My Perceptron algorithm gives me a single training example prediction on Python for Iris dataset

Question

I'm new to Machine Learning. My Perceptron code works but just for the first training example. Also, the b_final is n-dimensional array instead of being a scalar. Can you help me with these issues? The X.shape = (150,4) and y.shape = (100,).

The dataset is "Iris dataset".

m, n = X.shape

def weighted_sum(X,w,b):

    for i in range(m):
        f_Xwb = np.dot(X[i],w) + b
        return f_Xwb

def Prediction(X,initial_w,initial_b):
    return np.where(weighted_sum(X,initial_w,initial_b) > 0 , 1 , -1)

def weights(X , y , w , b , epochs , eta):

    for i in range(epochs):

        for xi,yi in zip(X,y):
            w = w - eta * ((Prediction(xi,initial_w,initial_b)-yi) * xi)
            b = b - eta * (Prediction(xi,initial_w,initial_b)-yi)
    return w , b


initial_w = np.ones(n)
initial_b = 1

w_final , b_final = weights(X,y,initial_w,initial_b,50,0.3)
print(w_final,b_final)

y_prediction = Prediction(X,w_final,b_final)
y_prediction

Also, the y_prediction is n-dimensional array instead of being a scalar.