mirror of
https://github.com/vale981/ray
synced 2025-03-06 02:21:39 -05:00
83e06cd30a
* WIP. * WIP. * WIP. * WIP. * WIP. * Fix * WIP. * Add TD3 quick Pendulum regresison. * Cleanup. * Fix. * LINT. * Fix. * Sort quick_learning test cases, add TD3. * Sort quick_learning test cases, add TD3. * Revert test_checkpoint_restore.py (debugging) changes. * Fix old soft_q settings in documentation and test configs. * More doc fixes. * Fix test case. * Fix test case. * Lower test load. * WIP.
57 lines
1.5 KiB
YAML
57 lines
1.5 KiB
YAML
# This configuration can expect to reach 90 reward in 10k-20k timesteps
|
|
mountaincarcontinuous-ddpg:
|
|
env: MountainCarContinuous-v0
|
|
run: DDPG
|
|
stop:
|
|
episode_reward_mean: 90
|
|
time_total_s: 600 # 10 minutes
|
|
config:
|
|
# === Model ===
|
|
actor_hiddens: [32, 64]
|
|
critic_hiddens: [64, 64]
|
|
n_step: 3
|
|
model: {}
|
|
gamma: 0.99
|
|
env_config: {}
|
|
|
|
# === Exploration ===
|
|
exploration_config:
|
|
initial_scale: 1.0
|
|
final_scale: 0.02
|
|
scale_timesteps: 40000
|
|
ou_base_scale: 0.75
|
|
ou_theta: 0.15
|
|
ou_sigma: 0.2
|
|
|
|
timesteps_per_iteration: 1000
|
|
|
|
target_network_update_freq: 0
|
|
tau: 0.01
|
|
|
|
# === Replay buffer ===
|
|
buffer_size: 50000
|
|
prioritized_replay: False
|
|
prioritized_replay_alpha: 0.6
|
|
prioritized_replay_beta: 0.4
|
|
prioritized_replay_eps: 0.000001
|
|
clip_rewards: False
|
|
|
|
# === Optimization ===
|
|
actor_lr: 0.001
|
|
critic_lr: 0.001
|
|
use_huber: False
|
|
huber_threshold: 1.0
|
|
l2_reg: 0.00001
|
|
learning_starts: 1000
|
|
sample_batch_size: 1
|
|
train_batch_size: 64
|
|
|
|
# === Parallelism ===
|
|
num_workers: 0
|
|
num_gpus_per_worker: 0
|
|
per_worker_exploration: False
|
|
worker_side_prioritization: False
|
|
|
|
# === Evaluation ===
|
|
evaluation_interval: 5
|
|
evaluation_num_episodes: 10
|