导入scikit learn时出现ImportError: cannot import name _remove_dead_weakref错误求助

Let’s break down how to troubleshoot this frustrating error, especially since it works on one identical machine but not yours. Here are the most likely fixes to try:

1. Verify Version Compatibility (Critical for Python 2) #

Since you’re using Python 2’s C API, remember that scikit-learn dropped support for Python 2 after version 0.20.4. If you’re trying to use a newer sklearn version with Python 2, that’s almost certainly the issue:

• Check your sklearn version in a working Python shell: import sklearn; print(sklearn.__version__)
• If it’s newer than 0.20.4, reinstall the compatible version:

  pip uninstall -y scikit-learn
  pip install scikit-learn==0.20.4
  

2. Ensure Your Python Environments Match Exactly #

Even if you think the environments are identical, tiny differences can cause this. Compare the package lists between your working and broken machine:

• On both machines, run pip freeze > requirements.txt and diff the files. Look for mismatched versions of sklearn, numpy, scipy, or even Python itself.
• Double-check that both machines are using the same Python interpreter (not a system Python vs. a virtualenv Python, for example).

3. Fix Python C API Initialization Issues #

Since you’re calling Python from Objective-C via the C API, incorrect environment setup is a common culprit:

• Make sure Py_Initialize() is called before any module imports, and that you’re initializing the correct Python environment (the one where sklearn is installed).
• Explicitly add the site-packages directory of your target Python environment to sys.path in your C API code. For example:

  PyObject *sys_path = PySys_GetObject("path");
  PyList_Append(sys_path, PyUnicode_FromString("/path/to/your/python/site-packages"));
  

• Ensure the PYTHONPATH environment variable is set correctly in the Objective-C process (it might not inherit the same path as your shell).

4. Repair a Corrupted Python Environment #

If even a simple import sklearn fails in a regular Python shell, your environment is likely corrupted:

• First, test if the weakref module itself is working: Run import weakref; print(weakref._remove_dead_weakref) in a Python shell. If this throws an error, your Python installation is damaged—reinstall Python 2.7 (make sure to use the same minor version as the working machine).
• If weakref works, reinstall scikit-learn and its dependencies with Python 2.7-compatible versions:

  pip uninstall -y scikit-learn numpy scipy
  pip install numpy==1.16.6 scipy==1.2.3 scikit-learn==0.20.4
  

5. Check for Conflicting Local Files #

Sometimes, a local file named weakref.py or sklearn.py can override the standard library or scikit-learn modules. Double-check that your project directory doesn’t have any files with these names—they’ll cause silent import conflicts.

内容的提问来源于stack exchange，提问作者kmn