ray/rllib/tuned_examples/regression_tests/cartpole-ars.yaml

16 lines
377 B
YAML

cartpole-ars:
env: CartPole-v0
run: ARS
stop:
episode_reward_mean: 50
timesteps_total: 500000
config:
noise_stdev: 0.02
num_rollouts: 50
rollouts_used: 25
num_workers: 2
sgd_stepsize: 0.01
noise_size: 25000000
eval_prob: 0.5
model:
fcnet_hiddens: [] # a linear policy