training_parameters.yaml

ACTOR_LR : 0.00001                 # learning rate for actor net
CRITIC_LR : 0.00001                # learning rate for critic net
GAMMA : 0.99                       # decay factor of reward, range from 0.9 to 0.99
TAU : 0.01                         # parameter for soft updating
MEMORY_SIZE : 1000000              # size of replay memory
MEMORY_WARMUP_SIZE : 100000        # this parameter decide how many pieces of data we need to restore before training
REWARD_SCALE : 0.01                # scaling factor for reward
BATCH_SIZE : 256                   # batch size
TRAIN_TOTAL_STEPS : 1000000        # total steps for training
TEST_EVERY_STEPS : 10000           # intervals we test the model
EPSILON : 0.1                      # epsilon greed