863a9dcee691e850e79d97a16abb977e24324fa1,tests/deconvolution_test.py,DeconvTest,test_deconvlayer_2d_bn_reg_valid_shape,#DeconvTest#,285

Before Change

            self.assertAllClose((2, 32, 32, 10), out_2d.shape)

    def test_deconvlayer_2d_bn_reg_valid_shape(self):
        x_2d = self.get_2d_input()
        conv_reg = DeconvolutionalLayer(
            10, 3, 2,
            padding="VALID",
            w_regularizer=regularizers.l2_regularizer(0.5),
            with_bias=False,
            with_bn=True,
            acti_func="prelu")
        conv_reg_out = conv_reg(x_2d, is_training=True, keep_prob=0.4)
        print(conv_reg)
        with self.test_session() as sess:
            sess.run(tf.global_variables_initializer())
            out_2d = sess.run(conv_reg_out)
            self.assertAllClose((2, 33, 33, 10), out_2d.shape)


if __name__ == "__main__":
    tf.test.main()

After Change

                                             is_training=False,
                                             dropout_prob=1.0)

    def test_deconvlayer_2d_bn_reg_valid_shape(self):
        input_param = {"n_output_chns": 10,
                       "kernel_size": [4, 3],
                       "stride": [1, 2],
                       "with_bias": False,
                       "with_bn": True,
                       "acti_func": "prelu",
                       "padding": "VALID",
                       "w_regularizer": regularizers.l2_regularizer(0.5)}
        self._test_deconv_layer_output_shape(rank=2,
                                             param_dict=input_param,
                                             output_shape=(2, 19, 33, 10),
                                             is_training=True,
                                             dropout_prob=0.4)
        self._test_deconv_layer_output_shape(rank=2,
                                             param_dict=input_param,
                                             output_shape=(2, 19, 33, 10),
                                             is_training=False,
                                             dropout_prob=1.0)


if __name__ == "__main__":
    tf.test.main()

Instances