gryf/coach
1
0
Fork 0
mirror of https://github.com/gryf/coach.git synced 2026-03-11 03:55:52 +01:00
Code Issues Projects Releases Wiki Activity

ACER algorithm (#184)

Browse Source 
* initial ACER commit

* Code cleanup + several fixes

* Q-retrace bug fix + small clean-ups

* added documentation for acer

* ACER benchmarks

* update benchmarks table

* Add nightly running of golden and trace tests. (#202)

Resolves #200

* comment out nightly trace tests until values reset.

* remove redundant observe ignore (#168)

* ensure nightly test env containers exist. (#205)

Also bump integration test timeout

* wxPython removal (#207)

Replacing wxPython with Python's Tkinter.
Also removing the option to choose multiple files as it is unused and causes errors, and fixing the load file/directory spinner.

* Create CONTRIBUTING.md (#210)

* Create CONTRIBUTING.md.  Resolves #188

* run nightly golden tests sequentially. (#217)

Should reduce resource requirements and potential CPU contention but increases
overall execution time.

* tests: added new setup configuration + test args (#211)

- added utils for future tests and conftest
- added test args

* new docs build

* golden test update
This commit is contained in:
shadiendrawis
2019-02-20 23:52:34 +02:00
committed by GitHub
parent 7253f511ed
commit 2b5d1dabe6
175 changed files with 2327 additions and 664 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
rl_coach/memories/episodic/episodic_experience_replay.py
View File

@@ -75,18 +75,27 @@ class EpisodicExperienceReplay(Memory):
def num_transitions_in_complete_episodes(self):
return self._num_transitions_in_complete_episodes
def sample(self, size: int) -> List[Transition]:
def sample(self, size: int, is_consecutive_transitions=False) -> List[Transition]:
"""
Sample a batch of transitions form the replay buffer. If the requested size is larger than the number
Sample a batch of transitions from the replay buffer. If the requested size is larger than the number
of samples available in the replay buffer then the batch will return empty.
:param size: the size of the batch to sample
:param is_consecutive_transitions: if set True, samples a batch of consecutive transitions.
:return: a batch (list) of selected transitions from the replay buffer
"""
self.reader_writer_lock.lock_writing()
if self.num_complete_episodes() >= 1:
transitions_idx = np.random.randint(self.num_transitions_in_complete_episodes(), size=size)
batch = [self.transitions[i] for i in transitions_idx]
if is_consecutive_transitions:
episode_idx = np.random.randint(0, self.num_complete_episodes())
if self._buffer[episode_idx].length() <= size:
batch = self._buffer[episode_idx].transitions
else:
transition_idx = np.random.randint(size, self._buffer[episode_idx].length())
batch = self._buffer[episode_idx].transitions[transition_idx-size:transition_idx]
else:
transitions_idx = np.random.randint(self.num_transitions_in_complete_episodes(), size=size)
batch = [self.transitions[i] for i in transitions_idx]
else:
raise ValueError("The episodic replay buffer cannot be sampled since there are no complete episodes yet. "