mirror of
https://github.com/vale981/ray
synced 2025-03-06 10:31:39 -05:00
46 lines
1.3 KiB
YAML
46 lines
1.3 KiB
YAML
|
# Pendulum SAC can attain -150+ reward in 6-7k
|
||
|
|
# Configurations are the similar to original softlearning/sac codebase
|
||
|
|
pendulum-cql:
|
||
|
|
env: Pendulum-v0
|
||
|
|
run: CQL
|
||
|
|
stop:
|
||
|
|
episode_reward_mean: -300
|
||
|
|
#timesteps_total: 10000
|
||
|
|
config:
|
||
|
|
# Works for both torch and tf.
|
||
|
|
framework: torch
|
||
|
|
#input: sampler
|
||
|
|
horizon: 200
|
||
|
|
soft_horizon: true
|
||
|
|
Q_model:
|
||
|
|
fcnet_activation: relu
|
||
|
|
fcnet_hiddens: [256, 256]
|
||
|
|
policy_model:
|
||
|
|
fcnet_activation: relu
|
||
|
|
fcnet_hiddens: [256, 256]
|
||
|
|
tau: 0.005
|
||
|
|
target_entropy: auto
|
||
|
|
no_done_at_end: true
|
||
|
|
n_step: 3
|
||
|
|
rollout_fragment_length: 1
|
||
|
|
prioritized_replay: false
|
||
|
|
train_batch_size: 256
|
||
|
|
target_network_update_freq: 1
|
||
|
|
timesteps_per_iteration: 1000
|
||
|
|
learning_starts: 256
|
||
|
|
optimization:
|
||
|
|
actor_learning_rate: 0.0003
|
||
|
|
critic_learning_rate: 0.0003
|
||
|
|
entropy_learning_rate: 0.0003
|
||
|
|
num_workers: 0
|
||
|
|
num_gpus: 1
|
||
|
|
clip_actions: False
|
||
|
|
normalize_actions: true
|
||
|
|
evaluation_num_workers: 1
|
||
|
|
evaluation_interval: 1
|
||
|
|
metrics_smoothing_episodes: 5
|
||
|
|
bc_iters: 0
|
||
|
|
evaluation_config:
|
||
|
|
input: sampler
|
||
|
|
|