Scikit-learn(0.18.1) unable to unpickle a model saved from scikit-learn pre-0.18

[JarvisTao]

Hello, thanks for you excellent work. Recently, when i test the CBMV using your example command:

python main.py --l ./datasets/ArtL/im0.png --r ./datasets/ArtL/im1.png --w 694 --h 554 --d 128 --model ./modelall_4m.rf --prob_save_path ./

but it returns that UserWarning: Trying to unpickle estimator DecisionTreeClassifier from version pre-0.18 when using version 0.18.1. This might lead to breaking code or invalid results. Use at your own risk.

After searching for method to resolve it, i found that it's the conflict of scikit-learn version. Can you show us the scikit-learn version you use to save the model?

Or could you save the model by scikit-learn 0.18.1, and reload the new version of Model.

Thanks.

[ccj5351]

Hi, we used the 0.17.1 version scikit-learn for the model. Please check the details in the README.md file of the master branch. You can uninstall scikit-learn using 'pip uninstall scikit-learn' and install this version via pip install -v scikit-learn==0.17.1.

[JarvisTao]

Thanks for your help, it can work.

[JarvisTao]

Hi, we used the 0.17.1 version scikit-learn for the model. Please check the details in the README.md file of the master branch. You can uninstall scikit-learn using 'pip uninstall scikit-learn' and install this version via pip install -v scikit-learn==0.17.1.

Hello, i have found another problem. When i run the test code, it cannot use the GPU, is there some parameters to set? Thanks.

[ccj5351]

The GPU code is only used for cost-volume-optimization, if you choose the MC-CNN method. As for the random forest training and testing, no GPUs are involved. Please see the README.md, I think I already add instruction about that.