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

Fixed bug with initialization of buffer

Browse Source
This commit is contained in:
Dominik Moritz Roth 2022-07-01 20:02:09 +02:00
parent 0dc9edf112
commit 91f64c10d7
1 changed files with 5 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
metastable_baselines/misc/rollout_buffer.py
View File

@ -34,17 +34,18 @@ class GaussianRolloutBuffer(RolloutBuffer):
n_envs: int = 1, n_envs: int = 1,
cov_shape=None, cov_shape=None,
): ):
super().__init__(buffer_size, observation_space, action_space,
device, n_envs=n_envs, gae_lambda=gae_lambda, gamma=gamma)
self.means, self.stds = None, None self.means, self.stds = None, None
# TODO: Correct shape for full cov matrix # TODO: Correct shape for full cov matrix
# self.action_space.shape + self.action_space.shape # self.action_space.shape + self.action_space.shape
if cov_shape == None: if cov_shape == None:
cov_shape = self.action_space.shape cov_shape = action_space.shape
self.cov_shape = cov_shape self.cov_shape = cov_shape
# It is ugly, but necessary to put this at the bottom of the init...
super().__init__(buffer_size, observation_space, action_space,
device, n_envs=n_envs, gae_lambda=gae_lambda, gamma=gamma)
def reset(self) -> None: def reset(self) -> None:
self.means = np.zeros( self.means = np.zeros(
(self.buffer_size, self.n_envs) + self.action_space.shape, dtype=np.float32) (self.buffer_size, self.n_envs) + self.action_space.shape, dtype=np.float32)