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

#[allow(unused_imports)]
use crate::{
    dtypes::*,
    shapes::{RemoveDimTo, ReplaceDimTo, Shape},
    tensor::{launch_cfg, Cuda, Storage, Tensor},
};
use cudarc::driver::{DeviceSlice, LaunchAsync};

const GATHER_PTX_SRC: &str = include_str!(concat!(env!("OUT_DIR"), "/gather.ptx"));
const SELECT_PTX_SRC: &str = include_str!(concat!(env!("OUT_DIR"), "/select.ptx"));

macro_rules! impl_cuda_kernels {
    ($TypeName:ty, $GatherMod:tt, $GatherFwd:tt, $GatherBwd:tt, $SelectMod:tt, $SelectFwd:tt, $SelectBwd:tt) => {
        impl super::ReplaceDimKernel<$TypeName> for Cuda {
            fn forward<Src: Shape, Dst: Shape, Idx: Shape>(
                &self,
                inp: &Tensor<Src, $TypeName, Self>,
                idx: &Tensor<Idx, usize, Self>,
            ) -> Result<Tensor<Dst, $TypeName, Self>, Self::Err>
            where
                Src: ReplaceDimTo<Dst, Idx>,
            {
                if !self.dev.has_func($GatherMod, $GatherFwd) {
                    self.dev.load_ptx(
                        GATHER_PTX_SRC.into(),
                        $GatherMod,
                        &[$GatherFwd, $GatherBwd],
                    )?;
                }

                let dst = inp.shape.replace(idx.shape);
                let numel = dst.num_elements();
                let mut storage = unsafe { self.alloc_empty::<$TypeName>(numel) }?;
                self.dev.memset_zeros(&mut storage)?;

                let inp_dims = self.dev.htod_copy(inp.shape.concrete().into())?;
                let idx_dims = self.dev.htod_copy(idx.shape.concrete().into())?;
                let inp_strides = self.dev.htod_copy(inp.strides.into())?;
                let idx_strides = self.dev.htod_copy(idx.strides.into())?;

                let fwd_fn = self.dev.get_func($GatherMod, $GatherFwd).unwrap();
                let cfg = launch_cfg::<128>(numel as u32);
                let params = (
                    numel,             // const size_t numel,
                    inp.data.as_ref(), // const float *inp,
                    Src::NUM_DIMS,     // const size_t inp_num_dims,
                    &inp_dims,         // const size_t *inp_dims,
                    &inp_strides,      // const size_t *inp_strides,
                    idx.data.as_ref(), // const float *idx,
                    Idx::NUM_DIMS,     // const size_t idx_num_dims,
                    &idx_dims,         // const size_t *idx_dims,
                    &idx_strides,      // const size_t *idx_strides,
                    &mut storage,      // float *out,
                    Dst::NUM_DIMS,     // const size_t out_num_dims,
                );
                unsafe { fwd_fn.launch(cfg, params) }?;

                Ok(self.build_tensor(dst, dst.strides(), storage))
            }

            fn backward<Src: Shape, Dst: Shape, Idx: Shape>(
                &self,
                inp: &Tensor<Src, $TypeName, Self>,
                grad_inp: &mut <Self as Storage<$TypeName>>::Vec,
                idx: &Tensor<Idx, usize, Self>,
                _: &Tensor<Dst, $TypeName, Self>,
                grad_out: &<Self as Storage<$TypeName>>::Vec,
            ) -> Result<(), Self::Err>
            where
                Src: ReplaceDimTo<Dst, Idx>,
            {
                let bwd_fn = self.dev.get_func($GatherMod, $GatherBwd).unwrap();
                let numel = grad_out.len();

                let inp_dims = self.dev.htod_copy(inp.shape.concrete().into())?;
                let idx_dims = self.dev.htod_copy(idx.shape.concrete().into())?;
                let inp_strides = self.dev.htod_copy(inp.strides.into())?;
                let idx_strides = self.dev.htod_copy(idx.strides.into())?;

                let cfg = launch_cfg::<128>(numel as u32);
                let params = (
                    numel,             // const size_t numel,
                    grad_inp,          // float *grad_inp,
                    Src::NUM_DIMS,     // const size_t inp_num_dims,
                    &inp_dims,         // const size_t *inp_dims,
                    &inp_strides,      // const size_t *inp_strides,
                    idx.data.as_ref(), // const float *idx,
                    Idx::NUM_DIMS,     // const size_t idx_num_dims,
                    &idx_dims,         // const size_t *idx_dims,
                    &idx_strides,      // const size_t *idx_strides,
                    grad_out,          // const float *grad_out,
                    Dst::NUM_DIMS,     // const size_t out_num_dims,
                );
                unsafe { bwd_fn.launch(cfg, params) }?;
                Ok(())
            }
        }

        impl super::RemoveDimKernel<$TypeName> for Cuda {
            fn forward<Src: Shape, Dst: Shape, Idx: Shape>(
                &self,
                inp: &Tensor<Src, $TypeName, Self>,
                idx: &Tensor<Idx, usize, Self>,
            ) -> Result<Tensor<Dst, $TypeName, Self>, Self::Err>
            where
                Src: RemoveDimTo<Dst, Idx>,
            {
                if !self.dev.has_func($SelectMod, $SelectFwd) {
                    self.dev.load_ptx(
                        SELECT_PTX_SRC.into(),
                        $SelectMod,
                        &[$SelectFwd, $SelectBwd],
                    )?;
                }

                let dst = inp.shape.remove(idx.shape);
                let numel = dst.num_elements();
                let mut storage = unsafe { self.alloc_empty::<$TypeName>(numel) }?;
                self.dev.memset_zeros(&mut storage)?;

                let inp_dims = self.dev.htod_copy(inp.shape.concrete().into())?;
                let idx_dims = self.dev.htod_copy(idx.shape.concrete().into())?;
                let dst_dims = self.dev.htod_copy(dst.concrete().into())?;
                let inp_strides = self.dev.htod_copy(inp.strides.into())?;
                let idx_strides = self.dev.htod_copy(idx.strides.into())?;
                let dst_strides = self.dev.htod_copy(dst.strides().into())?;

                let fwd_fn = self.dev.get_func($SelectMod, $SelectFwd).unwrap();
                let cfg = launch_cfg::<128>(numel as u32);
                let params = (
                    numel,             // const size_t numel,
                    inp.data.as_ref(), // const float *inp,
                    Src::NUM_DIMS,     // const size_t inp_num_dims,
                    &inp_dims,         // const size_t *inp_dims,
                    &inp_strides,      // const size_t *inp_strides,
                    idx.data.as_ref(), // const float *idx,
                    Idx::NUM_DIMS,     // const size_t idx_num_dims,
                    &idx_dims,         // const size_t *idx_dims,
                    &idx_strides,      // const size_t *idx_strides,
                    &mut storage,      // float *out,
                    &dst_dims,         // const size_t *out_dims,
                    &dst_strides,      // const size_t *out_strides
                );
                unsafe { fwd_fn.launch(cfg, params) }?;

                Ok(self.build_tensor(dst, dst.strides(), storage))
            }

            fn backward<Src: Shape, Dst: Shape, Idx: Shape>(
                &self,
                inp: &Tensor<Src, $TypeName, Self>,
                grad_inp: &mut <Self as Storage<$TypeName>>::Vec,
                idx: &Tensor<Idx, usize, Self>,
                out: &Tensor<Dst, $TypeName, Self>,
                grad_out: &<Self as Storage<$TypeName>>::Vec,
            ) -> Result<(), Self::Err>
            where
                Src: RemoveDimTo<Dst, Idx>,
            {
                let bwd_fn = self.dev.get_func($SelectMod, $SelectBwd).unwrap();
                let numel = grad_out.len();

                let inp_dims = self.dev.htod_copy(inp.shape.concrete().into())?;
                let idx_dims = self.dev.htod_copy(idx.shape.concrete().into())?;
                let out_dims = self.dev.htod_copy(out.shape.concrete().into())?;
                let inp_strides = self.dev.htod_copy(inp.strides.into())?;
                let idx_strides = self.dev.htod_copy(idx.strides.into())?;
                let out_strides = self.dev.htod_copy(out.strides.into())?;

                let cfg = launch_cfg::<128>(numel as u32);
                let params = (
                    numel,             // const size_t numel,
                    grad_inp,          // float *grad_inp,
                    Src::NUM_DIMS,     // const size_t inp_num_dims,
                    &inp_dims,         // const size_t *inp_dims,
                    &inp_strides,      // const size_t *inp_strides,
                    idx.data.as_ref(), // const float *idx,
                    Idx::NUM_DIMS,     // const size_t idx_num_dims,
                    &idx_dims,         // const size_t *idx_dims,
                    &idx_strides,      // const size_t *idx_strides,
                    grad_out,          // const float *grad_out,
                    &out_dims,         // const size_t *out_dims,
                    &out_strides,      // const size_t *out_strides
                );
                unsafe { bwd_fn.launch(cfg, params) }?;
                Ok(())
            }
        }
    };
}

#[cfg(feature = "f16")]
impl_cuda_kernels!(
    f16,
    "gather_f16",
    "gather_fwd_f16",
    "gather_bwd_f16",
    "select_f16",
    "select_fwd_f16",
    "select_bwd_f16"
);

#[cfg(feature = "f16")]
impl_cuda_kernels!(
    AMP<f16>,
    "gather_f16",
    "gather_fwd_f16",
    "gather_bwd_f16",
    "select_f16",
    "select_fwd_f16",
    "select_bwd_f16"
);

impl_cuda_kernels!(
    f32,
    "gather_f32",
    "gather_fwd_f32",
    "gather_bwd_f32",
    "select_f32",
    "select_fwd_f32",
    "select_bwd_f32"
);
impl_cuda_kernels!(
    f64,
    "gather_f64",
    "gather_fwd_f64",
    "gather_bwd_f64",
    "select_f64",
    "select_fwd_f64",
    "select_bwd_f64"
);