Files
ray/python/ray/rllib/tuned_examples/hopper-ppo.yaml
T
Eric LiangandGitHub b41bdcefa0 [rllib] Update RLlib to work with new actor scheduling behavior (#1754)
* Mon Mar 19 21:23:01 PDT 2018

* Mon Mar 19 21:23:07 PDT 2018

* Mon Mar 19 21:30:49 PDT 2018

* Mon Mar 19 21:32:05 PDT 2018

* Mon Mar 19 21:35:43 PDT 2018

* fix cpu limits

* Mon Mar 19 22:25:07 PDT 2018
2018-03-20 19:29:52 -07:00

10 lines
472 B
YAML

hopper-ppo:
env: Hopper-v1
run: PPO
resources:
cpu: 65
gpu: 4
driver_cpu_limit: 1
driver_gpu_limit: 4
config: {"gamma": 0.995, "kl_coeff": 1.0, "num_sgd_iter": 20, "sgd_stepsize": .0001, "sgd_batchsize": 32768, "devices": ["/gpu:0", "/gpu:1", "/gpu:2", "/gpu:3"], "tf_session_args": {"device_count": {"GPU": 4}, "log_device_placement": false, "allow_soft_placement": true}, "timesteps_per_batch": 160000, "num_workers": 64}