从sklearn.externals导入joblib失败及模型加载报错求助

Hey there, let's work through these two issues you're facing with your Flask ML prediction service—they're both typical when moving models between environments or updating libraries!

What's causing this? #

Scikit-learn removed the sklearn.externals.joblib import path starting in version 0.23. If you're running a newer version of scikit-learn now (compared to 2 months ago), that import will fail because it's no longer supported.

Fix: #

Replace:

from sklearn.externals import joblib

With:

import joblib

Make sure you have the joblib package installed in your environment—if not, run:

pip install joblib

What's causing this? #

This error almost always stems from version mismatches between the environment where you saved the classifier.joblib model and your current environment. Specifically:

• The scikit-learn version you used to save the model is different from the one you're using now
• The joblib version is incompatible with either the saved model or your current scikit-learn version
• Rarely, a mismatch between 32-bit and 64-bit system architectures can trigger this too

Fixes to try: #

• Match the exact library versions from when you saved the model
  If you remember the scikit-learn and joblib versions used 2 months ago, install those exact versions:

  pip install scikit-learn==[your-old-version] joblib==[corresponding-joblib-version]
  

  Tip: If you don't remember the versions, check old requirements.txt files or environment snapshots (like conda env exports) if you have them.

• Retrain and re-save the model in your current environment
  If you can't access the old versions, the most reliable fix is to re-run your model training code in your current Flask environment, then save a new classifier.joblib file. This ensures the model is fully compatible with your current library versions.

• Verify system architecture consistency
  Make sure your current environment is running on the same architecture (32-bit vs 64-bit) as the machine where you saved the model. Different architectures can cause dtype mismatches in serialized files.

内容的提问来源于stack exchange，提问作者Dilipan M