Using pytorch 2.0.1 and CUDA 11.8. Due to numpy type (np.int, np.float, etc...) is deprecated after numpy 1.20, and sklearn uses a lot in their code, so I upgrade the sklearn version to the latest (1.2.2), not sure if some bugs is caused by this upgrade.

1. fix some typos in README.md, and delete the version limitation of sklearn.
2. delete the version limitation of sklearn in requirements.txt.
3. since I upgrade the sklearn version, need to update 1 import line in camouflage_HRNet.py.
4. fix HRNet.py to solve runtime error: one of the variables needed for gradient computation has been modified by an inplace operation....
5. fix the file paths (the input_path) in params_*.py files.
6. fix the numpy type deprecation problem in hidden_recommend.py.