Change Log

Unreleased

v10.4.2

  • Fix: Nstep accepts JAX array

v10.4.1

  • Fix: GitHub Action binary build

v10.4.0

  • Add: Support JAX array (aka. jaxlib.xla_extension.DeviceArray) input.

v10.3.2

  • Fix: macOS binary build

v10.3.1

  • Add: Binary distibution for macOS x8664, and Linux AArch64

v10.3.0

  • Add: ReverseReplayBuffer

v10.2.0

  • Add: LaBERmean, LaBERlazy, and LaBERmax

v10.1.1

  • Fix: Numpy version at build from source

v10.1.0

  • Add: save_transitions / load_transitions (138)

v10.0.0

  • Breaking Change: Nstep no longer returns discounts (137)

v9.4.6

  • Fix: stackcompress cache (135)

v9.4.5

  • Fix: NumPy ABI change (131)

v9.4.4

  • Fix: PrioritizedReplayBuffer, MPPrioritizedReplayBuffer support non C-convertible dtype (e.g. np.float16) (130)

v9.4.3

  • Fix: PrioritizedReplayBuffer, MPPrioritizedReplayBuffer adds read only priority (128)

v9.4.2

  • Fix: MPPrioritizedReplayBuffer lock mechanism, which makes faster

v9.4.1

  • Fix: MPReplayBuffer and MPPrioritizedReplayBuffer with spawn multiprocessing

v9.4.0

  • Add: MPReplayBuffer and MPPrioritizedReplayBuffer

v9.3.3

  • Fix: Walkaround next_of at Nstep
  • Fix: Input dict modification (116)

v9.3.2

  • Fix: CodeQL reported warnings
    • Unused import (1,2)
    • Empty except (1)

v9.3.1

  • Fix: cache for nextof and stackcompress (114)

v9.3.0

  • Add: get_all_trainsitions(shuffle=True) for shuffle transitions

v9.2.1

  • Fix: Windows binary distribution

v9.2.0

  • Add: Large buffer exceeding memory using mmap

v9.1.2

  • Fix: Logger writer multiple same massages
  • Fix: buffer dtypes were mis-converted (112)

v9.1.1

  • Fix: Add Warning to train
  • Fix: Embed signature at train

v9.1.0

  • Add: New free function train for simple train loop (beta)

v9.0.5

  • Fix: Nstep PrioritizedReplayBuffer raises exception at on_episode_end (111)

v9.0.4

  • Fix: Nstep PrioritizedReplayBuffer raises exception at on_episode_end (111) This was not fixed correctly.

v9.0.3

  • Fix: Cache for "next_of" when episode is smaller than stack (110)

v9.0.2

  • Fix: Cache for "next_of" when "stack_compress" at last item (108)
  • Remove: author email address from setup.py, which is not necessary for PyPI

v9.0.1

  • Fix: Cache for "next_of" with "stack_compress" at non last item (108)

v9.0.0

  • Remove: cpprb.gym sub-package, which is not cpprb main purpose

v8.4.8

  • Fix: Support PyVirtualDisplay 1.0 above.
  • Fix: Try to acquire physical memory in ReplayBuffer initialization (#105)

v8.4.7

  • Add: Pruning update chain in Segment Tree (#103)

v8.4.6

  • Fix: Windows binary build on GitHub Actions

v8.4.5

  • Fix: Cache non last transitions. (#97)
  • Fix: Clear caches when ReplayBuffer.clear (#98)

v8.4.4

  • Fix: PER priority shape mismatch (#96)

v8.4.3

  • Fix: Unrequire done (#61)

v8.4.2

  • Fix: Unrequire gym

v8.4.1

  • Fix: PyPI description

v8.4.0

  • Add: ReplayBuffer.get_all_transitions

v8.3.2

  • Fix: Add gym into requirements, since create_env_dict, create_before_func use

v8.3.1

  • Fix: GitHub Action settings (Recover Windows binary)

v8.3.0

  • Add: Hepler functions create_env_dict, create_before_func

v8.2.5

  • Fix: GitHub Actions syntax

v8.2.4

  • Fix: README on PyPI (overwritten upload from GitHub)

v8.2.3

  • Fix: Binary upload on Github Actions

v8.2.2

  • Add: Distribute binary of Windows and Linux

v8.2.1

  • Fix: gym module is comparison with literal
  • Fix: Build without pre-install cython

v8.2.0

  • Add: debug build with environment DEBUG_CPPRB

v8.1.3

  • Fix: compile argument at setup.py for Windows.

v8.1.2

  • Fix: Missing <numeric> header at ReplayBuffer.hh

v8.1.1

  • Fix: Accept float shape at PrioritizedReplayBuffer.unchangesincesample

v8.1.0

  • Add: check_for_update for PrioritizedReplayBuffer