def get_stats(self, key):
"""
Lets an algorithm ask the logger for mean/std/min/max of a diagnostic.
"""
v = self.epoch_dict[key]
vals = np.concatenate(v) if isinstance(v[0], np.ndarray) and len(v[0].shape)>0 else v
return mpi_statistics_scalar(vals)
def log_tabular(self, key, val=None, with_min_and_max=False, average_only=False):
"""
Log a value or possibly the mean/std/min/max values of a diagnostic.
Args:
key (string): The name of the diagnostic. If you are logging a
diagnostic whose state has previously been saved with
``store``, the key here has to match the key you used there.
val: A value for the diagnostic. If you have previously saved
values for this key via ``store``, do *not* provide a ``val``
here.
with_min_and_max (bool): If true, log min and max values of the
diagnostic over the epoch.
average_only (bool): If true, do not log the standard deviation
of the diagnostic over the epoch.
"""
if val is not None:
super().log_tabular(key,val)
else:
v = self.epoch_dict[key]
vals = np.concatenate(v) if isinstance(v[0], np.ndarray) and len(v[0].shape)>0 else v
stats = mpi_statistics_scalar(vals, with_min_and_max=with_min_and_max)
super().log_tabular(key if average_only else 'Average' + key, stats[0])
if not(average_only):
super().log_tabular('Std'+key, stats[1])
if with_min_and_max:
super().log_tabular('Max'+key, stats[3])
super().log_tabular('Min'+key, stats[2])
self.epoch_dict[key] = []
def get(self):
assert self.ptr == self.max_size
self.ptr, self.path_start_idx = 0, 0
# Normalization
adv_mean, adv_std = mpi_statistics_scalar(self.adv_buf)
self.adv_buf = (self.adv_buf - adv_mean) / adv_std
return [
self.obs_buf, self.act_buf, self.adv_buf, self.ret_buf,
self.logp_buf
]
def get(self):
"""
return all the samples in this epoch and ten start training
:return:
"""
assert self.ptr == self.max_size
self.ptr, self.path_start_idx = 0, 0
# advantage normalization
adv_mean, adv_std = mpi_statistics_scalar(self.adv_buf)
self.adv_buf = (self.adv_buf - adv_mean) / adv_std
return [
self.obs_buf, self.act_buf, self.adv_buf, self.ret_buf,
self.logp_buf
] + core.values_as_sorted_list(self.info_bufs)
def get(self):
"""
call at the end of epoch to get all data from buffer
:return:
"""
assert self.ptr == self.max_size
self.ptr, self.path_start_idx = 0, 0
# advantage normalization
adv_mean, adv_std = mpi_statistics_scalar(self.adv_buf)
self.adv_buf = (self.adv_buf - adv_mean) / adv_std
return [
self.obs_buf, self.act_buf, self.adv_buf, self.ret_buf,
self.logp_buf
]