TensorFlow utils

1、

def model_stats():
    print("============================================================")
    print("List of all Trainable Variables")
    tvars = tf.trainable_variables()
    all_params = []
    for idx, v in enumerate(tvars):
        print(" var {:3}: {:15} {}".format(idx, str(v.get_shape()), v.name))
        num_params = 1
        param_list = v.get_shape().as_list()
        if(len(param_list)>1):
            for p in param_list:
                if(p>0):
                    num_params = num_params * int(p)
        else:
            all_params.append(param_list[0])
        all_params.append(num_params)
    print("Total number of trainable parameters {}".format(np.sum(all_params)))

 

转载于:https://www.cnblogs.com/niuxichuan/p/10161320.html