: Tensor<Rank1<16>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank2<2, 8>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank3<4, 2, 2>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank4<2, 2, 2, 2>, TestDtype, _> = t.reshape();

        let t: Tensor<Rank4<2, 2, 2, 2>, TestDtype, _> = dev.zeros();
        let _: Tensor<Rank1<16>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank2<2, 8>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank3<2, 2, 4>, TestDtype, _> = t.clone().reshape();
        let _: Tensor<Rank4<4, 1, 2, 2>, TestDtype, _> = t.reshape();
    }

    #[test]
    fn test_1d_reshape() {
        let dev: TestDevice = Default::default();
        let a = dev
            .tensor([0.1, 0.2, 0.3, 0.4, 0.5, 0.6])
            .to_dtype::<TestDtype>();
        let b = a.leaky_trace().reshape::<Rank2<2, 3>>();
        assert_close_to_literal!(b, [[0.1, 0.2, 0.3], [0.4, 0.5, 0.6]]);
        let g = b.exp().mean().backward();
        assert_close_to_literal!(
            g.get(&a),
            [0.18419516, 0.20356713, 0.22497648, 0.24863747, 0.2747869, 0.3036865]
        )
    }

    #[test]
    fn test_1d_reshape_non_contiguous() {
        let dev: TestDevice = Default::default();
        let a = dev
            .tensor([[0.1, 0.2, 0.3], [0.4, 0.5, 0.6]])
            .to_dtype::<TestDtype>();
        let b = a
            .leaky_trace()
            .permute::<Rank2<3, 2>, _>()
            .reshape::<Rank1<6>>();
        assert_close_to_literal!(b, [0.1, 0.4, 0.2, 0.5, 0.3, 0.6]);
        let g = b.exp().mean().backward();
        assert_close_to_literal!(
            g.get(&a),
            [
                [0.18419516, 0.20356713, 0.22497648],
                [0.24863747, 0.2747869, 0.3036865],
            ]
        )
    }

    #[test]
    fn test_reshape_broadcasted() {
        let dev: TestDevice = Default::default();
        let a: Tensor<Rank2<2, 3>, TestDtype, _> =
            dev.tensor([1., 2., 3.]).to_dtype::<TestDtype>().broadcast();
        let b: Tensor<Rank2<3, 2>, TestDtype, _> = a.clone().reshape();

        #[cfg(feature = "cuda")]
        use cudarc::driver::DeviceSlice;

        assert_eq!(b.data.len(), 6);
        assert_eq!(a.as_vec(), b.as_vec());
        assert_close_to_literal!(b, [[1., 2.], [3., 1.], [2., 3.]]);
    }

    #[test]
    fn test_contiguous() {
        let dev: TestDevice = Default::default();

        let a = dev
            .tensor([[0.1, 0.2, 0.3], [0.4, 0.5, 0.6]])
            .to_dtype::<TestDtype>();

        let b1 = a.clone().contiguous();
        assert_eq!(a.strides, b1.strides);

        let b2 = a.permute::<Rank2<3, 2>, _>().contiguous();
        assert_eq!(b2.strides, [2, 1]);
    }
}