dodox/metastable-baselines
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Fixed a bug with KL-proj

Browse Source
This commit is contained in:
Dominik Moritz Roth 2022-08-28 20:48:02 +02:00
parent 4080ad8135
commit 2f05474091
1 changed files with 4 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
test.py
View File

@ -27,9 +27,10 @@ def main(env_name='ColumbusCandyland_Aux10-v0', timesteps=1_000_000, showRes=Tru
ppo = PPO(
MlpPolicyPPO,
env,
projection=BaseProjectionLayer(), # KLProjectionLayer(trust_region_coeff=0.01),
policy_kwargs={'dist_kwargs': {'neural_strength': Strength.NONE, 'cov_strength': Strength.FULL, 'parameterization_type':
ParametrizationType.CHOL, 'enforce_positive_type': EnforcePositiveType.ABS, 'prob_squashing_type': ProbSquashingType.NONE}},
# KLProjectionLayer(trust_region_coeff=0.01),
projection=KLProjectionLayer(trust_region_coeff=0.01),
policy_kwargs={'dist_kwargs': {'neural_strength': Strength.NONE, 'cov_strength': Strength.DIAG, 'parameterization_type':
ParametrizationType.NONE, 'enforce_positive_type': EnforcePositiveType.ABS, 'prob_squashing_type': ProbSquashingType.NONE}},
verbose=0,
tensorboard_log=root_path+"/logs_tb/" +
env_name+"/ppo"+(['', '_sde'][use_sde])+"/",