Codes for the paper arXiv:...
Requirement:
- python 3.8
- CUDA 11.0
Install all required packages
- Clone this repo and change the direction to this repo in your terminal.
- Run
pip install -r Requirements.txtto install all required packages.
net_train.py is the code to train the CAE.
There are three parameters for you to make a better a training:
T: Step of training set.
BC: Step of training.
XLLC: How many turn need to train.
We have show a set of value of these three parameters in the file, and you do not need to change them, you can just use it directly.
When finish training, we can get the encoder mod-lx1-100-1.pt and decoder mod-lx2-100-1.pt. (We have save them in the 'mod-lx1-100-1.pt' and 'mod-lx2-100-1.pt.', you can use them directly.)
Use python accuracy on test set.py to test the accuracy of the CAE.