d14ad44d0f425c04be5a6cdca13fc513fba53843,test_models_train.py,,,#,11

Before Change

if __name__ == "__main__":
    network_inputs = np.random.uniform(size=(3, 16, 16, 1))

    model = convolutional_model(input_shapes=list(network_inputs[0].shape),
                                num_frames=len(network_inputs))

    model.compile(optimizer="adam",
                  loss=deep_speaker_loss,
                  metrics=["accuracy"])

    inputs = list(np.expand_dims(network_inputs, axis=1))
    model.fit(inputs, np.expand_dims([0] * len(inputs), axis=1))

    print(model.summary())

After Change

if __name__ == "__main__":
    network_inputs = np.random.uniform(size=(BATCH_SIZE, NUM_FRAMES, 16, 16, 1))

    model = convolutional_model(batch_input_shape=(BATCH_SIZE * NUM_FRAMES, 16, 16, 1))

    model.compile(optimizer="adam",
                  loss="mse",
                  metrics=["accuracy"])

    network_inputs = np.reshape(network_inputs, (-1, 16, 16, 1))

    output = model.predict(network_inputs)

    // stub_targets = np.expand_dims([0] * BATCH_SIZE * NUM_FRAMES, axis=1)
    stub_targets = np.random.uniform(size=(BATCH_SIZE * NUM_FRAMES, 512))
    print(model.train_on_batch(network_inputs, stub_targets))

    // from triplet_loss import deep_speaker_loss

Instances