Blame - examples/graph_vgg19.cpp - ml/ComputeLibrary

2017-11-03 17:16:20 +0000

[diff] [blame]

1

/*

Michele Di Giorgio

e3fba0a

2018-02-14 14:18:01 +0000

[diff] [blame]

2

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

3

*

4

* SPDX-License-Identifier: MIT

5

*

6

* Permission is hereby granted, free of charge, to any person obtaining a copy

7

* of this software and associated documentation files (the "Software"), to

8

* deal in the Software without restriction, including without limitation the

9

* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or

10

* sell copies of the Software, and to permit persons to whom the Software is

11

* furnished to do so, subject to the following conditions:

12

*

13

* The above copyright notice and this permission notice shall be included in all

14

* copies or substantial portions of the Software.

15

*

16

* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR

17

* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,

18

* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE

19

* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER

20

* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,

21

* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE

22

* SOFTWARE.

23

*/

Georgios Pinitas

2017-12-22 15:27:52 +0000

[diff] [blame]

24

#include "arm_compute/graph2.h"

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

25

#include "support/ToolchainSupport.h"

26

#include "utils/GraphUtils.h"

27

#include "utils/Utils.h"

#include <cstdlib>

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

31

using namespace arm_compute::utils;

Georgios Pinitas

2017-12-22 15:27:52 +0000

[diff] [blame]

32

using namespace arm_compute::graph2::frontend;

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

33

using namespace arm_compute::graph_utils;

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

34

35

/** Example demonstrating how to implement VGG19's network using the Compute Library's graph API

36

*

37

* @param[in] argc Number of arguments

Gian Marco

bfa3b52

2017-12-12 10:08:38 +0000

[diff] [blame]

38

* @param[in] argv Arguments ( [optional] Target (0 = NEON, 1 = OpenCL), [optional] Path to the weights folder, [optional] image, [optional] labels )

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

39

*/

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

40

class GraphVGG19Example : public Example

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

41

{

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

42

public:

43

void do_setup(int argc, char **argv) override

44

{

45

std::string data_path; /* Path to the trainable data */

46

std::string image; /* Image data */

47

std::string label; /* Label data */

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

48

Georgios Pinitas

140fdc7

2018-02-16 11:42:38 +0000

[diff] [blame]

49

// Create a preprocessor object

50

const std::array<float, 3> mean_rgb{ { 123.68f, 116.779f, 103.939f } };

51

std::unique_ptr<IPreprocessor> preprocessor = arm_compute::support::cpp14::make_unique<CaffePreproccessor>(mean_rgb);

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

52

Michele Di Giorgio

e3fba0a

2018-02-14 14:18:01 +0000

[diff] [blame]

53

// Set target. 0 (NEON), 1 (OpenCL), 2 (OpenCL with Tuner). By default it is NEON

Georgios Pinitas

9a8c672

2018-03-21 17:52:35 +0000

[diff] [blame]

54

const int target = argc > 1 ? std::strtol(argv[1], nullptr, 10) : 0;

55

Target target_hint = set_target_hint2(target);

56

ConvolutionMethod convolution_hint = ConvolutionMethod::DIRECT;

Gian Marco

bfa3b52

2017-12-12 10:08:38 +0000

[diff] [blame]

57

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

// Parse arguments

if(argc < 2)

{

// Print help

std::cout << "Usage: " << argv[0] << " [target] [path_to_data] [image] [labels]\n\n";

63

std::cout << "No data folder provided: using random values\n\n";

}

else if(argc == 2)

{

std::cout << "Usage: " << argv[0] << " " << argv[1] << " [path_to_data] [image] [labels]\n\n";

68

std::cout << "No data folder provided: using random values\n\n";

}

else if(argc == 3)

{

data_path = argv[2];

std::cout << "Usage: " << argv[0] << " " << argv[1] << " " << argv[2] << " [image] [labels]\n\n";

74

std::cout << "No image provided: using random values\n\n";

}

else if(argc == 4)

{

data_path = argv[2];

image = argv[3];

std::cout << "Usage: " << argv[0] << " " << argv[1] << " " << argv[2] << " " << argv[3] << " [labels]\n\n";

81

std::cout << "No text file with labels provided: skipping output accessor\n\n";

}

else

{

data_path = argv[2];

image = argv[3];

label = argv[4];

}

graph << target_hint

<< convolution_hint

Georgios Pinitas

2017-12-22 15:27:52 +0000

[diff] [blame]

92

<< InputLayer(TensorDescriptor(TensorShape(224U, 224U, 3U, 1U), DataType::F32),

93

get_input_accessor(image, std::move(preprocessor)))

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

// Layer 1

<< ConvolutionLayer(

3U, 3U, 64U,

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv1_1_w.npy"),

98

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv1_1_b.npy"),

99

PadStrideInfo(1, 1, 1, 1))

100

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

101

<< ConvolutionLayer(

102

3U, 3U, 64U,

103

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv1_2_w.npy"),

104

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv1_2_b.npy"),

105

PadStrideInfo(1, 1, 1, 1))

106

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

107

<< PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))

// Layer 2

<< ConvolutionLayer(

3U, 3U, 128U,

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv2_1_w.npy"),

112

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv2_1_b.npy"),

113

PadStrideInfo(1, 1, 1, 1))

114

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

115

<< ConvolutionLayer(

116

3U, 3U, 128U,

117

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv2_2_w.npy"),

118

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv2_2_b.npy"),

119

PadStrideInfo(1, 1, 1, 1))

120

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

121

<< PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))

// Layer 3

<< ConvolutionLayer(

3U, 3U, 256U,

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_1_w.npy"),

126

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_1_b.npy"),

127

PadStrideInfo(1, 1, 1, 1))

128

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

129

<< ConvolutionLayer(

130

3U, 3U, 256U,

131

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_2_w.npy"),

132

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_2_b.npy"),

133

PadStrideInfo(1, 1, 1, 1))

134

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

135

<< ConvolutionLayer(

136

3U, 3U, 256U,

137

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_3_w.npy"),

138

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_3_b.npy"),

139

PadStrideInfo(1, 1, 1, 1))

140

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

141

<< ConvolutionLayer(

142

3U, 3U, 256U,

143

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_4_w.npy"),

144

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv3_4_b.npy"),

145

PadStrideInfo(1, 1, 1, 1))

146

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

147

<< PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))

// Layer 4

<< ConvolutionLayer(

3U, 3U, 512U,

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_1_w.npy"),

152

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_1_b.npy"),

153

PadStrideInfo(1, 1, 1, 1))

154

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

155

<< ConvolutionLayer(

156

3U, 3U, 512U,

157

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_2_w.npy"),

158

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_2_b.npy"),

159

PadStrideInfo(1, 1, 1, 1))

160

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

161

<< ConvolutionLayer(

162

3U, 3U, 512U,

163

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_3_w.npy"),

164

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_3_b.npy"),

165

PadStrideInfo(1, 1, 1, 1))

166

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

167

<< ConvolutionLayer(

168

3U, 3U, 512U,

169

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_4_w.npy"),

170

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv4_4_b.npy"),

171

PadStrideInfo(1, 1, 1, 1))

172

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

173

<< PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))

// Layer 5

<< ConvolutionLayer(

3U, 3U, 512U,

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_1_w.npy"),

178

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_1_b.npy"),

179

PadStrideInfo(1, 1, 1, 1))

180

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

181

<< ConvolutionLayer(

182

3U, 3U, 512U,

183

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_2_w.npy"),

184

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_2_b.npy"),

185

PadStrideInfo(1, 1, 1, 1))

186

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

187

<< ConvolutionLayer(

188

3U, 3U, 512U,

189

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_3_w.npy"),

190

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_3_b.npy"),

191

PadStrideInfo(1, 1, 1, 1))

192

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

193

<< ConvolutionLayer(

194

3U, 3U, 512U,

195

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_4_w.npy"),

196

get_weights_accessor(data_path, "/cnn_data/vgg19_model/conv5_4_b.npy"),

197

PadStrideInfo(1, 1, 1, 1))

198

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

199

<< PoolingLayer(PoolingLayerInfo(PoolingType::MAX, 2, PadStrideInfo(2, 2, 0, 0)))

200

// Layer 6

201

<< FullyConnectedLayer(

202

4096U,

203

get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc6_w.npy"),

204

get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc6_b.npy"))

205

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

206

// Layer 7

207

<< FullyConnectedLayer(

208

4096U,

209

get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc7_w.npy"),

210

get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc7_b.npy"))

211

<< ActivationLayer(ActivationLayerInfo(ActivationLayerInfo::ActivationFunction::RELU))

212

// Layer 8

213

<< FullyConnectedLayer(

214

1000U,

215

get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc8_w.npy"),

216

get_weights_accessor(data_path, "/cnn_data/vgg19_model/fc8_b.npy"))

217

// Softmax

218

<< SoftmaxLayer()

Georgios Pinitas

2017-12-22 15:27:52 +0000

[diff] [blame]

219

<< OutputLayer(get_output_accessor(label, 5));

Gian Marco

c1b6e37

2018-02-21 18:03:26 +0000

[diff] [blame]

220

Georgios Pinitas

2017-12-22 15:27:52 +0000

[diff] [blame]

221

// Finalize graph

Georgios Pinitas

9a8c672

2018-03-21 17:52:35 +0000

[diff] [blame]

222

GraphConfig config;

223

config.use_function_memory_manager = true;

224

config.use_tuner = (target == 2);

225

graph.finalize(target_hint, config);

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

226

}

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

227

void do_run() override

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

228

{

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

229

// Run graph

230

graph.run();

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

231

}

232

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

233

private:

Georgios Pinitas

2017-12-22 15:27:52 +0000

[diff] [blame]

234

Stream graph{ 0, "VGG19" };

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

235

};

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

236

237

/** Main program for VGG19

238

*

239

* @param[in] argc Number of arguments

Gian Marco

bfa3b52

2017-12-12 10:08:38 +0000

[diff] [blame]

240

* @param[in] argv Arguments ( [optional] Target (0 = NEON, 1 = OpenCL), [optional] Path to the weights folder, [optional] image, [optional] labels )

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

241

*/

Anthony Barbier

6db0ff5

2018-01-05 10:59:12 +0000

[diff] [blame]

242

int main(int argc, char **argv)

Isabella Gottardi

2017-11-03 17:16:20 +0000

[diff] [blame]

243

{

Michalis Spyrou

2018-01-10 14:08:50 +0000

[diff] [blame]

244

return arm_compute::utils::run_example<GraphVGG19Example>(argc, argv);

Isabella Gottardi