fa7ee697076d2061faeaf7ebf20cdcd5da352eae,tests/application_driver_test.py,ApplicationDriverTest,test_multi_device_multi_optimiser_gradients,#ApplicationDriverTest#,193

Before Change

                g_array = np.concatenate([g_0.reshape((1, -1)),
                                          g_1.reshape((1, -1)),
                                          g_2.reshape((1, -1)),
                                          g_3.reshape((1, -1))], axis=0)
                g_ave = g_ave.reshape(-1)
                g_np_ave = np.mean(g_array, axis=0)
                self.assertAllClose(g_np_ave, g_ave)

After Change

                ])

                // query sample discriminator gradient to check
                gen_0, gen_1, gen_2, gen_3, gen_ave = sess.run([
                    tf.get_default_graph().get_tensor_by_name(
                        "worker_0/ComputeGradients/gradients_1/worker_0/tinynet/G/conv/conv_/conv/ExpandDims_1_grad/Reshape:0"),
                    tf.get_default_graph().get_tensor_by_name(
                        "worker_1/ComputeGradients/gradients_1/worker_1/tinynet/G/conv/conv_/conv/ExpandDims_1_grad/Reshape:0"),
                    tf.get_default_graph().get_tensor_by_name(
                        "worker_2/ComputeGradients/gradients_1/worker_2/tinynet/G/conv/conv_/conv/ExpandDims_1_grad/Reshape:0"),
                    tf.get_default_graph().get_tensor_by_name(
                        "worker_3/ComputeGradients/gradients_1/worker_3/tinynet/G/conv/conv_/conv/ExpandDims_1_grad/Reshape:0"),
                    tf.get_default_graph().get_tensor_by_name(
                        "ApplyGradients_1/AveOverDevices_8:0")
                ])
                self.check_gradients(gen_0, gen_1, gen_2, gen_3, gen_ave)
                self.check_gradients(dis_0, dis_1, dis_2, dis_3, dis_ave)
            SESS_FINISHED.send(test_driver.app, itermsg=None)
            test_driver.app.stop()

Instances