config_ma2c_cu_tragedy.ini

[MODEL_CONFIG]
rmsp_alpha = 0.99
rmsp_epsilon = 1e-5
max_grad_norm = 40
gamma = 0.99
lr_init = 5e-4
lr_decay = constant
entropy_coef = 0.05
value_coef = 0.5
num_lstm = 64
num_fc = 64
batch_size = 50
reward_norm = 0.025
reward_clip = -1

[TRAIN_CONFIG]
total_step = 1e4
test_interval = 5e5
log_interval = 1e4

[ENV_CONFIG]
agent = ma2c_cu
scenario = tragedy
n_agent = 3
episode_length_sec = 100
control_interval_sec = 1
coop_gamma = -1
seed = 0
test_seeds = 10000,20000,30000