为什么有时将.h5文件转换为.tflite文件，但为什么当我改变训练范围时却不这样做

Question

model = keras.Sequential([keras.layers.Dense(1, input_shape=(1, 4))])
model.compile(optimizer='sgd', loss='mean_squared_error')
#
xs = np.array([[[-1.0, 1.0, -1.0, 1.0]],
               [[0.0, 2.0, 0.0, 2.0]],
               [[1.0, 3.0, 1.0, 3.0]],
               [[2.0, 4.0, 2.0, 4.0]],
               [[3.0, 5.0, 3.0, 5.0]],
               [[4.0, 6.0, 4.0, 6.0]]], dtype=np.float32)
ys = np.array([-3.0, -1.0, 0.0, 3.0, 5.0, 7.0], dtype=np.float32)
#
model.fit(xs, ys, epochs=500)
#
# # Write out to keras save file
keras_file = 'linear.h5'
# keras.models.save_model(model, keras_file)

# input to predict
model_input = np.array([[[8.0, 10.0, 8.0, 10.0]]], dtype=np.float32)

# make predictions from the h5 file
h5_model = keras.models.load_model(keras_file)
h5_prediction = h5_model.predict(model_input)

# h5_prediction will always be around 19, which is correct
print('Prediction from h5 model: {}'.format(h5_prediction))

# Convert the keras file to TensorFlow Lite
converter = lite.TocoConverter.from_keras_model_file(keras_file)
tflite_model = converter.convert()
open('model_test.tflite', 'wb').write(tflite_model)

有错误：

  

tensorflow / lite / toco / tooling_util.cc：625]检查失败：暗>> 1（0对1）

或

  

I tensorflow / lite / toco / graph_transformations / graph_transformations.cc：39]在一般图形转换之前：17个运算符，31个数组（0个量化）。