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

update nec and value optimization agents to work with recurrent middleware

Browse Source
This commit is contained in:
Zach Dwiel
2017-11-03 13:58:42 -07:00
parent 93a54c7e8e
commit 6c79a442f2
12 changed files with 138 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
presets.py
View File

@@ -1,5 +1,5 @@
#
# Copyright (c) 2017 Intel Corporation
# Copyright (c) 2017 Intel Corporation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@@ -907,6 +907,19 @@ class Doom_Health_DQN(Preset):
self.agent.num_steps_between_copying_online_weights_to_target = 1000
class Pong_NEC_LSTM(Preset):
def __init__(self):
Preset.__init__(self, NEC, Atari, ExplorationParameters)
self.env.level = 'PongDeterministic-v4'
self.learning_rate = 0.001
self.agent.num_transitions_in_experience_replay = 1000000
self.agent.middleware_type = MiddlewareTypes.LSTM
self.exploration.initial_epsilon = 0.5
self.exploration.final_epsilon = 0.1
self.exploration.epsilon_decay_steps = 1000000
self.num_heatup_steps = 500
class Pong_NEC(Preset):
def __init__(self):
Preset.__init__(self, NEC, Atari, ExplorationParameters)