gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2026-03-31 00:53:32 +02:00
Code Issues Projects Releases Wiki Activity

network_imporvements branch merge

Browse Source
This commit is contained in:
Shadi Endrawis
2018-10-02 13:41:46 +03:00
parent 72ea933384
commit 51726a5b80
110 changed files with 1639 additions and 1161 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
rl_coach/presets/BitFlip_DQN.py
View File

@@ -1,11 +1,10 @@
from rl_coach.agents.dqn_agent import DQNAgentParameters
from rl_coach.architectures.tensorflow_components.architecture import Dense
from rl_coach.architectures.tensorflow_components.embedders.embedder import InputEmbedderParameters
from rl_coach.architectures.tensorflow_components.layers import Dense
from rl_coach.base_parameters import VisualizationParameters, EmbedderScheme, \
PresetValidationParameters
from rl_coach.architectures.tensorflow_components.embedders.embedder import InputEmbedderParameters
from rl_coach.core_types import TrainingSteps, EnvironmentEpisodes, EnvironmentSteps
from rl_coach.environments.gym_environment import Mujoco
from rl_coach.environments.gym_environment import GymVectorEnvironment
from rl_coach.graph_managers.basic_rl_graph_manager import BasicRLGraphManager
from rl_coach.graph_managers.graph_manager import ScheduleParameters
from rl_coach.memories.memory import MemoryGranularity
@@ -28,7 +27,7 @@ schedule_params.heatup_steps = EnvironmentSteps(0)
agent_params = DQNAgentParameters()
agent_params.network_wrappers['main'].learning_rate = 0.001
agent_params.network_wrappers['main'].batch_size = 128
agent_params.network_wrappers['main'].middleware_parameters.scheme = [Dense([256])]
agent_params.network_wrappers['main'].middleware_parameters.scheme = [Dense(256)]
agent_params.network_wrappers['main'].input_embedders_parameters = {
'state': InputEmbedderParameters(scheme=EmbedderScheme.Empty),
'desired_goal': InputEmbedderParameters(scheme=EmbedderScheme.Empty)
@@ -45,12 +44,10 @@ agent_params.exploration.evaluation_epsilon = 0
###############
# Environment #
###############
env_params = Mujoco()
env_params.level = 'rl_coach.environments.toy_problems.bit_flip:BitFlip'
env_params = GymVectorEnvironment(level='rl_coach.environments.toy_problems.bit_flip:BitFlip')
env_params.additional_simulator_parameters = {'bit_length': bit_length, 'mean_zero': True}
# env_params.custom_reward_threshold = -bit_length + 1
env_params.custom_reward_threshold = -bit_length + 1
vis_params = VisualizationParameters()
########
# Test #
@@ -61,7 +58,7 @@ preset_validation_params.min_reward_threshold = -7.9
preset_validation_params.max_episodes_to_achieve_reward = 10000
graph_manager = BasicRLGraphManager(agent_params=agent_params, env_params=env_params,
schedule_params=schedule_params, vis_params=vis_params,
schedule_params=schedule_params, vis_params=VisualizationParameters(),
preset_validation_params=preset_validation_params)