gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2025-12-17 19:20:19 +01:00
Code Issues Projects Releases Wiki Activity

benchmark update (#250)

Browse Source
This commit is contained in:
shadiendrawis
2019-03-17 15:33:28 +02:00
committed by GitHub
parent 537b549e1d
commit f03bd7ad93
4 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

BIN
benchmarks/acer/breakout_acer_16_workers.png
View File

Binary file not shown.
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 66 KiB

After

Width:  |  Height:  |  Size: 65 KiB

BIN
benchmarks/acer/space_invaders_acer_16_workers.png
View File

Binary file not shown.
Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 65 KiB

After

Width:  |  Height:  |  Size: 60 KiB

2
rl_coach/agents/acer_agent.py
View File

@@ -68,7 +68,7 @@ class ACERAlgorithmParameters(AlgorithmParameters):
self.num_steps_between_gradient_updates = 5000 self.num_steps_between_gradient_updates = 5000
self.ratio_of_replay = 4 self.ratio_of_replay = 4
self.num_transitions_to_start_replay = 10000 self.num_transitions_to_start_replay = 10000
self.rate_for_copying_weights_to_target = 0.99 self.rate_for_copying_weights_to_target = 0.01
self.importance_weight_truncation = 10.0 self.importance_weight_truncation = 10.0
self.use_trust_region_optimization = True self.use_trust_region_optimization = True
self.max_KL_divergence = 1.0 self.max_KL_divergence = 1.0

2
rl_coach/presets/Atari_ACER.py
View File

@@ -11,7 +11,7 @@ from rl_coach.memories.memory import MemoryGranularity
# Graph Scheduling # # Graph Scheduling #
#################### ####################
schedule_params = ScheduleParameters() schedule_params = ScheduleParameters()
schedule_params.improve_steps = TrainingSteps(10000000000) schedule_params.improve_steps = TrainingSteps(10000000)
schedule_params.steps_between_evaluation_periods = EnvironmentEpisodes(100) schedule_params.steps_between_evaluation_periods = EnvironmentEpisodes(100)
schedule_params.evaluation_steps = EnvironmentEpisodes(3) schedule_params.evaluation_steps = EnvironmentEpisodes(3)
schedule_params.heatup_steps = EnvironmentSteps(0) schedule_params.heatup_steps = EnvironmentSteps(0)