cpprb

cpprb v10.1.0 can save & load transitions

From cpprb version 10.1.0, ReplayBuffer and its sub-classes can save and load transitions. from cpprb import ReplayBuffer rb1 = ReplayBuffer(256, {"obs": {"shape": 3}, "act": {}, "rew": {}, "done": {}}, next_of="obs")

ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject

When import cpprb, I got ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject. Recently, NumPy has changed its ABI at version

Speed up Ape-X implementation on single machine

Alghough neural network is optimized for GPU, environments for reinforcement learning (e.g. simulator) are not always GPU friendly. One of the method to speed up reinforcement learning is to run

Benchmark and comparison with other replay buffer implementation

For a long time, I wanted to take benchmark of cpprb and compare other replay buffer implementations. Finally, today (24th Feb. 2020) I released comparison result pages. By searching online,

Introduce new helper function for creating ReplayBuffer easier.

From version 8.3, cpprb introduces new helper functions create_env_dict and create_before_add_func, which enable users to use ReplayBuffer class easier. create_env_dict The function create_env_dict takes gym.Env (environment) and returns dict which

setup.py for Cython / Numpy based project

This post is a English version of this Japanese post. One of the common troubles for developing Cython/ Numpy based C/C++ extension package is setup.py. Ideally, all packages should be

cpprb

Replay buffer classes for reinforcement learning.

Binary distribution using manylinux

Happily, cpprb started binary package distribution for Linux at PyPI. Unlike macOS and Windows, Linux has a lot of distros. Unfortunately setuptools names these wheel binary on Linux just as

New cpprb logo!

From this post, I will write blog in English. Several days ago, I changed the cpprb logo. The reason I changed was that I would like to match the logo

cpprb のテスト環境をDockerfile化

cpprb の v8 へのメジャーアップデートに向けてテスト環境をDocke