Bug fixes

famura · Aug 24, 2020 · 48c975d · 48c975d
1 parent e84a9da
commit 48c975d
Show file tree

Hide file tree

Showing 6 changed files with 15 additions and 13 deletions.
diff --git a/Pyrado/pyrado/__init__.py b/Pyrado/pyrado/__init__.py
@@ -35,7 +35,7 @@
 
 
 # Pyrado version number
-VERSION = '0.1'
+VERSION = '0.2'
 
 # Provide global data directories
 PERMA_DIR = osp.join(osp.dirname(__file__), '..', 'data', 'perma')

diff --git a/Pyrado/pyrado/exploration/stochastic_params.py b/Pyrado/pyrado/exploration/stochastic_params.py
@@ -151,11 +151,13 @@ def __init__(self,
  super().__init__(param_dim)
 
  if full_cov:
- self._noise = FullNormalNoise(noise_dim=param_dim, std_init=std_init, std_min=std_min,
- train_mean=train_mean)
+ self._noise = FullNormalNoise(
+ use_cuda=False, noise_dim=param_dim, std_init=std_init, std_min=std_min, train_mean=train_mean
+ )
  else:
- self._noise = DiagNormalNoise(noise_dim=param_dim, std_init=std_init, std_min=std_min,
- train_mean=train_mean)
+ self._noise = DiagNormalNoise(
+ use_cuda=False, noise_dim=param_dim, std_init=std_init, std_min=std_min, train_mean=train_mean
+ )
 
  @property
  def noise(self) -> [FullNormalNoise, DiagNormalNoise]:

diff --git a/Pyrado/scripts/training/qbb_spota_hc.py b/Pyrado/scripts/training/qbb_spota_hc.py
@@ -81,10 +81,10 @@
  expl_factor=1.1,
  expl_std_init=0.5,
  )
- subrtn_hparam_cand = subrtn_hparam_cand
+ subrtn_hparam_refs = deepcopy(subrtn_hparam_cand)
 
  sr_cand = HCNormal(ex_dir, env, policy, **subrtn_hparam_cand)
- sr_refs = HCNormal(ex_dir, env, deepcopy(policy), **subrtn_hparam_cand)
+ sr_refs = HCNormal(ex_dir, env, deepcopy(policy), **subrtn_hparam_refs)
 
  spota_hparam = dict(
  max_iter=10,
@@ -109,7 +109,7 @@
  save_list_of_dicts_to_yaml([
  dict(env=env_hparams, seed=ex_dir.seed),
  dict(policy=policy_hparam),
- dict(subrtn_cand=subrtn_hparam_cand, subrtn_refs=subrtn_hparam_cand, subrtn_name=HCNormal.name),
+ dict(subrtn_name=HCNormal.name, subrtn_cand=subrtn_hparam_cand, subrtn_refs=subrtn_hparam_refs),
  dict(algo=spota_hparam, algo_name=SPOTA.name)],
  ex_dir
  )

diff --git a/Pyrado/scripts/training/qbb_spota_ppo.py b/Pyrado/scripts/training/qbb_spota_ppo.py
@@ -108,10 +108,10 @@
  max_grad_norm=1.,
  lr=1e-4,
  )
- subrtn_hparam_cand = subrtn_hparam_cand
+ subrtn_hparam_refs = deepcopy(subrtn_hparam_cand)
 
  sr_cand = PPO(ex_dir, env, policy, critic_cand, **subrtn_hparam_cand)
- sr_refs = PPO(ex_dir, env, deepcopy(policy), critic_refs, **subrtn_hparam_cand)
+ sr_refs = PPO(ex_dir, env, deepcopy(policy), critic_refs, **subrtn_hparam_refs)
 
  # Meta-Algorithm
  spota_hparam = dict(
@@ -139,7 +139,7 @@
  dict(env=env_hparams, seed=ex_dir.seed),
  dict(policy=policy_hparam),
  dict(critic_cand_and_ref=critic_hparam),
- dict(subrtn_cand=subrtn_hparam_cand, subrtn_refs=subrtn_hparam_cand, subrtn_name=PPO.name),
+ dict(subrtn_name=PPO.name, subrtn_cand=subrtn_hparam_cand, subrtn_refs=subrtn_hparam_refs),
  dict(algo=spota_hparam, algo_name=SPOTA.name)],
  ex_dir
  )

diff --git a/Pyrado/scripts/training/wam-bic_bayrn_cem.py b/Pyrado/scripts/training/wam-bic_bayrn_cem.py
@@ -115,7 +115,7 @@
  num_eval_rollouts_real=100 if isinstance(env_real, WAMBallInCupSim) else 5,
  num_eval_rollouts_sim=100,
  # policy_param_init=policy_init.param_values.data,
- subroutine_snapshot_mode='latest'
+ subrtn_snapshot_mode='latest'
  )
 
  # Save the environments and the hyper-parameters (do it before the init routine of BDR)

diff --git a/RcsPySim/setup.py b/RcsPySim/setup.py
@@ -89,7 +89,7 @@ def build_extension(self, ext):
 
 setup(
  name='RcsPySim',
- version='0.1',
+ version='0.2',
  description='Robotic simulation environments backed by Rcs',
  author='Fabio Muratore & Felix Treede',
  author_email='fabio.muratore@famura.net',