Skip to content

Latest commit

 

History

History
56 lines (40 loc) · 1.22 KB

pg.rst

File metadata and controls

56 lines (40 loc) · 1.22 KB
Raw

VPG

Example

 from rlzoo.common.env_wrappers import build_env
 from rlzoo.common.utils import call_default_params
 from rlzoo.algorithms import PG

 AlgName = 'PG'
 EnvName = 'PongNoFrameskip-v4'
 EnvType = 'atari'

 # EnvName = 'CartPole-v0'
 # EnvType = 'classic_control'

 # EnvName = 'BipedalWalker-v2'
 # EnvType = 'box2d'

 # EnvName = 'Ant-v2'
 # EnvType = 'mujoco'

 # EnvName = 'FetchPush-v1'
 # EnvType = 'robotics'

 # EnvName = 'FishSwim-v0'
 # EnvType = 'dm_control'

 # EnvName = 'ReachTarget'
 # EnvType = 'rlbench'

 env = build_env(EnvName, EnvType)
 alg_params, learn_params = call_default_params(env, EnvType, AlgName)
 alg = eval(AlgName+'(**alg_params)')
 alg.learn(env=env, mode='train', render=False, **learn_params)
 alg.learn(env=env, mode='test', render=True, **learn_params)

Vanilla Policy Gradient

.. autoclass:: rlzoo.algorithms.pg.pg.PG
   :members:
   :undoc-members:

Default Hyper-parameters

.. automodule:: rlzoo.algorithms.pg.default
   :members:
   :undoc-members:
   :show-inheritance: