How to use kernel method in autotest
Best Python code snippet using autotest_python
multi_lead_branch_fusion.py
Source:multi_lead_branch_fusion.py 
1# -*- coding: utf-8 -*-2"""Multi Lead Branch Fusion3Automatically generated by Colaboratory.4Original file is located at5 https://colab.research.google.com/drive/1RfMexPUadW9T6txCWITOdRUBhsP-DOE86"""7input = Input(shape=(30000,6), dtype='float32', name= 'input')8x = Conv1D(12,kernel_size=3,strides=1, padding='same')(input)9x = LeakyReLU(alpha=0.3)(x)10x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)11x = LeakyReLU(alpha=0.3)(x)12x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)13x = LeakyReLU(alpha=0.3)(x)14x = Dropout(0.2)(x)15x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)16x = LeakyReLU(alpha=0.3)(x)17x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)18x = LeakyReLU(alpha=0.3)(x)19x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)20x = LeakyReLU(alpha=0.3)(x)21x = Dropout(0.2)(x)22x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)23x = LeakyReLU(alpha=0.3)(x)24x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)25x = LeakyReLU(alpha=0.3)(x)26x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)27x = LeakyReLU(alpha=0.3)(x)28x = Dropout(0.2)(x) 29x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)30x = LeakyReLU(alpha=0.3)(x)31x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)32x = LeakyReLU(alpha=0.3)(x)33x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)34x = LeakyReLU(alpha=0.3)(x)35x = Dropout(0.2)(x)36x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)37x = LeakyReLU(alpha=0.3)(x)38x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)39x = LeakyReLU(alpha=0.3)(x)40x = Conv1D(12,kernel_size=48, strides = 2, padding='same')(x)41x = LeakyReLU(alpha=0.3)(x)42x = Dropout(0.2)(x)43x1= Bidirectional(GRU(12, input_shape=(938,6),return_sequences=True,return_state=False),merge_mode='concat')(x)44x = LeakyReLU(alpha=0.3)(x1)45x = Dropout(0.2)(x)46x = AttentionWithContext()(x)47x = BatchNormalization()(x)48x = LeakyReLU(alpha=0.3)(x)49x = Dropout(0.2)(x)50x = Dense(7,activation='sigmoid')(x)51 52 #SECOND MODEL53x = Conv1D(12,kernel_size=3,strides=1, padding='same')(input)54x = LeakyReLU(alpha=0.3)(x)55x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)56x = LeakyReLU(alpha=0.3)(x)57x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)58x = LeakyReLU(alpha=0.3)(x)59x = Dropout(0.2)(x)60x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)61x = LeakyReLU(alpha=0.3)(x)62x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)63x = LeakyReLU(alpha=0.3)(x)64x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)65x = LeakyReLU(alpha=0.3)(x)66x = Dropout(0.5)(x)67x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)68x = LeakyReLU(alpha=0.3)(x)69x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)70x = LeakyReLU(alpha=0.3)(x)71x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)72x = LeakyReLU(alpha=0.3)(x)73x = Dropout(0.2)(x) 74x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)75x = LeakyReLU(alpha=0.3)(x)76x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)77x = LeakyReLU(alpha=0.3)(x)78x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)79x = LeakyReLU(alpha=0.3)(x)80x = Dropout(0.2)(x)81x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)82x = LeakyReLU(alpha=0.3)(x)83x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)84x = LeakyReLU(alpha=0.3)(x)85x = Conv1D(12,kernel_size=48, strides = 2, padding='same')(x)86x = LeakyReLU(alpha=0.3)(x)87x = Dropout(0.2)(x)88x2 = Bidirectional(GRU(12, input_shape=(938,6),return_sequences=True,return_state=False),merge_mode='concat')(x)89x = LeakyReLU(alpha=0.3)(x2)90x = Dropout(0.2)(x)91x = AttentionWithContext()(x)92x = BatchNormalization()(x)93x = LeakyReLU(alpha=0.3)(x)94x = Dropout(0.2)(x)95x = Dense(7,activation='sigmoid')(x)96 #THIRD MODEL97x = Conv1D(12,kernel_size=3,strides=1, padding='same')(input)98x = LeakyReLU(alpha=0.3)(x)99x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)100x = LeakyReLU(alpha=0.3)(x)101x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)102x = LeakyReLU(alpha=0.3)(x)103x = Dropout(0.2)(x)104x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)105x = LeakyReLU(alpha=0.3)(x)106x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)107x = LeakyReLU(alpha=0.3)(x)108x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)109x = LeakyReLU(alpha=0.3)(x)110x = Dropout(0.2)(x)111x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)112x = LeakyReLU(alpha=0.3)(x)113x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)114x = LeakyReLU(alpha=0.3)(x)115x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)116x = LeakyReLU(alpha=0.3)(x)117x = Dropout(0.2)(x) 118x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)119x = LeakyReLU(alpha=0.3)(x)120x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)121x = LeakyReLU(alpha=0.3)(x)122x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)123x = LeakyReLU(alpha=0.3)(x)124x = Dropout(0.2)(x)125x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)126x = LeakyReLU(alpha=0.3)(x)127x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)128x = LeakyReLU(alpha=0.3)(x)129x = Conv1D(12,kernel_size=48, strides = 2, padding='same')(x)130x = LeakyReLU(alpha=0.3)(x)131x= Dropout(0.2)(x)132x3= Bidirectional(GRU(12, input_shape=(938,6),return_sequences=True,return_state=False),merge_mode='concat')(x)133x = LeakyReLU(alpha=0.3)(x3)134x = Dropout(0.2)(x)135x = AttentionWithContext()(x)136x = BatchNormalization()(x)137x = LeakyReLU(alpha=0.3)(x)138x = Dropout(0.2)(x)139x = Dense(7,activation='sigmoid')(x)140 #MODEL 4141x = Conv1D(12,kernel_size=3,strides=1, padding='same')(input)142x = LeakyReLU(alpha=0.3)(x)143x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)144x = LeakyReLU(alpha=0.3)(x)145x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)146x = LeakyReLU(alpha=0.3)(x)147x = Dropout(0.2)(x)148x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)149x = LeakyReLU(alpha=0.3)(x)150x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)151x = LeakyReLU(alpha=0.3)(x)152x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)153x = LeakyReLU(alpha=0.3)(x)154x = Dropout(0.2)(x)155x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)156x = LeakyReLU(alpha=0.3)(x)157x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)158x = LeakyReLU(alpha=0.3)(x)159x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)160x = LeakyReLU(alpha=0.3)(x)161x = Dropout(0.2)(x) 162x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)163x = LeakyReLU(alpha=0.3)(x)164x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)165x = LeakyReLU(alpha=0.3)(x)166x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)167x = LeakyReLU(alpha=0.3)(x)168x = Dropout(0.2)(x)169x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)170x = LeakyReLU(alpha=0.3)(x)171x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)172x = LeakyReLU(alpha=0.3)(x)173x = Conv1D(12,kernel_size=48, strides = 2, padding='same')(x)174x = LeakyReLU(alpha=0.3)(x)175x= Dropout(0.2)(x)176x4= Bidirectional(GRU(12, input_shape=(938,6),return_sequences=True,return_state=False),merge_mode='concat')(x)177x = LeakyReLU(alpha=0.3)(x4)178x = Dropout(0.2)(x)179x = AttentionWithContext()(x)180x = BatchNormalization()(x)181x = LeakyReLU(alpha=0.3)(x)182x = Dropout(0.2)(x)183x = Dense(7,activation='sigmoid')(x)184 #MODEL 5185x = Conv1D(12,kernel_size=3,strides=1, padding='same')(input)186x = LeakyReLU(alpha=0.3)(x)187x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)188x = LeakyReLU(alpha=0.3)(x)189x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)190x = LeakyReLU(alpha=0.3)(x)191x = Dropout(0.2)(x)192x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)193x = LeakyReLU(alpha=0.3)(x)194x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)195x = LeakyReLU(alpha=0.3)(x)196x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)197x = LeakyReLU(alpha=0.3)(x)198x = Dropout(0.2)(x)199x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)200x = LeakyReLU(alpha=0.3)(x)201x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)202x = LeakyReLU(alpha=0.3)(x)203x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)204x = LeakyReLU(alpha=0.3)(x)205x = Dropout(0.2)(x) 206x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)207x = LeakyReLU(alpha=0.3)(x)208x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)209x = LeakyReLU(alpha=0.3)(x)210x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)211x = LeakyReLU(alpha=0.3)(x)212x = Dropout(0.2)(x)213x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)214x = LeakyReLU(alpha=0.3)(x)215x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)216x = LeakyReLU(alpha=0.3)(x)217x = Conv1D(12,kernel_size=48, strides = 2, padding='same')(x)218x = LeakyReLU(alpha=0.3)(x)219x = Dropout(0.2)(x)220x5= Bidirectional(GRU(12, input_shape=(938,6),return_sequences=True,return_state=False),merge_mode='concat')(x)221x = LeakyReLU(alpha=0.3)(x5)222x = Dropout(0.2)(x)223x = AttentionWithContext()(x)224x = BatchNormalization()(x)225x = LeakyReLU(alpha=0.3)(x)226x = Dropout(0.2)(x)227x= Dense(7,activation='sigmoid')(x)228 229 230 #MODEL 6231x = Conv1D(12,kernel_size=3,strides=1, padding='same')(input)232x = LeakyReLU(alpha=0.3)(x)233x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)234x = LeakyReLU(alpha=0.3)(x)235x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)236x = LeakyReLU(alpha=0.3)(x)237x = Dropout(0.2)(x)238x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)239x = LeakyReLU(alpha=0.3)(x)240x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)241x = LeakyReLU(alpha=0.3)(x)242x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)243x = LeakyReLU(alpha=0.3)(x)244x = Dropout(0.2)(x)245x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)246x = LeakyReLU(alpha=0.3)(x)247x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)248x = LeakyReLU(alpha=0.3)(x)249x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)250x = LeakyReLU(alpha=0.3)(x)251x = Dropout(0.2)(x) 252x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)253x = LeakyReLU(alpha=0.3)(x)254x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)255x = LeakyReLU(alpha=0.3)(x)256x = Conv1D(12,kernel_size=24, strides = 2, padding='same')(x)257x = LeakyReLU(alpha=0.3)(x)258x = Dropout(0.2)(x)259x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)260x = LeakyReLU(alpha=0.3)(x)261x = Conv1D(12,kernel_size=3,strides=1, padding='same')(x)262x = LeakyReLU(alpha=0.3)(x)263x = Conv1D(12,kernel_size=48, strides = 2, padding='same')(x)264x = LeakyReLU(alpha=0.3)(x)265x = Dropout(0.2)(x)266x6= Bidirectional(GRU(12, input_shape=(938,6),return_sequences=True,return_state=False),merge_mode='concat')(x)267x = LeakyReLU(alpha=0.3)(x6)268x = Dropout(0.2)(x)269x = AttentionWithContext()(x)270x = BatchNormalization()(x)271x = LeakyReLU(alpha=0.3)(x)272x = Dropout(0.2)(x)273x= Dense(7,activation='sigmoid')(x)274 #MERGE INPUT MODELS275z=concatenate([x1,x2,x3,x4,x5,x6])276print(z.shape)277#FINAL ATTENTIOON MODULE278x = AttentionWithContext()(z)279x = BatchNormalization()(x)280x = LeakyReLU(alpha=0.3)(x)281x = Dropout(0.2)(x)282x = Dense(50,activation='sigmoid')(x)283output = Dense(7,activation='sigmoid')(x)...
blocks_masked_conv2d_test.py
Source:blocks_masked_conv2d_test.py
1# Copyright 2017 The TensorFlow Authors All Rights Reserved.2#3# Licensed under the Apache License, Version 2.0 (the "License");4# you may not use this file except in compliance with the License.5# You may obtain a copy of the License at6#7# http://www.apache.org/licenses/LICENSE-2.08#9# Unless required by applicable law or agreed to in writing, software10# distributed under the License is distributed on an "AS IS" BASIS,11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.12# See the License for the specific language governing permissions and13# limitations under the License.14# ==============================================================================15"""Tests of the 2D masked convolution blocks."""16from __future__ import division17from __future__ import unicode_literals18import numpy as np19from six.moves import xrange20import tensorflow as tf21import blocks_masked_conv2d22class MaskedConv2DTest(tf.test.TestCase):23 def testRasterScanKernel(self):24 kernel_size = 525 input_depth = 126 output_depth = 127 kernel_shape = [kernel_size, kernel_size, input_depth, output_depth]28 # pylint: disable=bad-whitespace29 kernel_feed = [[ 1.0, 2.0, 3.0, 4.0, 5.0],30 [ 6.0, 7.0, 8.0, 9.0, 10.0],31 [11.0, 12.0, 13.0, 14.0, 15.0],32 [16.0, 17.0, 18.0, 19.0, 20.0],33 [21.0, 22.0, 23.0, 24.0, 25.0]]34 kernel_feed = np.reshape(kernel_feed, kernel_shape)35 kernel_expected = [[ 1.0, 2.0, 3.0, 4.0, 5.0],36 [ 6.0, 7.0, 8.0, 9.0, 10.0],37 [11.0, 12.0, 0.0, 0.0, 0.0],38 [ 0.0, 0.0, 0.0, 0.0, 0.0],39 [ 0.0, 0.0, 0.0, 0.0, 0.0]]40 kernel_expected = np.reshape(kernel_expected, kernel_shape)41 # pylint: enable=bad-whitespace42 init_kernel = lambda s, t: tf.constant(kernel_feed, dtype=t, shape=s)43 masked_conv2d = blocks_masked_conv2d.RasterScanConv2D(44 output_depth, [kernel_size] * 2, [1] * 2, 'SAME',45 initializer=init_kernel)46 x = tf.placeholder(dtype=tf.float32, shape=[10] * 3 + [input_depth])47 _ = masked_conv2d(x)48 with self.test_session():49 tf.global_variables_initializer().run()50 kernel_value = masked_conv2d._kernel.eval()51 self.assertAllEqual(kernel_expected, kernel_value)52 def testDepthOrderKernel(self):53 kernel_size = 154 input_depth = 755 output_depth = input_depth56 kernel_shape = [kernel_size, kernel_size, input_depth, output_depth]57 kernel_feed = np.ones(kernel_shape)58 x_shape = [5] * 3 + [input_depth]59 x_feed = np.ones(x_shape)60 y_expected = np.zeros(x_shape[0:3] + [output_depth])61 y_expected[:, :, :] = np.arange(output_depth)62 init_kernel = lambda s, t: tf.constant(kernel_feed, dtype=t, shape=s)63 masked_conv2d = blocks_masked_conv2d.DepthOrderConv2D(64 output_depth, [kernel_size] * 2, [1] * 2, 'SAME',65 strict_order=True,66 initializer=init_kernel)67 x = tf.placeholder(dtype=tf.float32, shape=x_shape)68 y = masked_conv2d(x)69 with self.test_session():70 tf.global_variables_initializer().run()71 y_value = y.eval(feed_dict={x: x_feed})72 self.assertAllEqual(y_expected, y_value)73 def testGroupRasterScanKernel(self):74 kernel_size = 375 input_depth = 476 input_group_size = 277 output_depth = 278 output_group_size = 179 kernel_shape = [kernel_size, kernel_size, input_depth, output_depth]80 kernel_feed = np.ones(shape=kernel_shape)81 height = 582 width = 583 x_shape = [1, height, width, input_depth]84 x_feed = np.ones(shape=x_shape)85 # pylint: disable=bad-whitespace86 y_expected = [87 [[ 0, 2], [ 4, 6], [ 4, 6], [ 4, 6], [ 4, 6]],88 [[ 8, 10], [16, 18], [16, 18], [16, 18], [12, 14]],89 [[ 8, 10], [16, 18], [16, 18], [16, 18], [12, 14]],90 [[ 8, 10], [16, 18], [16, 18], [16, 18], [12, 14]],91 [[ 8, 10], [16, 18], [16, 18], [16, 18], [12, 14]],92 ]93 y_expected = np.reshape(y_expected, [1, height, width, output_depth])94 # pylint: enable=bad-whitespace95 init_kernel = lambda s, t: tf.constant(kernel_feed, dtype=t, shape=s)96 masked_conv2d = blocks_masked_conv2d.GroupRasterScanConv2D(97 output_depth, [kernel_size] * 2, [1] * 2, 'SAME',98 strict_order=True,99 input_group_size=input_group_size,100 output_group_size=output_group_size,101 initializer=init_kernel)102 x = tf.placeholder(dtype=tf.float32, shape=x_shape)103 y = masked_conv2d(x)104 with self.test_session():105 tf.global_variables_initializer().run()106 y_value = y.eval(feed_dict={x: x_feed})107 self.assertAllEqual(y_expected, y_value)108 def testInFillingKernel(self):109 kernel_size = 5110 input_depth = 1111 output_depth = 1112 kernel_shape = [kernel_size, kernel_size, input_depth, output_depth]113 # pylint: disable=bad-whitespace114 kernel_feed = [[ 1.0, 2.0, 3.0, 4.0, 5.0],115 [ 6.0, 7.0, 8.0, 9.0, 10.0],116 [11.0, 12.0, 13.0, 14.0, 15.0],117 [16.0, 17.0, 18.0, 19.0, 20.0],118 [21.0, 22.0, 23.0, 24.0, 25.0]]119 kernel_feed = np.reshape(kernel_feed, kernel_shape)120 kernel_expected = [[ 1.0, 2.0, 3.0, 4.0, 5.0],121 [ 6.0, 7.0, 8.0, 9.0, 10.0],122 [11.0, 12.0, 0.0, 14.0, 15.0],123 [16.0, 17.0, 18.0, 19.0, 20.0],124 [21.0, 22.0, 23.0, 24.0, 25.0]]125 kernel_expected = np.reshape(kernel_expected, kernel_shape)126 # pylint: enable=bad-whitespace127 init_kernel = lambda s, t: tf.constant(kernel_feed, dtype=t, shape=s)128 masked_conv2d = blocks_masked_conv2d.InFillingConv2D(129 output_depth, [kernel_size] * 2, [1] * 2, 'SAME',130 initializer=init_kernel)131 x = tf.placeholder(dtype=tf.float32, shape=[10] * 3 + [input_depth])132 _ = masked_conv2d(x)133 with self.test_session():134 tf.global_variables_initializer().run()135 kernel_value = masked_conv2d._kernel.eval()136 self.assertAllEqual(kernel_expected, kernel_value)137 def testConv2DMaskedNumerics(self):138 kernel_size = 5139 input_shape = [1, 10, 10, 1]140 filter_shape = [kernel_size, kernel_size, 1, 1]141 strides = [1, 1, 1, 1]142 output_shape = [1, 10, 10, 1]143 conv = blocks_masked_conv2d.RasterScanConv2D(144 depth=filter_shape[-1],145 filter_size=filter_shape[0:2],146 strides=strides[1:3],147 padding='SAME',148 initializer=tf.constant_initializer(value=1.0))149 x = tf.placeholder(dtype=tf.float32, shape=input_shape)150 y = conv(x)151 x_feed = - np.ones(input_shape, dtype=float)152 y_expected = np.ones(output_shape, dtype=float)153 for i in xrange(input_shape[1]):154 for j in xrange(input_shape[2]):155 x_feed[0, i, j, 0] = 10 * (j + 1) + i156 v = 0157 ki_start = max(i - kernel_size // 2, 0)158 kj_start = max(j - kernel_size // 2, 0)159 kj_end = min(j + kernel_size // 2, input_shape[2] - 1)160 for ki in range(ki_start, i + 1):161 for kj in range(kj_start, kj_end + 1):162 if ki > i:163 continue164 if ki == i and kj >= j:165 continue166 v += 10 * (kj + 1) + ki167 y_expected[0, i, j, 0] = v168 with self.test_session():169 tf.global_variables_initializer().run()170 y_value = y.eval(feed_dict={x: x_feed})171 self.assertAllEqual(y_expected, y_value)172if __name__ == '__main__':...
models.py
Source:models.py
1from keras.layers import Dense, Input, LSTM, Dropout2from keras.models import Model, Sequential3from keras.initializers import Constant, RandomNormal, RandomUniform4from keras.optimizers import Adam5from keras.regularizers import l1, l26def create_model_stateless(seq_len, seq_dim, kernel_initializer=RandomUniform(), kernel_regularizer=None,7 hidden_units=50, hidden_layers=1, activation="tanh", regularization=0.01):8 x = Input(shape=(seq_len, seq_dim))9 if hidden_layers == 1:10 h = LSTM(hidden_units, return_sequences = False, activation=activation, kernel_initializer=kernel_initializer,11 kernel_regularizer=kernel_regularizer)(x)12 else:13 h = LSTM(hidden_units, return_sequences = True, activation=activation, kernel_initializer=kernel_initializer,14 kernel_regularizer=kernel_regularizer)(x)15 for i in range(1, hidden_layers):16 if i == hidden_layers - 1:17 h = LSTM(hidden_units, return_sequences = False, activation=activation, kernel_initializer=kernel_initializer,18 kernel_regularizer=kernel_regularizer)(h)19 else:20 h = LSTM(hidden_units, return_sequences = True, activation=activation, kernel_initializer=kernel_initializer,21 kernel_regularizer=kernel_regularizer)(h)22 y = Dense(seq_dim, activation = "linear", kernel_initializer=kernel_initializer, kernel_regularizer=None)(h)23 model = Model(x, y)24 model.compile(optimizer="adam", loss="mean_squared_error")25 return model26def create_model_stateful(batch_size, seq_len, seq_dim, kernel_initializer=RandomUniform(), kernel_regularizer=None,27 hidden_units=10, hidden_layers=1, activation="relu"):28 x = Input(batch_shape=(batch_size, seq_len, seq_dim))29 if hidden_layers ==1:30 h = LSTM(hidden_units, activation=activation, stateful=True, return_sequences=False,31 kernel_regularizer=kernel_regularizer, kernel_initializer=kernel_initializer)(x)32 else:33 h = LSTM(hidden_units, activation=activation, stateful=True, return_sequences=True,34 kernel_regularizer=kernel_regularizer, kernel_initializer=kernel_initializer)(x)35 for i in range(1, hidden_layers):36 if i == hidden_layers - 1:37 h = LSTM(hidden_units, activation=activation, stateful=True, return_sequences=False,38 kernel_regularizer=kernel_regularizer, kernel_initializer=kernel_initializer)(h)39 else:40 h = LSTM(hidden_units, activation=activation, stateful=True, return_sequences=True,41 kernel_regularizer=kernel_regularizer, kernel_initializer=kernel_initializer)(h)42 y = Dense(seq_dim, activation="linear", kernel_initializer=kernel_initializer)(h)43 model = Model(x, y)44 model.compile(optimizer="adam", loss="mean_squared_error")45 return model46def create_dropout_rnn(seq_len, seq_dim, kernel_initializer=RandomNormal(), kernel_regularizer=l2(0.0001), hidden_units=100,47 hidden_layers=1, activation="relu", dropout_recurrent=0.1, dropout_dense=0.1,48 dropout_input=0.1, dropout_lstm=0.1):49 x = Input(shape=(seq_len, seq_dim))50 h = Dropout(dropout_input)(x)51 h = LSTM(hidden_units, activation=activation, kernel_regularizer=kernel_regularizer, recurrent_regularizer=kernel_regularizer,52 recurrent_dropout=dropout_recurrent, dropout=dropout_lstm, bias_regularizer=kernel_regularizer,53 kernel_initializer=kernel_initializer, recurrent_initializer=kernel_initializer,54 bias_initializer=kernel_initializer)(h)55 h = Dropout(dropout_dense)(h)56 y = Dense(1, activation="linear", kernel_regularizer=kernel_regularizer, bias_regularizer=kernel_regularizer,57 kernel_initializer=kernel_initializer, bias_initializer=kernel_initializer)(h)58 model = Model(x, y)59 model.compile(optimizer="adam", loss="mean_squared_error")60 return model61def create_dropout_rnn_stateful(seq_len, seq_dim, kernel_initializer=RandomNormal(), kernel_regularizer=l2(0.0001), hidden_units=100,62 hidden_layers=1, activation="relu", dropout_recurrent=0.1, dropout_dense=0.1,63 dropout_input=0.1, dropout_lstm=0.1):64 x = Input(batch_shape=(1, seq_len, seq_dim))65 if hidden_layers == 1:66 h = Dropout(dropout_input)(x)67 h = LSTM(hidden_units, activation=activation, kernel_regularizer=kernel_regularizer, recurrent_regularizer=kernel_regularizer,68 recurrent_dropout=dropout_recurrent, dropout=dropout_lstm, bias_regularizer=kernel_regularizer, stateful=True,69 kernel_initializer=kernel_initializer, recurrent_initializer=kernel_initializer,70 bias_initializer=kernel_initializer, return_sequences=False)(h)71 for i in range(hidden_layers - 1):72 h = LSTM(hidden_units, activation=activation, kernel_regularizer=kernel_regularizer,73 recurrent_regularizer=kernel_regularizer,74 recurrent_dropout=dropout_recurrent, dropout=dropout_lstm, bias_regularizer=kernel_regularizer,75 stateful=True,76 kernel_initializer=kernel_initializer, recurrent_initializer=kernel_initializer,77 bias_initializer=kernel_initializer, return_sequences=True)(h)78 if hidden_layers > 1:79 h = LSTM(hidden_units, activation=activation, kernel_regularizer=kernel_regularizer,80 recurrent_regularizer=kernel_regularizer,81 recurrent_dropout=dropout_recurrent, dropout=dropout_lstm, bias_regularizer=kernel_regularizer,82 stateful=True,83 kernel_initializer=kernel_initializer, recurrent_initializer=kernel_initializer,84 bias_initializer=kernel_initializer, return_sequences=False)(h)85 h = Dropout(dropout_dense)(h)86 y = Dense(1, activation="linear", kernel_regularizer=kernel_regularizer, bias_regularizer=kernel_regularizer,87 kernel_initializer=kernel_initializer, bias_initializer=kernel_initializer)(h)88 model = Model(x, y)89 model.compile(optimizer="adam", loss="mean_squared_error")90 return model91def reinstantiate_model(config, weights):92 model = Model.from_config(config)93 model.set_weights(weights)94 return model95# class CustomLSTM(LSTM):96# def call(self, inputs, mask=None, training=None, initial_state=None):97# self.cell._generate_dropout_mask(inputs, training=training)98# self.cell._generate_recurrent_dropout_mask(inputs, training=training)99# return RNN.call(inputs, mask=mask, training=training,...
Automation Testing Tutorials
Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.
LambdaTest Learning Hubs:
- JUnit Tutorial
- TestNG Tutorial
- Webdriver Tutorial
- WebDriverIO Tutorial
- Protractor Tutorial
- Selenium 4 Tutorial
- Jenkins Tutorial
- NUnit Tutorial
- Jest Tutorial
- Playwright Tutorial
- Cypress Tutorial
- PyTest Tutorial
YouTube
You could also refer to video tutorials over LambdaTest YouTube channel to get step by step demonstration from industry experts.
Run autotest automation tests on LambdaTest cloud grid
Perform automation testing on 3000+ real desktop and mobile devices online.
Try LambdaTest Now !!
Get 100 minutes of automation test minutes FREE!!
Was this article helpful?
