for learning run it with: KERAS_BACKEND=theano THEANO_FLAGS=floatX=float32,device=gpu,force_device=True,cuda.root=/usr/local/cuda,lib.cnmem=0.2 python ./qlearn.py -m "Train" -l 1
if you want to use it with existing traing weights file run it with: KERAS_BACKEND=theano THEANO_FLAGS=floatX=float32,device=gpu,force_device=True,cuda.root=/usr/local/cuda,lib.cnmem=0.2 python ./qlearn.py -m "Run"
for learning by imitating player method run it with: KERAS_BACKEND=theano THEANO_FLAGS=floatX=float32,device=gpu,force_device=True,cuda.root=/usr/local/cuda,lib.cnmem=0.2 python ./qlearn_player.py -m "Train" -l 5
for learning by player combined with classical q learn method run it with: KERAS_BACKEND=theano THEANO_FLAGS=floatX=float32,device=gpu,force_device=True,cuda.root=/usr/local/cuda,lib.cnmem=0.2 python ./qlearn_player.py -m "Train" -l 6
you need to change the game file for different game