/
CNN1DGradientCheckTest.java
342 lines (280 loc) · 16.1 KB
/
CNN1DGradientCheckTest.java
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
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
package org.deeplearning4j.gradientcheck;
import org.deeplearning4j.BaseDL4JTest;
import org.deeplearning4j.TestUtils;
import org.deeplearning4j.nn.conf.ConvolutionMode;
import org.deeplearning4j.nn.conf.MultiLayerConfiguration;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.distribution.NormalDistribution;
import org.deeplearning4j.nn.conf.inputs.InputType;
import org.deeplearning4j.nn.conf.layers.*;
import org.deeplearning4j.nn.conf.layers.convolutional.Cropping1D;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.nn.weights.WeightInit;
import org.junit.Test;
import org.nd4j.linalg.activations.Activation;
import org.nd4j.linalg.api.buffer.DataBuffer;
import org.nd4j.linalg.api.buffer.util.DataTypeUtil;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.learning.config.NoOp;
import org.nd4j.linalg.lossfunctions.LossFunctions;
import static org.junit.Assert.assertEquals;
import static org.junit.Assert.assertTrue;
public class CNN1DGradientCheckTest extends BaseDL4JTest {
private static final boolean PRINT_RESULTS = true;
private static final boolean RETURN_ON_FIRST_FAILURE = false;
private static final double DEFAULT_EPS = 1e-6;
private static final double DEFAULT_MAX_REL_ERROR = 1e-3;
private static final double DEFAULT_MIN_ABS_ERROR = 1e-8;
static {
Nd4j.setDataType(DataBuffer.Type.DOUBLE);
}
@Test
public void testCnn1DWithLocallyConnected1D() {
Nd4j.getRandom().setSeed(1337);
int[] minibatchSizes = {2, 3};
int length = 7;
int convNIn = 2;
int convNOut1 = 3;
int convNOut2 = 4;
int finalNOut = 4;
int[] kernels = {1};
int stride = 1;
int padding = 0;
Activation[] activations = {Activation.SIGMOID};
for (Activation afn : activations) {
for (int minibatchSize : minibatchSizes) {
for (int kernel : kernels) {
INDArray input = Nd4j.rand(new int[]{minibatchSize, convNIn, length});
INDArray labels = Nd4j.zeros(minibatchSize, finalNOut, length);
for (int i = 0; i < minibatchSize; i++) {
for (int j = 0; j < length; j++) {
labels.putScalar(new int[]{i, i % finalNOut, j}, 1.0);
}
}
MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
.updater(new NoOp()).weightInit(WeightInit.DISTRIBUTION)
.dist(new NormalDistribution(0, 1)).convolutionMode(ConvolutionMode.Same).list()
.layer(new Convolution1DLayer.Builder().activation(afn).kernelSize(kernel)
.stride(stride).padding(padding).nIn(convNIn).nOut(convNOut1)
.build())
.layer(new LocallyConnected1D.Builder().activation(afn).kernelSize(kernel)
.stride(stride).padding(padding).nIn(convNOut1).nOut(convNOut2).hasBias(false)
.build())
.layer(new RnnOutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
.activation(Activation.SOFTMAX).nOut(finalNOut).build())
.setInputType(InputType.recurrent(convNIn, length)).build();
String json = conf.toJson();
MultiLayerConfiguration c2 = MultiLayerConfiguration.fromJson(json);
assertEquals(conf, c2);
MultiLayerNetwork net = new MultiLayerNetwork(conf);
net.init();
String msg = "Minibatch=" + minibatchSize + ", activationFn="
+ afn + ", kernel = " + kernel;
if (PRINT_RESULTS) {
System.out.println(msg);
for (int j = 0; j < net.getnLayers(); j++)
System.out.println("Layer " + j + " # params: " + net.getLayer(j).numParams());
}
boolean gradOK = GradientCheckUtil.checkGradients(net, DEFAULT_EPS, DEFAULT_MAX_REL_ERROR,
DEFAULT_MIN_ABS_ERROR, PRINT_RESULTS, RETURN_ON_FIRST_FAILURE, input, labels);
assertTrue(msg, gradOK);
TestUtils.testModelSerialization(net);
}
}
}
}
@Test
public void testCnn1DWithCropping1D() {
Nd4j.getRandom().setSeed(1337);
int[] minibatchSizes = {1, 3};
int length = 7;
int convNIn = 2;
int convNOut1 = 3;
int convNOut2 = 4;
int finalNOut = 4;
int[] kernels = {1, 2, 4};
int stride = 1;
int padding = 0;
int cropping = 1;
int croppedLength = length - 2 * cropping;
Activation[] activations = {Activation.SIGMOID};
SubsamplingLayer.PoolingType[] poolingTypes =
new SubsamplingLayer.PoolingType[]{SubsamplingLayer.PoolingType.MAX,
SubsamplingLayer.PoolingType.AVG, SubsamplingLayer.PoolingType.PNORM};
for (Activation afn : activations) {
for (SubsamplingLayer.PoolingType poolingType : poolingTypes) {
for (int minibatchSize : minibatchSizes) {
for (int kernel : kernels) {
INDArray input = Nd4j.rand(new int[]{minibatchSize, convNIn, length});
INDArray labels = Nd4j.zeros(minibatchSize, finalNOut, croppedLength);
for (int i = 0; i < minibatchSize; i++) {
for (int j = 0; j < croppedLength; j++) {
labels.putScalar(new int[]{i, i % finalNOut, j}, 1.0);
}
}
MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
.updater(new NoOp()).weightInit(WeightInit.DISTRIBUTION)
.dist(new NormalDistribution(0, 1)).convolutionMode(ConvolutionMode.Same).list()
.layer(new Convolution1DLayer.Builder().activation(afn).kernelSize(kernel)
.stride(stride).padding(padding).nIn(convNIn).nOut(convNOut1)
.build())
.layer(new Cropping1D.Builder(cropping).build())
.layer(new Convolution1DLayer.Builder().activation(afn).kernelSize(kernel)
.stride(stride).padding(padding).nIn(convNOut1).nOut(convNOut2)
.build())
.layer(new RnnOutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
.activation(Activation.SOFTMAX).nOut(finalNOut).build())
.setInputType(InputType.recurrent(convNIn, length)).build();
String json = conf.toJson();
MultiLayerConfiguration c2 = MultiLayerConfiguration.fromJson(json);
assertEquals(conf, c2);
MultiLayerNetwork net = new MultiLayerNetwork(conf);
net.init();
String msg = "PoolingType=" + poolingType + ", minibatch=" + minibatchSize + ", activationFn="
+ afn + ", kernel = " + kernel;
if (PRINT_RESULTS) {
System.out.println(msg);
for (int j = 0; j < net.getnLayers(); j++)
System.out.println("Layer " + j + " # params: " + net.getLayer(j).numParams());
}
boolean gradOK = GradientCheckUtil.checkGradients(net, DEFAULT_EPS, DEFAULT_MAX_REL_ERROR,
DEFAULT_MIN_ABS_ERROR, PRINT_RESULTS, RETURN_ON_FIRST_FAILURE, input, labels);
assertTrue(msg, gradOK);
TestUtils.testModelSerialization(net);
}
}
}
}
}
@Test
public void testCnn1DWithZeroPadding1D() {
Nd4j.getRandom().setSeed(1337);
int[] minibatchSizes = {1, 3};
int length = 7;
int convNIn = 2;
int convNOut1 = 3;
int convNOut2 = 4;
int finalNOut = 4;
int[] kernels = {1, 2, 4};
int stride = 1;
int pnorm = 2;
int padding = 0;
int zeroPadding = 2;
int paddedLength = length + 2 * zeroPadding;
Activation[] activations = {Activation.SIGMOID};
SubsamplingLayer.PoolingType[] poolingTypes =
new SubsamplingLayer.PoolingType[]{SubsamplingLayer.PoolingType.MAX,
SubsamplingLayer.PoolingType.AVG, SubsamplingLayer.PoolingType.PNORM};
for (Activation afn : activations) {
for (SubsamplingLayer.PoolingType poolingType : poolingTypes) {
for (int minibatchSize : minibatchSizes) {
for (int kernel : kernels) {
INDArray input = Nd4j.rand(new int[]{minibatchSize, convNIn, length});
INDArray labels = Nd4j.zeros(minibatchSize, finalNOut, paddedLength);
for (int i = 0; i < minibatchSize; i++) {
for (int j = 0; j < paddedLength; j++) {
labels.putScalar(new int[]{i, i % finalNOut, j}, 1.0);
}
}
MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
.updater(new NoOp()).weightInit(WeightInit.DISTRIBUTION)
.dist(new NormalDistribution(0, 1)).convolutionMode(ConvolutionMode.Same).list()
.layer(new Convolution1DLayer.Builder().activation(afn).kernelSize(kernel)
.stride(stride).padding(padding).nIn(convNIn).nOut(convNOut1)
.build())
.layer(new ZeroPadding1DLayer.Builder(zeroPadding).build())
.layer(new Convolution1DLayer.Builder().activation(afn).kernelSize(kernel)
.stride(stride).padding(padding).nIn(convNOut1).nOut(convNOut2)
.build())
.layer(new ZeroPadding1DLayer.Builder(0).build())
.layer(new Subsampling1DLayer.Builder(poolingType).kernelSize(kernel)
.stride(stride).padding(padding).pnorm(pnorm).build())
.layer(new RnnOutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
.activation(Activation.SOFTMAX).nOut(finalNOut).build())
.setInputType(InputType.recurrent(convNIn, length)).build();
String json = conf.toJson();
MultiLayerConfiguration c2 = MultiLayerConfiguration.fromJson(json);
assertEquals(conf, c2);
MultiLayerNetwork net = new MultiLayerNetwork(conf);
net.init();
String msg = "PoolingType=" + poolingType + ", minibatch=" + minibatchSize + ", activationFn="
+ afn + ", kernel = " + kernel;
if (PRINT_RESULTS) {
System.out.println(msg);
for (int j = 0; j < net.getnLayers(); j++)
System.out.println("Layer " + j + " # params: " + net.getLayer(j).numParams());
}
boolean gradOK = GradientCheckUtil.checkGradients(net, DEFAULT_EPS, DEFAULT_MAX_REL_ERROR,
DEFAULT_MIN_ABS_ERROR, PRINT_RESULTS, RETURN_ON_FIRST_FAILURE, input, labels);
assertTrue(msg, gradOK);
TestUtils.testModelSerialization(net);
}
}
}
}
}
@Test
public void testCnn1DWithSubsampling1D() {
Nd4j.getRandom().setSeed(12345);
int[] minibatchSizes = {1, 3};
int length = 7;
int convNIn = 2;
int convNOut1 = 3;
int convNOut2 = 4;
int finalNOut = 4;
int[] kernels = {1, 2, 4};
int stride = 1;
int padding = 0;
int pnorm = 2;
Activation[] activations = {Activation.SIGMOID, Activation.TANH};
SubsamplingLayer.PoolingType[] poolingTypes =
new SubsamplingLayer.PoolingType[]{SubsamplingLayer.PoolingType.MAX,
SubsamplingLayer.PoolingType.AVG, SubsamplingLayer.PoolingType.PNORM};
for (Activation afn : activations) {
for (SubsamplingLayer.PoolingType poolingType : poolingTypes) {
for (int minibatchSize : minibatchSizes) {
for (int kernel : kernels) {
INDArray input = Nd4j.rand(new int[]{minibatchSize, convNIn, length});
INDArray labels = Nd4j.zeros(minibatchSize, finalNOut, length);
for (int i = 0; i < minibatchSize; i++) {
for (int j = 0; j < length; j++) {
labels.putScalar(new int[]{i, i % finalNOut, j}, 1.0);
}
}
MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
.updater(new NoOp()).weightInit(WeightInit.DISTRIBUTION)
.dist(new NormalDistribution(0, 1)).convolutionMode(ConvolutionMode.Same).list()
.layer(0, new Convolution1DLayer.Builder().activation(afn).kernelSize(kernel)
.stride(stride).padding(padding).nIn(convNIn).nOut(convNOut1)
.build())
.layer(1, new Convolution1DLayer.Builder().activation(afn).kernelSize(kernel)
.stride(stride).padding(padding).nIn(convNOut1).nOut(convNOut2)
.build())
.layer(2, new Subsampling1DLayer.Builder(poolingType).kernelSize(kernel)
.stride(stride).padding(padding).pnorm(pnorm).build())
.layer(3, new RnnOutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
.activation(Activation.SOFTMAX).nOut(finalNOut).build())
.setInputType(InputType.recurrent(convNIn, length)).build();
String json = conf.toJson();
MultiLayerConfiguration c2 = MultiLayerConfiguration.fromJson(json);
assertEquals(conf, c2);
MultiLayerNetwork net = new MultiLayerNetwork(conf);
net.init();
String msg = "PoolingType=" + poolingType + ", minibatch=" + minibatchSize + ", activationFn="
+ afn + ", kernel = " + kernel;
if (PRINT_RESULTS) {
System.out.println(msg);
for (int j = 0; j < net.getnLayers(); j++)
System.out.println("Layer " + j + " # params: " + net.getLayer(j).numParams());
}
boolean gradOK = GradientCheckUtil.checkGradients(net, DEFAULT_EPS, DEFAULT_MAX_REL_ERROR,
DEFAULT_MIN_ABS_ERROR, PRINT_RESULTS, RETURN_ON_FIRST_FAILURE, input, labels);
assertTrue(msg, gradOK);
TestUtils.testModelSerialization(net);
}
}
}
}
}
}