2020-09-09 00:34:34 -07:00
|
|
|
hopper-mbmpo:
|
|
|
|
|
env: ray.rllib.examples.env.mbmpo_env.HopperWrapper
|
|
|
|
|
run: MBMPO
|
|
|
|
|
stop:
|
|
|
|
|
training_iteration: 500
|
|
|
|
|
config:
|
2020-10-06 20:28:16 +02:00
|
|
|
# Only supported in torch right now.
|
2020-09-09 00:34:34 -07:00
|
|
|
framework: torch
|
|
|
|
|
# 200 in paper, 1000 will take forever
|
|
|
|
|
horizon: 200
|
|
|
|
|
num_envs_per_worker: 20
|
|
|
|
|
inner_adaptation_steps: 1
|
|
|
|
|
maml_optimizer_steps: 8
|
|
|
|
|
gamma: 0.99
|
|
|
|
|
lambda: 1.0
|
|
|
|
|
lr: 0.001
|
|
|
|
|
clip_param: 0.5
|
|
|
|
|
kl_target: 0.003
|
|
|
|
|
kl_coeff: 0.0000000001
|
|
|
|
|
num_workers: 20
|
|
|
|
|
num_gpus: 1
|
|
|
|
|
inner_lr: 0.001
|
|
|
|
|
clip_actions: False
|
|
|
|
|
num_maml_steps: 15
|
|
|
|
|
model:
|
|
|
|
|
fcnet_hiddens: [32, 32]
|
|
|
|
|
free_log_std: True
|
