TensorFlow导入报错：找不到tensorflow.python.eager.polymorphic_function模块

问题详情 #

尝试多种方法均无法解决该错误，特寻求帮助。导入TensorFlow Object Detection模块时触发以下错误：

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
c:\Users\lmumelo\Documents\GitHub\NAR_Tensorflow_OD\scripts\modelling\NAR Object Detection Project.ipynb Cell 3 in <cell line: 25>()
     20 #from tflite_model_maker import model_spec
     21 #from tflite_model_maker import object_detector
     23 sys.path.append("..")
---> 25 from object_detection.utils import visualization_utils as vis_util
     26 from object_detection.utils import dataset_util, label_map_util, config_util
     27 from object_detection.protos import string_int_label_map_pb2

File c:\Users\lmumelo\Anaconda3\envs\tensorflow\lib\site-packages\object_detection\utils\visualization_utils.py:39, in <module>
     37 from six.moves import range
     38 from six.moves import zip
---> 39 import tensorflow.compat.v1 as tf
     41 from object_detection.core import keypoint_ops
     42 from object_detection.core import standard_fields as fields

File ~\AppData\Roaming\Python\Python39\site-packages\tensorflow\__init__.py:92, in <module>
     90 from tensorflow.python.eager.backprop import GradientTape
     91 from tensorflow.python.eager.context import executing_eagerly
---> 92 from tensorflow.python.eager.polymorphic_function.polymorphic_function import function
     93 from tensorflow.python.framework.constant_op import constant
     94 from tensorflow.python.framework.device_spec import DeviceSpecV2 as DeviceSpec

ModuleNotFoundError: No module named 'tensorflow.python.eager.polymorphic_function'

触发错误的导入代码 #

from object_detection.utils import visualization_utils as vis_util
from object_detection.utils import dataset_util, label_map_util, config_util
from object_detection.protos import string_int_label_map_pb2
from object_detection.utils import ops as utils_ops
from object_detection.protos import pipeline_pb2
from object_detection.builders import model_builder
from google.protobuf import text_format

import tensorflow as tf

当前已安装的相关包版本 #

tb-nightly                    2.11.0a20221023
tensorboard                   2.10.1
tensorboard-data-server       0.6.1
tensorboard-plugin-wit        1.8.0
tensorflow                    2.5.0
tensorflow-addons             0.18.0
tensorflow-datasets           4.7.0
tensorflow-estimator          2.5.0
tensorflow-hub                0.12.0
tensorflow-io                 0.27.0
tensorflow-io-gcs-filesystem  0.27.0
tensorflow-metadata           1.10.0
tensorflow-model-optimization 0.7.3
tensorflow-text               2.10.0
termcolor                     1.1.0
terminado                     0.13.1
testpath                      0.6.0
text-unidecode                1.3
tf-estimator-nightly          2.12.0.dev2022102308
tf-models-official            2.10.0
tf-nightly                    2.12.0.dev20221023
tf_nightly_intel              2.12.0.dev20221023

已尝试的无效操作 #

pip uninstall tensorflow
pip install tensorflow --upgrade

注：使用Anaconda环境开发，所有包均安装在该环境内。

问题根源 #

环境中同时存在正式版TensorFlow（2.5.0）和多个Nightly预发行版（tf-nightly、tf_nightly_intel等），版本冲突导致TensorFlow内部模块路径混乱，无法找到polymorphic_function模块。此外，tf-models-official（2.10.0）与TensorFlow（2.5.0）版本不匹配，进一步加剧兼容性问题。

修复步骤 #

1. 完全清理冲突的TensorFlow包
   激活Anaconda环境后，执行以下命令卸载所有TensorFlow相关的正式版和Nightly版：

   pip uninstall -y tensorflow tf-nightly tf_nightly_intel tf-estimator-nightly tb-nightly
   conda uninstall -y tensorflow
   

   执行完成后，用pip list | findstr tensorflow（Windows）或pip list | grep tensorflow（Linux/macOS）确认无残留的TensorFlow相关包。

2. 安装版本匹配的TensorFlow包
   当前tf-models-official版本为2.10.0，需安装对应版本的TensorFlow：

   pip install tensorflow==2.10.0 tensorflow-estimator==2.10.0
   

3. 验证环境一致性
   安装完成后，检查核心包版本是否匹配：

   pip list | findstr "tensorflow tf-models-official"
   

   确保输出中tensorflow和tf-models-official版本均为2.10.x，无其他冲突版本。

4. 重新运行代码
   激活Anaconda环境后重新执行导入代码，确认错误消失。

内容的提问来源于stack exchange，提问作者LivingstoneM