How to set up to train continuous action? #2
Description
I set the startup script as follows
th main.lua -seed 1 -server false -resume true -discrete false -gpu 0 -game_config quickrace_continuous_slowmulti_ushite-city.xml -env TORCS.TorcsContinuousConstDamagePos -threads 2 -model models.Discrete_action -hiddenSize 256 -height 84 -width 84 -batchSize 5 -momentum 0.99 -rmsEpsilon 0.1 -steps 20000000 -optimiser sharedRmsProp -eta 0.00007 -gradClip 0 -rewardClip 0 -valSteps 18750 -valFreq 250000 -tensorType torch.FloatTensor -entropyBeta 0.01 -progFreq 5000 -verbose false -foldername train_TORCS_slowmulti_ushite_city -group 0 "$@"
But the system is stuck, is the model set to models.Discrete_action? What should the correct setting look like?