1
0
mirror of https://github.com/gryf/coach.git synced 2026-02-15 05:25:55 +01:00

introduce dockerfiles. (#169)

* introduce dockerfiles.

* ensure golden tests are run not just collected.

* Skip CI download of dockerfiles.

* add StarCraft environment and tests.

* add minimaps starcraft validation parameters.

* Add functional test running (from Ayoob)

* pin mujoco_py version to a 1.5 compatible release.

* fix config syntax issue.

* pin remaining mujoco_py install calls.

* Relax pin of gym version in gym Dockerfile.

* update makefile based on functional test filtering.
This commit is contained in:
Scott Leishman
2019-04-03 09:33:17 -07:00
committed by anabwan
parent 0b808f0794
commit f173e69187
11 changed files with 459 additions and 33 deletions

View File

@@ -1,7 +1,7 @@
from rl_coach.agents.actor_critic_agent import ActorCriticAgentParameters
from rl_coach.agents.policy_optimization_agent import PolicyGradientRescaler
from rl_coach.architectures.embedder_parameters import InputEmbedderParameters
from rl_coach.base_parameters import VisualizationParameters
from rl_coach.base_parameters import VisualizationParameters, PresetValidationParameters
from rl_coach.core_types import TrainingSteps, EnvironmentEpisodes, EnvironmentSteps
from rl_coach.environments.starcraft2_environment import StarCraft2EnvironmentParameters
from rl_coach.exploration_policies.additive_noise import AdditiveNoiseParameters
@@ -53,5 +53,17 @@ env_params = StarCraft2EnvironmentParameters(level='CollectMineralShards')
env_params.feature_screen_maps_to_use = [5]
env_params.feature_minimap_maps_to_use = [5]
########
# Test #
########
preset_validation_params = PresetValidationParameters()
preset_validation_params.test = True
preset_validation_params.min_reward_threshold = 50
preset_validation_params.max_episodes_to_achieve_reward = 200
preset_validation_params.num_workers = 1
graph_manager = BasicRLGraphManager(agent_params=agent_params, env_params=env_params,
schedule_params=schedule_params, vis_params=VisualizationParameters())
schedule_params=schedule_params,
vis_params=VisualizationParameters(),
preset_validation_params=preset_validation_params)

View File

@@ -3,7 +3,7 @@ from collections import OrderedDict
from rl_coach.agents.ddqn_agent import DDQNAgentParameters
from rl_coach.architectures.embedder_parameters import InputEmbedderParameters
from rl_coach.architectures.head_parameters import DuelingQHeadParameters
from rl_coach.base_parameters import VisualizationParameters
from rl_coach.base_parameters import VisualizationParameters, PresetValidationParameters
from rl_coach.core_types import TrainingSteps, EnvironmentEpisodes, EnvironmentSteps
from rl_coach.environments.starcraft2_environment import StarCraft2EnvironmentParameters
from rl_coach.filters.action.box_discretization import BoxDiscretization
@@ -53,6 +53,16 @@ env_params = StarCraft2EnvironmentParameters(level='CollectMineralShards')
env_params.feature_screen_maps_to_use = [5]
env_params.feature_minimap_maps_to_use = [5]
########
# Test #
########
preset_validation_params = PresetValidationParameters()
preset_validation_params.test = True
preset_validation_params.min_reward_threshold = 50
preset_validation_params.max_episodes_to_achieve_reward = 200
preset_validation_params.num_workers = 1
graph_manager = BasicRLGraphManager(agent_params=agent_params, env_params=env_params,
schedule_params=schedule_params, vis_params=VisualizationParameters())
schedule_params=schedule_params, vis_params=VisualizationParameters(),
preset_validation_params=preset_validation_params)