问训练好模型之后，进行预测时出现NotImplementedError怎么办？

test_gen = generator(list(yields['test'].keys()), yields['test'], 16)
X_test, y_test= next(test_gen)

IDS = list(yields['test'].keys())
sum = 0
for i in IDS:
    sum += yields['test'][i]
avg = sum / len(IDS)
print("Average Test Yield is ", avg)

model = load_model('CNN_LSTM_AVG_1000')

print(len(X_test),len(y_test))
a, b = model.evaluate(X_test, y_test, batch_size=16)
y_pred = model.predict(X_test, batch_size=16, verbose=1)

def generator(IDs, yields, batch_size, cutoff=None):
    import numpy as np
    import random

    # Create empty arrays to get batch of features and labels

    if cutoff != None:
        batch_features = np.zeros((batch_size, cutoff, 1, 256, 10))
        batch_yields = np.zeros((batch_size))
        while True:
            for i in range(batch_size):
                # choose random index in features
                index = random.choice(range(len(IDs)))
                ID = IDs[index]
                if np.sum(np.isnan(np.load('Data/PROCESSED_III/' + ID + '.npy'))) == 0:
                    batch_features[i, :, :, :, :] = np.load('Data/PROCESSED_III/' + ID + '.npy')[:cutoff, :, :, :]
                    # print('yes', ID)
                    batch_yields[i] = yields[ID]
                else:
                    print('no', ID)

            yield batch_features, batch_yields

    else:
        batch_features = np.zeros((batch_size, 14, 1, 32, 4))
        batch_yields = np.zeros((batch_size))
        while True:
            for i in range(batch_size):
                # choose random index in features
                index = random.choice(range(len(IDs)))
                ID = IDs[index]
                if np.sum(np.isnan(np.load('Data/data_1/' + ID + '.npy'))) == 0:
                    batch_features[i, :, :, :, :] = np.load('Data/data_1/' + ID + '.npy')
                    # print('yes', ID)
                    batch_yields[i] = yields[ID]
                else:
                    print('no', ID)
            yield batch_features, batch_yields

Traceback (most recent call last):
  File "F:/Crop-Yield-Prediction-Using-CNN-LSTM/Crop-Yield-Prediction-Using-CNN-LSTM--Temp/Code/Train_HB.py", line 174, in <module>
    a, b = model.evaluate(X_test, y_test, batch_size=16)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\keras\engine\training.py", line 66, in _method_wrapper
    return method(self, *args, **kwargs)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\keras\engine\training.py", line 1081, in evaluate
    tmp_logs = test_function(iterator)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\eager\def_function.py", line 580, in __call__
    result = self._call(*args, **kwds)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\eager\def_function.py", line 627, in _call
    self._initialize(args, kwds, add_initializers_to=initializers)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\eager\def_function.py", line 506, in _initialize
    *args, **kwds))
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\eager\function.py", line 2446, in _get_concrete_function_internal_garbage_collected
    graph_function, _, _ = self._maybe_define_function(args, kwargs)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\eager\function.py", line 2777, in _maybe_define_function
    graph_function = self._create_graph_function(args, kwargs)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\eager\function.py", line 2667, in _create_graph_function
    capture_by_value=self._capture_by_value),
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\framework\func_graph.py", line 981, in func_graph_from_py_func
    func_outputs = python_func(*func_args, **func_kwargs)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\eager\def_function.py", line 441, in wrapped_fn
    return weak_wrapped_fn().__wrapped__(*args, **kwds)
  File "D:\anaconda3\envs\DL\lib\site-packages\tensorflow\python\framework\func_graph.py", line 968, in wrapper
    raise e.ag_error_metadata.to_exception(e)
NotImplementedError: in user code: