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
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
use crate::{Graph, ShapeOrTensor, Tensor, ns_number_array_from_slice};
use objc2::msg_send;
use objc2::rc::Retained;
use objc2_foundation::NSString;
mod convolution_2d_op_descriptor;
mod convolution_3d_op_descriptor;
pub use convolution_2d_op_descriptor::Convolution2DOpDescriptor;
pub use convolution_3d_op_descriptor::Convolution3DOpDescriptor;
impl Graph {
/// Creates a 2-D (forward) convolution operation.
///
/// Corresponds to Objective-C selector
/// `convolution2DWithSourceTensor:weightsTensor:descriptor:name:`.
///
/// # Arguments
///
/// * `source_tensor` – Rank-4 source tensor (layout defined by
/// `descriptor.data_layout`).
/// * `weights_tensor` – Rank-4 weights tensor (layout defined by
/// `descriptor.weights_layout`).
/// * `descriptor` – Strides, dilation rates, padding, and layout
/// information.
/// * `name` – Optional debug label.
///
/// # Returns
///
/// A `Retained<Tensor>` containing the convolution result.
pub fn convolution_2d(
&self,
source_tensor: &Tensor,
weights_tensor: &Tensor,
descriptor: &Convolution2DOpDescriptor,
name: Option<&str>,
) -> Retained<Tensor> {
unsafe {
msg_send![
self,
convolution2DWithSourceTensor: source_tensor,
weightsTensor: weights_tensor,
descriptor: descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
}
/// Creates a 2-D convolution gradient operation with respect to the
/// *source* tensor.
///
/// Computes `dL/dS = dL/dR · dR/dS`, where `R` is the forward-pass result
/// and `incoming_gradient_tensor` supplies `dL/dR`.
///
/// # Arguments
///
/// * `incoming_gradient_tensor` – Loss gradient with respect to the forward
/// result.
/// * `weights_tensor` – Weights tensor from the forward pass.
/// * `output_shape` – Shape of the forward-pass source tensor (static
/// slice or dynamic tensor via [`ShapeOrTensor`]).
/// * `forward_convolution_descriptor` – Descriptor used in the forward op.
/// * `name` – Optional debug label.
///
/// # Returns
///
/// A `Retained<Tensor>` containing the data-gradient result.
pub fn convolution_2d_data_gradient<'a>(
&self,
incoming_gradient_tensor: &Tensor,
weights_tensor: &Tensor,
output_shape: ShapeOrTensor<'a>,
forward_convolution_descriptor: &Convolution2DOpDescriptor,
name: Option<&str>,
) -> Retained<Tensor> {
unsafe {
match output_shape {
ShapeOrTensor::Shape(output_shape) => {
msg_send![
self,
convolution2DDataGradientWithIncomingGradientTensor: incoming_gradient_tensor,
weightsTensor: weights_tensor,
outputShape: &*ns_number_array_from_slice(output_shape),
forwardConvolutionDescriptor: forward_convolution_descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
ShapeOrTensor::Tensor(output_shape_tensor) => {
msg_send![
self,
convolution2DDataGradientWithIncomingGradientTensor: incoming_gradient_tensor,
weightsTensor: weights_tensor,
outputShapeTensor: output_shape_tensor,
forwardConvolutionDescriptor: forward_convolution_descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
}
}
}
/// Creates a 2-D convolution gradient operation with respect to the
/// *weights* tensor.
///
/// Computes `dL/dW = dL/dR · dR/dW`, where `R` is the forward convolution
/// result.
///
/// # Arguments
///
/// * `incoming_gradient_tensor` – Loss gradient with respect to the forward
/// result.
/// * `source_tensor` – Source tensor from the forward pass.
/// * `output_shape` – Shape of the forward-pass weights tensor (static or
/// dynamic via [`ShapeOrTensor`]).
/// * `forward_convolution_descriptor` – Descriptor used in the forward op.
/// * `name` – Optional debug label.
///
/// # Returns
///
/// A `Retained<Tensor>` containing the weights-gradient result.
pub fn convolution_2d_weights_gradient<'a>(
&self,
incoming_gradient_tensor: &Tensor,
source_tensor: &Tensor,
output_shape: ShapeOrTensor<'a>,
forward_convolution_descriptor: &Convolution2DOpDescriptor,
name: Option<&str>,
) -> Retained<Tensor> {
unsafe {
match output_shape {
ShapeOrTensor::Shape(output_shape) => {
msg_send![
self,
convolution2DWeightsGradientWithIncomingGradientTensor: incoming_gradient_tensor,
sourceTensor: source_tensor,
outputShape: &*ns_number_array_from_slice(output_shape),
forwardConvolutionDescriptor: forward_convolution_descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
ShapeOrTensor::Tensor(output_shape_tensor) => {
msg_send![
self,
convolution2DWeightsGradientWithIncomingGradientTensor: incoming_gradient_tensor,
sourceTensor: source_tensor,
outputShapeTensor: output_shape_tensor,
forwardConvolutionDescriptor: forward_convolution_descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
}
}
}
/// Creates a 3-D (forward) convolution operation. Parameter semantics
/// mirror the 2-D variant.
///
/// # Arguments
///
/// * `source_tensor` – Rank-5 source tensor.
/// * `weights_tensor` – Rank-5 weights tensor.
/// * `descriptor` – Strides, dilation rates, padding, and layout
/// information.
/// * `name` – Optional debug label.
///
/// # Returns
///
/// A `Retained<Tensor>` containing the convolution result.
pub fn convolution_3d<'a>(
&self,
source_tensor: &Tensor,
weights_tensor: &Tensor,
descriptor: &Convolution3DOpDescriptor,
name: Option<&str>,
) -> Retained<Tensor> {
unsafe {
msg_send![
self,
convolution3DWithSourceTensor: source_tensor,
weightsTensor: weights_tensor,
descriptor: descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
}
/// Creates a 3-D convolution gradient operation with respect to the
/// *source* tensor.
///
/// Computes `dL/dS = dL/dR · dR/dS`, analogous to the 2-D variant.
///
/// # Arguments
///
/// * `incoming_gradient_tensor` – Loss gradient with respect to the forward
/// result.
/// * `weights_tensor` – Weights tensor from the forward pass.
/// * `output_shape` – Forward-pass source shape (static slice or dynamic
/// tensor).
/// * `forward_convolution_descriptor` – Descriptor used in the forward op.
/// * `name` – Optional debug label.
///
/// # Returns
///
/// A `Retained<Tensor>` containing the data-gradient result.
pub fn convolution_3d_data_gradient<'a>(
&self,
incoming_gradient_tensor: &Tensor,
weights_tensor: &Tensor,
output_shape: ShapeOrTensor<'a>,
forward_convolution_descriptor: &Convolution3DOpDescriptor,
name: Option<&str>,
) -> Retained<Tensor> {
unsafe {
match output_shape {
ShapeOrTensor::Shape(output_shape) => {
msg_send![
self,
convolution3DDataGradientWithIncomingGradientTensor: incoming_gradient_tensor,
weightsTensor: weights_tensor,
outputShape: &*ns_number_array_from_slice(output_shape),
forwardConvolutionDescriptor: forward_convolution_descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
ShapeOrTensor::Tensor(output_shape_tensor) => {
msg_send![
self,
convolution3DDataGradientWithIncomingGradientTensor: incoming_gradient_tensor,
weightsTensor: weights_tensor,
outputShapeTensor: output_shape_tensor,
forwardConvolutionDescriptor: forward_convolution_descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
}
}
}
/// Creates a 3-D convolution gradient operation with respect to the
/// *weights* tensor.
///
/// Computes `dL/dW = dL/dR · dR/dW`, analogous to the 2-D variant.
///
/// # Arguments
///
/// * `incoming_gradient_tensor` – Loss gradient with respect to the forward
/// result.
/// * `source_tensor` – Source tensor from the forward pass.
/// * `output_shape` – Forward-pass weights shape (static slice or dynamic
/// tensor).
/// * `forward_convolution_descriptor` – Descriptor used in the forward op.
/// * `name` – Optional debug label.
///
/// # Returns
///
/// A `Retained<Tensor>` containing the weights-gradient result.
pub fn convolution_3d_weights_gradient<'a>(
&self,
incoming_gradient_tensor: &Tensor,
source_tensor: &Tensor,
output_shape: ShapeOrTensor<'a>,
forward_convolution_descriptor: &Convolution3DOpDescriptor,
name: Option<&str>,
) -> Retained<Tensor> {
unsafe {
match output_shape {
ShapeOrTensor::Shape(output_shape) => {
msg_send![
self,
convolution3DWeightsGradientWithIncomingGradientTensor: incoming_gradient_tensor,
sourceTensor: source_tensor,
outputShape: &*ns_number_array_from_slice(output_shape),
forwardConvolutionDescriptor: forward_convolution_descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
ShapeOrTensor::Tensor(output_shape_tensor) => {
msg_send![
self,
convolution3DWeightsGradientWithIncomingGradientTensor: incoming_gradient_tensor,
sourceTensor: source_tensor,
outputShapeTensor: output_shape_tensor,
forwardConvolutionDescriptor: forward_convolution_descriptor,
name: name.map(NSString::from_str).as_deref()
]
}
}
}
}
}