grad_to_vect.py

def gradients_to_vector(gradients):
    """
    Roll all our gradients dictionary into a single vector satisfying our specific required shape.
    """
    
    count = 0
    params = []
    num_params = len(parameters)//2
    for l in range(1,num_params+1):
        params = params + ["dW"+str(l)]
        params = params + ["db"+str(l)]
        
    for key in params:
        # flatten parameter
        new_vector = np.reshape(gradients[key], (-1,1))
        
        if count == 0:
            theta = new_vector
        else:
            theta = np.concatenate((theta, new_vector), axis=0)
        count = count + 1

    return theta

vec_to_dict.py

def vector_to_dictionary(theta,parameters):
    """
    Unroll all our parameters dictionary from a single vector satisfying our specific required shape.
    """
    params = {}
    num_params = len(parameters)//2
    num_data_n = 0
    for l in range(1,num_params+1):
        num_data = num_data_n
        num_data_n = num_data + parameters["W"+str(l)].shape[0]*parameters["W"+str(l)].shape[1]
        params["W"+str(l)] = theta[num_data:num_data_n].reshape((parameters["W"+str(l)].shape[0],parameters["W"+str(l)].shape[1]))
        num_data = num_data_n
        num_data_n = num_data + parameters["b"+str(l)].shape[0]*parameters["b"+str(l)].shape[1]
        params["b"+str(l)] = theta[num_data:num_data_n].reshape((parameters["b"+str(l)].shape[0],parameters["b"+str(l)].shape[1]))

    return params