"Cannot feed value of shape" error with simple tensorflow + tflearn neural network

Question

I have seen a few people with similar errors, but none seem to pertain to me and as someone who is relatively new to python it's all a bit confusing.

I'm using Tensorflow + TFLearn to try and make a very simple network that can predict the price of avocados given the type, place of origin, and year (Don't ask), and It keeps throwing the error:

Cannot feed value of shape (10, 500) for Tensor 'TargetsData/Y:0', which has shape '(?, 1)'

I have n_class set to 500 because I'm working with some higher numbers and it kept throwing an error if it was too low, but from what I understand it isn't that, it is something to do with the "shape", but I don't really know what that means.

I'll paste in my full code, sorry if this is a bad question/a simple question, I'm just new to all this and a bit confused.

data, labels = tflearn.data_utils.load_csv('avocado.csv',
                                               target_column=2,
                                               categorical_labels=True,
                                               n_classes=500,
                                               columns_to_ignore=[0,1,3,4,5,6,7,8,9,10])
    for type in data:
        if type[0] == "conventional":
            type[0] = 1
        else:
            type[0] = 0

    for place in data:
        if place[2] == "Albany":
            place[2] = 0
        elif place[2] == "Atlanta":
            place[2] = 1
        elif place[2] == "BaltimoreWashington":
            place[2] = 2
        elif place[2] == "Boise":
            place[2] = 3
        elif place[2] == "Boston":
            place[2] = 4
        elif place[2] == "BuffaloRochester":
            place[2] = 5
        elif place[2] == "California":
            place[2] = 6
        elif place[2] == "Charlotte":
            place[2] = 7

    #this goes on for a while, just converting strings to int to work 
    #with TFLearn

    print(data[0])
    # define the input layer
    # 3 because we have 3 columns in the data set (year, location, and type)
    net = tflearn.input_data(shape=[None, 3])

    # adding hidden layers
    net = tflearn.fully_connected(net, 32)
    net = tflearn.fully_connected(net, 32)

    # the output layer
    net = tflearn.fully_connected(net, 1, activation="softmax")

    net = tflearn.regression(net)

    # define model
    model = tflearn.DNN(net)

    # start training
    model.fit(data, labels, n_epoch=10, batch_size=10, show_metric=True)

Thanks in advance for any help, and again sorry if this is a stupid question.

Answer 1

Your example code along with the problem description seems to be about regression, however you've set categorical_labels=True and along with n_classes=500 this will treat the avocado price as a categorical variable with 500 different possible values (instead of a continuous variable). This creates 500 class one-hot vectors which is incompatible with the last layer of your network (which has one output node instead of 500). For your regression problem you should set categorical_labels=False and you can leave out n_classes at all.