gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2025-12-18 11:40:18 +01:00
Code Issues Projects Releases Wiki Activity

Add RedisDataStore (#295)

Browse Source 
* GraphManager.set_session also sets self.sess

* make sure that GraphManager.fetch_from_worker uses training phase

* remove unnecessary phase setting in training worker

* reorganize rollout worker

* provide default name to GlobalVariableSaver.__init__ since it isn't really used anyway

* allow dividing TrainingSteps and EnvironmentSteps

* add timestamps to the log

* added redis data store

* conflict merge fix
This commit is contained in:
Zach Dwiel
2019-08-28 14:15:58 -04:00
committed by shadiendrawis
parent 34e1c04f29
commit 7b0fccb041
18 changed files with 528 additions and 120 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
rl_coach/presets/Mujoco_PPO.py
View File

@@ -13,8 +13,8 @@ from rl_coach.graph_managers.graph_manager import ScheduleParameters
# Graph Scheduling #
####################
schedule_params = ScheduleParameters()
schedule_params.improve_steps = TrainingSteps(10000000000)
schedule_params.steps_between_evaluation_periods = EnvironmentSteps(2000)
schedule_params.improve_steps = TrainingSteps(1e10)
schedule_params.steps_between_evaluation_periods = EnvironmentSteps(4000)
schedule_params.evaluation_steps = EnvironmentEpisodes(1)
schedule_params.heatup_steps = EnvironmentSteps(0)
@@ -22,8 +22,8 @@ schedule_params.heatup_steps = EnvironmentSteps(0)
# Agent #
#########
agent_params = PPOAgentParameters()
agent_params.network_wrappers['actor'].learning_rate = 0.001
agent_params.network_wrappers['critic'].learning_rate = 0.001
agent_params.network_wrappers['actor'].learning_rate = 5e-5
agent_params.network_wrappers['critic'].learning_rate = 5e-5
agent_params.network_wrappers['actor'].input_embedders_parameters['observation'].scheme = [Dense(64)]
agent_params.network_wrappers['actor'].middleware_parameters.scheme = [Dense(64)]
@@ -33,6 +33,9 @@ agent_params.network_wrappers['critic'].middleware_parameters.scheme = [Dense(64
agent_params.input_filter = InputFilter()
agent_params.input_filter.add_observation_filter('observation', 'normalize', ObservationNormalizationFilter())
agent_params.algorithm.initial_kl_coefficient = 0.2
agent_params.algorithm.gae_lambda = 1.0
# Distributed Coach synchronization type.
agent_params.algorithm.distributed_coach_synchronization_type = DistributedCoachSynchronizationType.SYNC
@@ -55,5 +58,3 @@ preset_validation_params.trace_test_levels = ['inverted_pendulum', 'hopper']
graph_manager = BasicRLGraphManager(agent_params=agent_params, env_params=env_params,
schedule_params=schedule_params, vis_params=VisualizationParameters(),
preset_validation_params=preset_validation_params)