I have encountered this problem in a TICTACTOE Neural Network in python. Just a simple AI in the making
It ocurrs after I return the last value from feed forward(the output) . I am kind of a beginner btw

    def feedForward(self, input, layer):
        #last activation is returned
        if layer == len(self.shape):
           * print(input) #* here the last activation is actually printed and works**
           * return input # *here the last Activation is returned but as Nonetype?**
        #output Vektor wird initialisiert mit shape
        self.output = np.zeros(self.shape[layer])
        print(input)
        #the Neurons are gone through
        for i in range(self.shape[layer]):
            self.neurons[layer][i].calculateActivations(input)
            self.output[i] = self.neurons[layer][i].getVar("activation")
            print("neuron fired")
        self.feedForward(self.output, layer + 1)

in this Method I print out what comes out at last from my FeedForward but it is just Nonetype
as you will see in the Console

    def runFF(self, inputs):
        print("output --->"+"n")

        *out = self.feedForward(inputs, 0)*
        *print(out)*

this is the Output:

[1 0 0 0 1 0 1 0 0 0 1 0 0 1 0 0 0 1 1 0 0 0 0 1 1 0 0]
#Inputs for Fields, e.g. 001 --- Empty? -> 0, is O? -> 0, is X?-> 1 = 001
neuron fired
neuron fired
neuron fired
neuron fired
neuron fired
[0.97608829 0.98383903 0.93957521 0.93431431 0.98965643] # these are the activations for the Layer 1
neuron fired
neuron fired
neuron fired
[0.98738308 0.96916289 0.94868291] Layer 2 and so on...
neuron fired
neuron fired
neuron fired
neuron fired
neuron fired
[0.7176812  0.885665   0.84968832 0.88980995 0.90426031]
neuron fired
neuron fired
neuron fired
neuron fired
neuron fired
neuron fired
neuron fired
neuron fired
neuron fired
[0.95916376 0.95721362 0.91595512 0.93022966 0.94000431 0.86405699   
 0.93726489 0.89166409 0.89323725]  # this is the last activation from the "print(input)" in feedForward()
*None* # this is what happens when I print the returned last activation

I have used 2 Classes for this NN, this is the Neuron class

class Neuron(object):

    def __init__(self):

        self.weights = []
        self.bias = 0.0
        self.activation = 0.0

    def set_weights(self, amount):
        self.weights = np.array([np.random.random_sample() for i in range( amount)])
    def set_bias(self):
        self.bias = np.random.random_sample()
    def getVar(self, var):
        if var == "weights":
            return self.weights
        elif var == "bias":
            return self.bias
        elif var == "activation":
            return self.activation
        elif var == "all":
            a = ""
            for i  in range (len(self.weights)):
                a = (a +" "+(str)(self.weights[i]))
            return   a +"/" + (str)(self.bias) + " "+ (str)(self.activation)


    def sigmoid(self, z):
        return 1 / (1 + np.exp(-z))

    def calculateActivations(self, inputs):
        z = np.dot(inputs, self.weights) + self.bias
        self.activation = self.sigmoid(z)

Here is the actual MLP (without backpropagation yet just feedForward)

class MLP(object):
    
    def __init__(self, Board):
        self.Board = Board
        self.shape = [5, 3, 5, 9]
        self.layout = [27, 5, 3, 5, 9]
        self.neurons = []

    def feedForward(self, input, layer):
        #rekursionsanker
        if layer == len(self.shape):
            print(input)
            return input
        #output Vektor wird initialisiert mit shape
        self.output = np.zeros(self.shape[layer])
        print(input)
        #Die Neuronen in der Layer werden durchgegangen und berechnet
        for i in range(self.shape[layer]):
            self.neurons[layer][i].calculateActivations(input)
            self.output[i] = self.neurons[layer][i].getVar("activation")
            print("neuron fired")
        self.feedForward(self.output, layer + 1)


    def createNeurons(self):
        for i in range(len(self.shape)):
            a = np.empty(self.shape[i], Neuron)
            for j in range(self.shape[i]):
                a[j] = Neuron()
            self.neurons.append(a)
        print(self.neurons)


    def initializeNeurons(self):
        for i in range(len(self.shape)):
            for j in range(self.shape[i]):
                self.neurons[i][j].set_weights(self.layout[i])
                self.neurons[i][j].set_bias()
                self.printAll(i, j)


    def printAll(self, i ,j):
        x = self.neurons[i][j].getVar("all")
        y = x.split('/')
        z = y[0].split(' ')
        for l in range(self.layout[i]):
            print("weight " + (str)(l + 1) + " = " + z[l+1] + "n")
        print("b/a : " + y[1])

    def runFF(self, inputs):
        print("output --->"+"n")

        out = self.feedForward(inputs, 0)
        print(out)

        def test(self):
        input = np.array([0, 0, 1, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 1])
        self.runFF(input)

there is a test method here so you can go right ahead if you want to

I have spent too much time looking for some sort of a mistake and I am lost
I have looked at other questions similar to mine but i do not seem to be able to draw a sensible conclusion from the others