No more than 10 conditional features about cond_rnn HOT 14 CLOSED

!pip install --upgrade cond-rnn

def create_conditions(NUM_SAMPLES, NUM_CLASSES):
    conditions = np.zeros(shape=[NUM_SAMPLES, NUM_CLASSES])
    for i, kk in enumerate(conditions):
             kk[i % NUM_CLASSES] = 1
    return conditions

def main():
    i = Input(shape=[time_steps, input_dim], name='input_0')
    c1 = Input(shape=[num_classes], name='input_1')
    c2 = Input(shape=[num_classes], name='input_2')
    c3 = Input(shape=[num_classes], name='input_3')
    c4 = Input(shape=[num_classes], name='input_4')
    c5 = Input(shape=[num_classes], name='input_5')
    c6 = Input(shape=[num_classes], name='input_6')
    c7 = Input(shape=[num_classes], name='input_7')
    c8 = Input(shape=[num_classes], name='input_8')
    c9 = Input(shape=[num_classes], name='input_9')
    c10 = Input(shape=[num_classes], name='input_10')
    # c11 = Input(shape=[num_classes], name='input_11')
        # add the condition tensor here.
    x = ConditionalRecurrent(LSTM(num_cells, return_sequences=True, name='cond_rnn_0'))([i,c1,c2,c3,c4,c5,c6,c7,c8,c9,c10])
        # and here too.
    x = ConditionalRecurrent(LSTM(num_cells, return_sequences=False, name='cond_rnn_1'))([x,c1,c2,c3,c4,c5,c6,c7,c8,c9,c10])
    x = Dense(units=num_classes, activation='softmax')(x)
    model = Model(inputs=[i,c1,c2,c3,c4,c5,c6,c7,c8,c9,c10], outputs=[x])

        # Define data.
    test_inputs = np.random.uniform(size=(num_samples, time_steps, input_dim))
    test_con = test_targets = create_conditions(num_samples, num_classes)
    #print(test_con)
    
    model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
    model.fit(verbose=2, x = [test_inputs,test_con,test_con,test_con,test_con
                              ,test_con,test_con,test_con,test_con,test_con,test_con], y=test_targets, epochs=10)

    pred1 = model.predict([test_inputs,test_con,test_con,test_con,test_con
                           ,test_con,test_con,test_con,test_con,test_con,test_con])
    #print(pred1)

num_samples = 10
    input_dim = 1
    num_classes = 3
    time_steps = 10
    num_cells = 6
    
if __name__ == '__main__':
    main()

!pip install --upgrade cond-rnn