API Change

exact n_episode for a list of n_episode limitation and save fake data in cache_buffer when self.buffer is None (#184)
add save_only_last_obs for replay buffer in order to save the memory. (#184)
remove default value in batch.split() and add merge_last argument (#185)
fix tensorboard logging: h-axis stands for env step instead of gradient step; add test results into tensorboard (#189)
add max_batchsize in onpolicy algorithms (#189)
keep only sumtree in segment tree implementation (#193)
add __contains__ and pop in batch: key in batch, batch.pop(key, deft) (#189)
remove dict return support for collector preprocess_fn (#189)
remove **kwargs in ReplayBuffer (#189)
add no_grad argument in collector.collect (#204)

Enhancement

add DQN Atari examples (#187)
change the type-checking order in batch.py and converter.py in order to meet the most often case first (#189)
Numba acceleration for GAE, nstep, and segment tree (#193)
add policy.eval() in all test scripts' "watch performance" (#189)
add test_returns (both GAE and nstep) (#189)
improve the code-coverage (from 90% to 95%) and remove the dead code (#189)
polish examples/box2d/bipedal_hardcore_sac.py (#207)

fix a bug in MAPolicy: buffer.rew = Batch() doesn't change buffer.rew (thanks mypy) (#207)
~~set policy.eval() before collector.collect (#204)~~ This is a bug
fix shape inconsistency for torch.Tensor in replay buffer (#189)
potential bugfix for subproc.wait (#189)
fix RecurrentActorProb (#189)
fix some incorrect type annotation (#189)
fix a bug in tictactoe set_eps (#193)
dirty fix for asyncVenv check_id test