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

Testing...

Browse Source
This commit is contained in:
Dominik Moritz Roth 2022-08-06 14:37:30 +02:00
parent 54113fd40c
commit fcd9953b37
1 changed files with 17 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
test.py
View File

@ -1,8 +1,5 @@
#!/usr/bin/python3 #!/usr/bin/python3
import gym import gym
from gym.envs.registration import register
import numpy as np
import os
import time import time
import datetime import datetime
@ -67,15 +64,15 @@ def main(env_name='ColumbusCandyland_Aux10-v0', timesteps=1_000_000, showRes=Tru
# saveModel, n_eval_episodes) # saveModel, n_eval_episodes)
def full(env_name='ColumbusCandyland_Aux10-v0', timesteps=35_000, saveModel=True, n_eval_episodes=4): def full(env_name='ColumbusCandyland_Aux10-v0', timesteps=200_000, saveModel=True, n_eval_episodes=4):
env = gym.make(env_name) env = gym.make(env_name)
use_sde = False use_sde = False
skip_num = 8 # 10 (/ start at index) skip_num = 4 # 10 (/ start at index)
sac = True sac = False
Model = [PPO, SAC][sac] Model = [PPO, SAC][sac]
Policy = [MlpPolicyPPO, MlpPolicySAC][sac] Policy = [MlpPolicyPPO, MlpPolicySAC][sac]
#projection = FrobeniusProjectionLayer() projection = FrobeniusProjectionLayer()
projection = BaseProjectionLayer() #projection = BaseProjectionLayer()
gen = enumerate(get_legal_setups( gen = enumerate(get_legal_setups(
allowedEPTs=[EnforcePositiveType.SOFTPLUS, EnforcePositiveType.ABS])) allowedEPTs=[EnforcePositiveType.SOFTPLUS, EnforcePositiveType.ABS]))
@ -87,21 +84,21 @@ def full(env_name='ColumbusCandyland_Aux10-v0', timesteps=35_000, saveModel=True
model = Model( model = Model(
Policy, Policy,
env, env,
# projection=projection, projection=projection,
policy_kwargs={'dist_kwargs': {'neural_strength': ps, 'cov_strength': cs, 'parameterization_type': policy_kwargs={'dist_kwargs': {'neural_strength': ps, 'cov_strength': cs, 'parameterization_type':
pt, 'enforce_positive_type': ept, 'prob_squashing_type': ProbSquashingType.NONE}}, pt, 'enforce_positive_type': ept, 'prob_squashing_type': ProbSquashingType.TANH}},
verbose=0, verbose=0,
tensorboard_log=root_path+"/logs_tb/" + tensorboard_log=root_path+"/logs_tb/" +
env_name+"/"+['ppo', 'sac'][sac]+"_" + env_name+"/"+['ppo', 'sac'][sac]+"_" + 'TANH_' +
("_".join([str(s) for s in setup])+['', '_sde'][use_sde])+"/", ("_".join([str(s) for s in setup])+['', '_sde'][use_sde])+"/",
# learning_rate=3e-4, learning_rate=3e-4,
# gamma=0.99, gamma=0.99,
# gae_lambda=0.95, gae_lambda=0.95,
# normalize_advantage=True, normalize_advantage=True,
# ent_coef=0.02, # 0.1 ent_coef=0.02, # 0.1
# vf_coef=0.5, vf_coef=0.5,
use_sde=use_sde, # False use_sde=use_sde, # False
# clip_range=1 # 0.2, clip_range=1 # 0.2,
) )
testModel(model, timesteps, False, testModel(model, timesteps, False,
@ -152,5 +149,5 @@ if __name__ == '__main__':
# main('ColumbusJustState-v0') # main('ColumbusJustState-v0')
# main('ColumbusStateWithBarriers-v0') # main('ColumbusStateWithBarriers-v0')
# full('ColumbusEasierObstacles-v0') # full('ColumbusEasierObstacles-v0')
# full('ColumbusStateWithBarriers-v0') full('ColumbusStateWithBarriers-v0')
full('LunarLanderContinuous-v2') # full('LunarLanderContinuous-v2')