ValueError: Cannot convert a partially known TensorShape

Question

I

code:

from tensorflow.keras.layers import Layer
        
class PCA(Layer) :

    def __init__(self, n_components):
        super(PCA,self).__init__()
        self.n_components = n_components
        self.components = None
        self.mean = None


    def build (self, X):
        #mean centering
        self.mean = tf.reduce_mean(tf.shape(X),axis=0,keepdims=True) #np.mean(X, axis=0)
        X = X - self.mean

        #covariance, function needs samples as columns
        cov = np.cov(X.T)

        # Eigenvectors, eigenvalues
        eigenvectors, eigenvalues = np.linalg.eig(cov)


        # Sort eigenvectors
        idxs = np.argsort(eigenvalues)[::-1]
        eigenvalues = eigenvalues[idxs]
        eigenvectors = eigenvectors[idxs]

        self.components = eigenvectors[:self.n_components]



    def call (self,X):
        #project data
        X = X - self.mean

        return np.dot(X, self.components.T)


def mpox(inputs):
    
    features = feature_extraction(inputs)
    pca = PCA(n_components=100)
    reduced_features = pca(features)
    classification_output = classifier(reduced_features)
    
    mpox_Model = Model(inputs = inputs, outputs = classification_output)
    return mpox_Model

I wanted to sandwich a custome PCA layer into a model right after the feature extraction function. I want to reduce the dimension of the extracted features before classification.But I seem to be getting ValueError: Cannot convert a partially known TensorShape all the time I have tried using the tf.shape but it isn't working.