wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-10 00:35:08 +08:00
Code Issues Packages Projects Releases Wiki Activity

[RLlib] PPO(torch) on CartPole not tuned well enough for consistent learning (#7556)

Browse Source
This commit is contained in:
Sven Mika
2020-03-12 04:31:27 +01:00
committed by GitHub
parent 932a749fa9
commit bc120730e5
2 changed files with 19 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rllib/tuned_examples/regression_tests/cartpole-ppo-tf.yaml
+8 -1
View File
@@ -5,6 +5,13 @@ cartpole-ppo-tf:
episode_reward_mean: 150
timesteps_total: 100000
config:
gamma: 0.99
lr: 0.0003
num_workers: 1
batch_mode: complete_episodes
observation_filter: MeanStdFilter
num_sgd_iter: 6
vf_share_layers: true
vf_loss_coeff: 0.01
model:
fcnet_hiddens: [32]
fcnet_activation: linear
rllib/tuned_examples/regression_tests/cartpole-ppo-torch.yaml
+11 -4
View File
@@ -1,11 +1,18 @@
cartpole-ppo-tf:
cartpole-ppo-torch:
env: CartPole-v0
run: PPO
stop:
episode_reward_mean: 150
timesteps_total: 100000
config:
num_workers: 1
batch_mode: complete_episodes
observation_filter: MeanStdFilter
use_pytorch: true
gamma: 0.99
lr: 0.0003
num_workers: 1
observation_filter: MeanStdFilter
num_sgd_iter: 6
vf_share_layers: true
vf_loss_coeff: 0.01
model:
fcnet_hiddens: [32]
fcnet_activation: linear