gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2026-03-14 05:35:55 +01:00
Code Issues Projects Releases Wiki Activity

Enable distributed SharedRunningStats (#81)

Browse Source 
- Use Redis pub/sub for updating SharedRunningStats.
This commit is contained in:
Balaji Subramaniam
2018-11-13 09:17:38 -08:00
committed by Gal Leibovich
parent 875d6ef017
commit a849c17e46
10 changed files with 76 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
rl_coach/presets/CartPole_ClippedPPO.py
View File

@@ -5,7 +5,6 @@ from rl_coach.core_types import TrainingSteps, EnvironmentEpisodes, EnvironmentS
from rl_coach.environments.gym_environment import GymVectorEnvironment, mujoco_v2
from rl_coach.exploration_policies.additive_noise import AdditiveNoiseParameters
from rl_coach.exploration_policies.e_greedy import EGreedyParameters
from rl_coach.filters.observation.observation_normalization_filter import ObservationNormalizationFilter
from rl_coach.graph_managers.basic_rl_graph_manager import BasicRLGraphManager
from rl_coach.graph_managers.graph_manager import ScheduleParameters
from rl_coach.schedules import LinearSchedule
@@ -49,8 +48,6 @@ agent_params.algorithm.distributed_coach_synchronization_type = DistributedCoach
agent_params.exploration = EGreedyParameters()
agent_params.exploration.epsilon_schedule = LinearSchedule(1.0, 0.01, 10000)
# agent_params.pre_network_filter.add_observation_filter('observation', 'normalize_observation',
# ObservationNormalizationFilter(name='normalize_observation'))
###############
# Environment #

2
rl_coach/presets/CartPole_DQN.py
View File

@@ -1,5 +1,5 @@
from rl_coach.agents.dqn_agent import DQNAgentParameters
from rl_coach.base_parameters import VisualizationParameters, PresetValidationParameters, DistributedCoachSynchronizationType
from rl_coach.base_parameters import VisualizationParameters, PresetValidationParameters
from rl_coach.core_types import TrainingSteps, EnvironmentEpisodes, EnvironmentSteps
from rl_coach.environments.gym_environment import GymVectorEnvironment
from rl_coach.graph_managers.basic_rl_graph_manager import BasicRLGraphManager

4
rl_coach/presets/Mujoco_ClippedPPO.py
View File

@@ -1,6 +1,6 @@
from rl_coach.agents.clipped_ppo_agent import ClippedPPOAgentParameters
from rl_coach.architectures.layers import Dense
from rl_coach.base_parameters import VisualizationParameters, PresetValidationParameters
from rl_coach.base_parameters import VisualizationParameters, PresetValidationParameters, DistributedCoachSynchronizationType
from rl_coach.core_types import TrainingSteps, EnvironmentEpisodes, EnvironmentSteps
from rl_coach.environments.environment import SingleLevelSelection
from rl_coach.environments.gym_environment import GymVectorEnvironment, mujoco_v2
@@ -43,6 +43,8 @@ agent_params.algorithm.gae_lambda = 0.95
agent_params.algorithm.discount = 0.99
agent_params.algorithm.optimization_epochs = 10
agent_params.algorithm.estimate_state_value_using_gae = True
# Distributed Coach synchronization type.
agent_params.algorithm.distributed_coach_synchronization_type = DistributedCoachSynchronizationType.SYNC
agent_params.input_filter = InputFilter()
agent_params.exploration = AdditiveNoiseParameters()

5
rl_coach/presets/Mujoco_PPO.py
View File

@@ -1,6 +1,6 @@
from rl_coach.agents.ppo_agent import PPOAgentParameters
from rl_coach.architectures.layers import Dense
from rl_coach.base_parameters import VisualizationParameters, PresetValidationParameters
from rl_coach.base_parameters import VisualizationParameters, PresetValidationParameters, DistributedCoachSynchronizationType
from rl_coach.core_types import TrainingSteps, EnvironmentEpisodes, EnvironmentSteps
from rl_coach.environments.environment import SingleLevelSelection
from rl_coach.environments.gym_environment import GymVectorEnvironment, mujoco_v2
@@ -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())
# Distributed Coach synchronization type.
agent_params.algorithm.distributed_coach_synchronization_type = DistributedCoachSynchronizationType.SYNC
###############
# Environment #
###############