tiny_dnn 1.0.0
A header only, dependency-free deep learning framework in C++11
Loading...
Searching...
No Matches
op_kernel.h
1/*
2 COPYRIGHT
3
4 All contributions by Taiga Nomi
5 Copyright (c) 2013, Taiga Nomi
6 All rights reserved.
7
8 All other contributions:
9 Copyright (c) 2013-2016, the respective contributors.
10 All rights reserved.
11
12 Each contributor holds copyright over their respective contributions.
13 The project versioning (Git) records all such contribution source information.
14
15 LICENSE
16
17 The BSD 3-Clause License
18
19
20 Redistribution and use in source and binary forms, with or without
21 modification, are permitted provided that the following conditions are met:
22
23 * Redistributions of source code must retain the above copyright notice, this
24 list of conditions and the following disclaimer.
25
26 * Redistributions in binary form must reproduce the above copyright notice,
27 this list of conditions and the following disclaimer in the documentation
28 and/or other materials provided with the distribution.
29
30 * Neither the name of tiny-dnn nor the names of its
31 contributors may be used to endorse or promote products derived from
32 this software without specific prior written permission.
33
34 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
35 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
36 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
37 DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
38 FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
39 DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
40 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
41 CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
42 OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
43 OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
44*/
45#pragma once
46
47#include "tiny_dnn/core/framework/device.fwd.h"
48#include "tiny_dnn/core/params/conv_params.h"
49
50namespace tiny_dnn {
51namespace core {
52
53class OpKernel; // delared below
54
56 public:
57 explicit OpKernelConstruction() {}
58 explicit OpKernelConstruction(Device* device, Params* params)
59 : device_(device), params_(params) {}
60
61 // Returns the device raw pointer
62 Device* device() const { return device_; }
63
64 // Returns the device raw pointer
65 Params* params() const { return params_; }
66
67 private:
68 Device* device_ = nullptr;
69 Params* params_ = nullptr;
70};
71
73 public:
74 struct OpParams {
75 // the op kernel being computed.
76 OpKernel* op_kernel_ptr = nullptr;
77
78 // the device on which the kernel is running.
79 Device* device_ptr = nullptr;
80
81 // the layer on which kernel is runnning
82 layer* layer_ptr_ = nullptr;
83
84 // the operation params
85 Params* params_ptr_ = nullptr;
86
87 // parallelize operation
88 bool parallelize = false;
89
90 backend_t engine = default_engine();
91 };
92
93 explicit OpKernelContext(const std::vector<tensor_t*>& in_data,
94 std::vector<tensor_t*>& out_data)
95 : in_data_(in_data), out_data_(out_data) {
96 op_params_ = std::unique_ptr<OpParams>(new OpParams());
97 }
98
99 explicit OpKernelContext(const std::vector<tensor_t*>& in_data,
100 const std::vector<tensor_t*>& out_data,
101 std::vector<tensor_t*>& out_grad,
102 std::vector<tensor_t*>& in_grad)
103 : in_data_(in_data)
104 , out_data_(out_data)
105 , out_grad_(out_grad)
106 , in_grad_(in_grad) {
107 op_params_ = std::unique_ptr<OpParams>(new OpParams());
108 }
109
110 tensor_t& input(const int idx) const {
111 return *in_data_[idx];
112 }
113
114 tensor_t& output(const int idx) const {
115 return *out_data_[idx];
116 }
117
118 tensor_t& input_grad(const int idx) const {
119 return *in_grad_[idx];
120 }
121
122 tensor_t& output_grad(const int idx) const {
123 return *out_grad_[idx];
124 }
125
126 void setParams(Params* params) {
127 op_params_->params_ptr_ = params;
128 }
129
130 Params* params() const {
131 return op_params_->params_ptr_;
132 }
133
134 void setParallelize(const bool parallelize) {
135 op_params_->parallelize = parallelize;
136 }
137
138 bool parallelize() const {
139 return op_params_->parallelize;
140 }
141
142 void setDevice(Device* device) {
143 op_params_->device_ptr = device;
144 }
145
146 Device* device() const {
147 return op_params_->device_ptr;
148 }
149
150 void setLayer(layer* layer) {
151 op_params_->layer_ptr_ = layer;
152 }
153
154 layer* Layer() const {
155 return op_params_->layer_ptr_;
156 }
157
158 backend_t engine() const {
159 return op_params_->engine;
160 }
161
162 void setEngine(const backend_t engine) {
163 op_params_->engine = engine;
164 }
165
166 private:
167 std::vector<tensor_t*> in_data_;
168 std::vector<tensor_t*> out_data_;
169 std::vector<tensor_t*> out_grad_;
170 std::vector<tensor_t*> in_grad_;
171
172 std::unique_ptr<OpParams> op_params_;
173};
174
175class OpKernel {
176 public:
177 explicit OpKernel() {}
178 explicit OpKernel(const OpKernelConstruction& context)
179 : device_(context.device())
180 , params_(context.params()) {}
181
182 virtual ~OpKernel() {}
183
184 virtual void compute(const OpKernelContext& context) = 0;
185
186 protected:
187 Device* device_ = nullptr;
188 Params* params_ = nullptr;
189};
190
191} // namespace core
192} // namespace tiny_dnn
Definition device.fwd.h:73
Definition op_kernel.h:55
Definition op_kernel.h:72
Definition op_kernel.h:175
Definition params.h:37
Simple image utility class.
Definition image.h:94
base class of all kind of NN layers
Definition layer.h:62