Windows11非Anaconda环境下导入TensorFlow报错的解决请求

问题背景 #

Windows 11系统，未使用Anaconda，多次重装Python版本搭建TensorFlow环境，导入tensorflow时出现模块导入级联错误，排查报错相关文件后无法定位问题。

运行代码 #

import pandas as pd
data = pd.read_csv('gpascore.csv')
blank = data.isnull().sum()
blank_keys = blank.keys()
for i in range(len(blank_keys)):
  if blank[blank_keys[i]] > 0:
    data = data.dropna()
    break
admit = data.admit.values
del data['admit']

dataset = []
for i, rows in data.iterrows():
  temp = [rows['gre'], rows['gpa'], rows['rank']]
  dataset.append(temp)

import numpy as np
import tensorflow as tf
from tensorflow import keras

model = tf.keras.models.Sequential([
  tf.keras.layers.Dense(64, activation='tanh'),
  tf.keras.layers.Dense(128, activation='tanh'),
  tf.keras.layers.Dense(1, activation='sigmoid'),
])
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
model.fit(np.array(dataset), np.array(admit), epochs=1000)
predict = model.predict([[750, 3.70, 3], [400, 2.20, 1]])
print(predict)

报错信息 #

Traceback (most recent call last):
  File "C:\\Users\\user\\AppData\\Local\\Programs\\Python\\Python311\\app.py", line 18, in <module>
import tensorflow as tf
  File "C:\\Users\\user\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\tensorflow\__init__.py", line 48, in <module>
from tensorflow._api.v2 import __internal__
  File "C:\\Users\\user\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\tensorflow_api\\v2_internal_\__init__.py", line 8, in <module>
from tensorflow._api.v2.__internal__ import autograph
  File "C:\\Users\\user\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\tensorflow_api\\v2_internal_\\autograph\__init__.py", line 8, in <module>
from tensorflow.python.autograph.core.ag_ctx import control_status_ctx # line: 34
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\\Users\\user\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\tensorflow\\python\\autograph\__init__.py", line 57, in <module>
remove_undocumented(__name__, _allowed_symbols)
  File "C:\\Users\\user\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\tensorflow\\python\\util\\all_util.py", line 99, in remove_undocumented
should_have = make_all(module_name, doc_string_modules)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\\Users\\user\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\tensorflow\\python\\util\\all_util.py", line 51, in make_all
for m in _reference_pattern.finditer(doc_module.__doc__)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: expected string or bytes-like object, got 'NoneType'

解决方案 #

从报错信息判断，这是TensorFlow安装文件损坏或版本与Python不兼容导致的问题，按以下步骤操作：

• 完全清理TensorFlow残留
  打开命令提示符，执行pip uninstall tensorflow -y卸载现有版本，然后手动删除Python安装目录下Lib\site-packages中所有带tensorflow、tensorflow_core、tensorflow_api的文件夹，避免残留文件干扰。

• 匹配Python与TensorFlow版本
  执行python --version查看当前Python版本，严格按照兼容表选择对应TensorFlow版本：比如Python 3.11需TensorFlow 2.13及以上，Python 3.10对应TensorFlow 2.10+，版本不匹配会引发各类导入错误。

• 重新安装TensorFlow
  先升级pip：python -m pip install --upgrade pip，再安装指定版本，比如pip install tensorflow==2.15.0（根据你的Python版本调整）；若安装缓慢，可加国内镜像源：pip install tensorflow==2.15.0 -i https://pypi.tuna.tsinghua.edu.cn/simple

• 验证安装
  新建空白Python文件，仅执行import tensorflow as tf，无报错则安装成功，再运行完整代码测试。

内容的提问来源于stack exchange，提问作者황영민