Component for training the AI.
Argument 1: match config as json Argument 2: team config for left team as json Argument 3: team config for right team as json Argument 4: learning rate Argument 5: discount rate
Argument 6: estimator config as json (will be used for both teams)
Argument 6: Directory with sub directories containing the experience data Argument 7: Frequency of experience epochs (every <value> epoch will be an experience replay epoch)
Argument 6: estimator config as json (will be used for both teams) Argument 7: Directory with sub directories containing the experience data Argument 8: Frequency of experience epochs (every <value> epoch will be an experience replay epoch)
You can choose between using Docker or manually installing all dependencies. Docker is the preferred method as it already installs the toolchain and all dependencies.
In the root directory of the project build the docker image ("kitraining" is the name of the container, this can be replaced by a different name):
docker build -t kitraining .
Now start the container, you need to map the external file (match.json) to an internal file:
docker run -v $(pwd)/match.json:match.json kitraining ./KiTraining /match.json
That's it you should now have a running docker instance.
If you need to debug the program it can be easier to do this outside of docker.
- A C++17 compatible Compiler (e.g. GCC-8)
- CMake (min 3.10) and GNU-Make
- Adress-Sanitizer for run time checks
- SopraGameLogic
- SopraMessages
- SopraUtil
- SopraAITools
- MLP
- Either a POSIX-Compliant OS or Cygwin (to use pthreads)
- Optional: Google Tests and Google Mock for Unit-Tests
In the root directory of the project create a new directory (in this example it will be called build), change in this directory.
Next generate a makefile using cmake:
cmake ..
if any error occurs recheck the prerequisites. Next compile the program:
make
you can now run the program by executing the created KiTraining file:
./KiTraining
Das Analyseergebniss von SonarQube ist hier auf SonarCloud zu finden.