steffen/FIKS_Entwicklungsumgebung
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

pomdp

Browse Source
This commit is contained in:
steffen-illium
2021-06-01 12:39:33 +02:00
parent 403d38dc24
commit 55b409c72f
6 changed files with 82 additions and 83 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
environments/logging/monitor.py
View File

@ -1,6 +1,5 @@
import pickle
from pathlib import Path
from collections import defaultdict
from stable_baselines3.common.callbacks import BaseCallback
@ -9,51 +8,6 @@ from environments.logging.plotting import prepare_plot
import pandas as pd
class FactoryMonitor:
def __init__(self, env):
self._env = env
self._monitor = defaultdict(lambda: defaultdict(lambda: 0))
self._last_vals = defaultdict(lambda: 0)
def __iter__(self):
for key, value in self._monitor.items():
yield key, dict(value)
def add(self, key, value, step=None):
assert step is None or step >= 1 # Is this good practice?
step = step or self._env.steps
self._last_vals[key] = self._last_vals[key] + value
self._monitor[key][step] = self._last_vals[key]
return self._last_vals[key]
def set(self, key, value, step=None):
assert step is None or step >= 1 # Is this good practice?
step = step or self._env.steps
self._last_vals[key] = value
self._monitor[key][step] = self._last_vals[key]
return self._last_vals[key]
def remove(self, key, value, step=None):
assert step is None or step >= 1 # Is this good practice?
step = step or self._env.steps
self._last_vals[key] = self._last_vals[key] - value
self._monitor[key][step] = self._last_vals[key]
return self._last_vals[key]
def to_dict(self):
return dict(self)
def to_pd_dataframe(self):
import pandas as pd
df = pd.DataFrame.from_dict(self.to_dict())
df.fillna(0)
return df
def reset(self):
raise RuntimeError("DO NOT DO THIS! Always initalize a new Monitor per Env-Run.")
class MonitorCallback(BaseCallback):
ext = 'png'
@ -62,6 +16,7 @@ class MonitorCallback(BaseCallback):
super(MonitorCallback, self).__init__()
self.filepath = Path(filepath)
self._monitor_df = pd.DataFrame()
self._monitor_dict = dict()
self.env = env
self.plotting = plotting
self.started = False
@ -113,12 +68,17 @@ class MonitorCallback(BaseCallback):
self.closed = True
def _on_step(self) -> bool:
for _, info in enumerate(self.locals.get('infos', [])):
self._monitor_dict[self.num_timesteps] = {key: val for key, val in info.items()
if key not in ['terminal_observation', 'episode']}
for env_idx, done in enumerate(self.locals.get('dones', [])):
if done:
env_monitor_df = self.locals['infos'][env_idx]['monitor'].to_pd_dataframe()
env_monitor_df = pd.DataFrame.from_dict(self._monitor_dict, orient='index')
self._monitor_dict = dict()
columns = [col for col in env_monitor_df.columns if col not in IGNORED_DF_COLUMNS]
env_monitor_df = env_monitor_df.aggregate(
{col: 'mean' if 'amount' in col or 'count' in col else 'sum' for col in columns}
{col: 'mean' if col.endswith('ount') else 'sum' for col in columns}
)
env_monitor_df['episode'] = len(self._monitor_df)
self._monitor_df = self._monitor_df.append([env_monitor_df])