Python wrappers.TimeDistributed方法代碼示例

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def GeneratorPretraining(V, E, H):
    '''
    Model for Generator pretraining. This model's weights should be shared with
        Generator.
    # Arguments:
        V: int, Vocabrary size
        E: int, Embedding size
        H: int, LSTM hidden size
    # Returns:
        generator_pretraining: keras Model
            input: word ids, shape = (B, T)
            output: word probability, shape = (B, T, V)
    '''
    # in comment, B means batch size, T means lengths of time steps.
    input = Input(shape=(None,), dtype='int32', name='Input') # (B, T)
    out = Embedding(V, E, mask_zero=True, name='Embedding')(input) # (B, T, E)
    out = LSTM(H, return_sequences=True, name='LSTM')(out)  # (B, T, H)
    out = TimeDistributed(
        Dense(V, activation='softmax', name='DenseSoftmax'),
        name='TimeDenseSoftmax')(out)    # (B, T, V)
    generator_pretraining = Model(input, out)
    return generator_pretraining 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def model_masking(discrete_time, init_alpha, max_beta):
    model = Sequential()

    model.add(Masking(mask_value=mask_value,
                      input_shape=(n_timesteps, n_features)))
    model.add(TimeDistributed(Dense(2)))
    model.add(Lambda(wtte.output_lambda, arguments={"init_alpha": init_alpha,
                                                    "max_beta_value": max_beta}))

    if discrete_time:
        loss = wtte.loss(kind='discrete', reduce_loss=False).loss_function
    else:
        loss = wtte.loss(kind='continuous', reduce_loss=False).loss_function

    model.compile(loss=loss, optimizer=RMSprop(
        lr=lr), sample_weight_mode='temporal')
    return model 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def drqn(input_shape, action_size, learning_rate):

        model = Sequential()
        model.add(TimeDistributed(Convolution2D(32, 8, 8, subsample=(4,4), activation='relu'), input_shape=(input_shape)))
        model.add(TimeDistributed(Convolution2D(64, 4, 4, subsample=(2,2), activation='relu')))
        model.add(TimeDistributed(Convolution2D(64, 3, 3, activation='relu')))
        model.add(TimeDistributed(Flatten()))

        # Use all traces for training
        #model.add(LSTM(512, return_sequences=True,  activation='tanh'))
        #model.add(TimeDistributed(Dense(output_dim=action_size, activation='linear')))

        # Use last trace for training
        model.add(LSTM(512,  activation='tanh'))
        model.add(Dense(output_dim=action_size, activation='linear'))

        adam = Adam(lr=learning_rate)
        model.compile(loss='mse',optimizer=adam)

        return model 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def a2c_lstm(input_shape, action_size, value_size, learning_rate):
        """Actor and Critic Network share convolution layers with LSTM
        """

        state_input = Input(shape=(input_shape)) # 4x64x64x3
        x = TimeDistributed(Convolution2D(32, 8, 8, subsample=(4,4), activation='relu'))(state_input)
        x = TimeDistributed(Convolution2D(64, 4, 4, subsample=(2,2), activation='relu'))(x)
        x = TimeDistributed(Convolution2D(64, 3, 3, activation='relu'))(x)
        x = TimeDistributed(Flatten())(x)

        x = LSTM(512, activation='tanh')(x)

        # Actor Stream
        actor = Dense(action_size, activation='softmax')(x)

        # Critic Stream
        critic = Dense(value_size, activation='linear')(x)

        model = Model(input=state_input, output=[actor, critic])

        adam = Adam(lr=learning_rate, clipnorm=1.0)
        model.compile(loss=['categorical_crossentropy', 'mse'], optimizer=adam, loss_weights=[1., 1.])

        return model 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def change_trainable(layer, trainable, verbose=False):
    """ Helper method that fixes some of Keras' issues with wrappers and
        trainability. Freezes or unfreezes a given layer.

    # Arguments:
        layer: Layer to be modified.
        trainable: Whether the layer should be frozen or unfrozen.
        verbose: Verbosity flag.
    """

    layer.trainable = trainable

    if type(layer) == Bidirectional:
        layer.backward_layer.trainable = trainable
        layer.forward_layer.trainable = trainable

    if type(layer) == TimeDistributed:
        layer.backward_layer.trainable = trainable

    if verbose:
        action = 'Unfroze' if trainable else 'Froze'
        print("{} {}".format(action, layer.name)) 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def build_cnn_to_lstm_model(self, input_shape, optimizer=Adam(lr=1e-6, decay=1e-5)):

        model = Sequential()

        model.add(TimeDistributed(Convolution2D(16, 3, 3), input_shape=input_shape))
        model.add(TimeDistributed(Activation('relu')))
        model.add(TimeDistributed(Convolution2D(16, 3, 3)))
        model.add(TimeDistributed(Activation('relu')))
        model.add(TimeDistributed(MaxPooling2D(pool_size=(2, 2))))
        model.add(TimeDistributed(Dropout(0.2)))
        model.add(TimeDistributed(Flatten()))
        model.add(TimeDistributed(Dense(200)))
        model.add(TimeDistributed(Dense(50, name="first_dense")))
        model.add(LSTM(20, return_sequences=False, name="lstm_layer"))
        model.add(Dense(2, activation='softmax'))

        model.compile(loss='categorical_crossentropy', optimizer=optimizer)

        self.model = model 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def test_large_batch_gpu(self):
        batch_size = 2049
        num_channels = 4
        kernel_size = 3

        model = Sequential()
        model.add(
            TimeDistributed(Dense(num_channels), input_shape=(batch_size, kernel_size))
        )

        model.set_weights(
            [(np.random.rand(*w.shape) - 0.5) / 5.0 for w in model.get_weights()]
        )

        self._test_keras_model(
            model, input_blob="data", output_blob="output", delta=1e-2
        ) 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def test_tiny_image_captioning(self):
        # use a conv layer as a image feature branch
        img_input_1 = Input(shape=(16, 16, 3))
        x = Convolution2D(2, 3, 3)(img_input_1)
        x = Flatten()(x)
        img_model = Model([img_input_1], [x])

        img_input = Input(shape=(16, 16, 3))
        x = img_model(img_input)
        x = Dense(8, name="cap_dense")(x)
        x = Reshape((1, 8), name="cap_reshape")(x)

        sentence_input = Input(shape=(5,))  # max_length = 5
        y = Embedding(8, 8, name="cap_embedding")(sentence_input)
        z = merge([x, y], mode="concat", concat_axis=1, name="cap_merge")
        z = LSTM(4, return_sequences=True, name="cap_lstm")(z)
        z = TimeDistributed(Dense(8), name="cap_timedistributed")(z)

        combined_model = Model([img_input, sentence_input], [z])
        self._test_keras_model(combined_model, one_dim_seq_flags=[False, True]) 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def test_large_batch_gpu(self):

        batch_size = 2049
        num_channels = 4
        kernel_size = 3

        model = Sequential()
        model.add(
            TimeDistributed(Dense(num_channels), input_shape=(batch_size, kernel_size))
        )

        model.set_weights(
            [(np.random.rand(*w.shape) - 0.5) * 0.2 for w in model.get_weights()]
        )

        self._test_model(model, delta=1e-2) 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def test_time_distributed_conv(self):
        model = Sequential()
        model.add(
            TimeDistributed(
                Conv2D(64, (3, 3), activation="relu"), input_shape=(1, 30, 30, 3)
            )
        )
        model.add(TimeDistributed(MaxPooling2D((2, 2), strides=(1, 1))))
        model.add(TimeDistributed(Conv2D(32, (4, 4), activation="relu")))
        model.add(TimeDistributed(MaxPooling2D((2, 2), strides=(2, 2))))
        model.add(TimeDistributed(Conv2D(32, (4, 4), activation="relu")))
        model.add(TimeDistributed(MaxPooling2D((2, 2), strides=(2, 2))))
        model.add(TimeDistributed(Flatten()))
        model.add(Dropout(0.5))
        model.add(LSTM(32, return_sequences=False, dropout=0.5))
        model.add(Dense(10, activation="sigmoid"))
        self._test_model(model) 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def test_tiny_image_captioning(self):
        # use a conv layer as a image feature branch
        img_input_1 = Input(shape=(16, 16, 3))
        x = Conv2D(2, (3, 3))(img_input_1)
        x = Flatten()(x)
        img_model = Model(inputs=[img_input_1], outputs=[x])

        img_input = Input(shape=(16, 16, 3))
        x = img_model(img_input)
        x = Dense(8, name="cap_dense")(x)
        x = Reshape((1, 8), name="cap_reshape")(x)

        sentence_input = Input(shape=(5,))  # max_length = 5
        y = Embedding(8, 8, name="cap_embedding")(sentence_input)
        z = concatenate([x, y], axis=1, name="cap_merge")
        z = LSTM(4, return_sequences=True, name="cap_lstm")(z)
        z = TimeDistributed(Dense(8), name="cap_timedistributed")(z)

        combined_model = Model(inputs=[img_input, sentence_input], outputs=[z])
        self._test_model(combined_model, one_dim_seq_flags=[False, True]) 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def test_regularizers():
    model = Sequential()
    model.add(wrappers.TimeDistributed(
        layers.Dense(2, kernel_regularizer='l1'), input_shape=(3, 4)))
    model.add(layers.Activation('relu'))
    model.compile(optimizer='rmsprop', loss='mse')
    assert len(model.layers[0].layer.losses) == 1
    assert len(model.layers[0].losses) == 1
    assert len(model.layers[0].get_losses_for(None)) == 1
    assert len(model.losses) == 1

    model = Sequential()
    model.add(wrappers.TimeDistributed(
        layers.Dense(2, activity_regularizer='l1'), input_shape=(3, 4)))
    model.add(layers.Activation('relu'))
    model.compile(optimizer='rmsprop', loss='mse')
    assert len(model.losses) == 1 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def GetLSTMEncoder(xin, uin, dense_size, lstm_size, dense_layers=1,
        lstm_layers=1):
    '''
    Get LSTM encoder.
    '''
    x = xin
    for _ in xrange(dense_layers):
        if uin is not None:
            x = Concatenate(axis=-1)([x, uin])
        x = TimeDistributed(Dense(dense_size))(x)
        x = TimeDistributed(Activation('relu'))(x)
    for i in xrange(lstm_layers):
        if i == lstm_layers - 1:
            sequence_out = False
        else:
            sequence_out = True
        #sequence_out = True
        x = LSTM(lstm_size, return_sequences=sequence_out)(x)
        x = Activation('relu')(x)
    return x 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def _buildDecoder(self, z, latent_rep_size, max_length, charset_length):
    h = Dense(latent_rep_size, name='latent_input', activation='relu')(z)
    h = RepeatVector(max_length, name='repeat_vector')(h)
    h = GRU(501, return_sequences=True, name='gru_1')(h)
    h = GRU(501, return_sequences=True, name='gru_2')(h)
    h = GRU(501, return_sequences=True, name='gru_3')(h)
    return TimeDistributed(
        Dense(charset_length, activation='softmax'), name='decoded_mean')(h) 

# 需要導入模塊: from keras.layers import wrappers [as 別名]
# 或者: from keras.layers.wrappers import TimeDistributed [as 別名]
def model_no_masking(discrete_time, init_alpha, max_beta):
    model = Sequential()
    model.add(TimeDistributed(Dense(2), input_shape=(n_timesteps, n_features)))

    model.add(Lambda(wtte.output_lambda, arguments={"init_alpha": init_alpha,
                                                    "max_beta_value": max_beta}))

    if discrete_time:
        loss = wtte.loss(kind='discrete').loss_function
    else:
        loss = wtte.loss(kind='continuous').loss_function

    model.compile(loss=loss, optimizer=RMSprop(lr=lr))

    return model