2020-11-30 12:41:24 +01:00
|
|
|
cartpole-sac:
|
|
|
|
|
env: CartPoleContinuousBulletEnv-v0
|
|
|
|
|
run: SAC
|
|
|
|
|
stop:
|
2021-03-26 11:45:17 -07:00
|
|
|
episode_reward_mean: 40
|
2020-11-30 12:41:24 +01:00
|
|
|
timesteps_total: 100000
|
|
|
|
|
config:
|
|
|
|
|
# Works for both torch and tf.
|
|
|
|
|
framework: tf
|
2021-03-15 14:03:54 +01:00
|
|
|
gamma: 0.95
|
|
|
|
|
no_done_at_end: false
|
|
|
|
|
horizon: 200
|
2021-03-29 17:20:29 +02:00
|
|
|
soft_horizon: true
|
2021-03-15 14:03:54 +01:00
|
|
|
n_step: 3
|
2022-05-17 13:43:49 +02:00
|
|
|
replay_buffer_config:
|
|
|
|
|
type: MultiAgentPrioritizedReplayBuffer
|
|
|
|
|
learning_starts: 256
|
2021-03-29 17:20:29 +02:00
|
|
|
initial_alpha: 0.2
|
2021-03-15 14:03:54 +01:00
|
|
|
clip_actions: false
|
2022-06-10 17:09:18 +02:00
|
|
|
min_sample_timesteps_per_iteration: 1000
|
2021-03-29 17:20:29 +02:00
|
|
|
optimization:
|
|
|
|
|
actor_learning_rate: 0.005
|
|
|
|
|
critic_learning_rate: 0.005
|
|
|
|
|
entropy_learning_rate: 0.0001
|
