Skip to content

Releases: agi-brain/xuance

XuanCe (1.2.4)

11 Dec 12:43
Compare
Choose a tag to compare
  • Synchronized updated to the MindSpore based on version 1.2.3.

  • Added support for multi-GPU distributed training.

  • Updated VDAC, COMA, MFQ, MFAC, DCG, QTRAN, and other algorithms for version of 1.2.x.

  • Added IAC multi-agent reinforcement learning baseline algorithm.

  • Improved the training mechanism for multi-agent adversarial tasks.

  • Further standardized and optimized the underlying APIs.

Full Changelog: v1.0.11...v1.2.4

XuanCe (1.2.3)

11 Dec 12:42
Compare
Choose a tag to compare
  • Synchronized updates to the MindSpore version based on version 1.2.2.

  • Further standardized and optimized the underlying APIs.

XuanCe (1.2.2)

11 Dec 12:41
Compare
Choose a tag to compare
  • Synchronized updates to the TensorFlow 2 version based on version 1.2.1.

  • Refactored the Agent module, distinguishing between on-policy and off-policy algorithms.

  • Further standardized and optimized the underlying APIs.

XuanCe (1.2.1)

11 Dec 12:35
Compare
Choose a tag to compare
  • Update the MARL algorithms based on version 1.2.0.

  • Add support for non-parameter-sharing in multi-agent systems, with individual agent models indexed by name.

  • Improve the algorithms like MADDPG and MASAC with support for RNNs.

XuanCe (1.2.0)

11 Dec 12:33
Compare
Choose a tag to compare
  • Modified the environment wrapper program to facilitate adding new environments.

  • Standardized the wrapping of existing single-agent and multi-agent environments.

  • Organized commonly used APIs into a unified structure.

  • Integrated the creation of representation, policy, and optimizer into each agent’s initialization method.

  • Added support for non-parameter-sharing in MARL algorithms.

  • Standardized the naming of parameters in config files.

XuanCe (1.1.1)

12 May 09:40
Compare
Choose a tag to compare
  • Improve some algorithms, such as SAC, MASAC, DDPG, and TD3, etc.
  • Add multi-robot warehouse (RWARE) environment.
  • n_size -> buffer_size;
  • The saving and loading of observation status;
  • Unify the names of attributes for both single-agent and multi-agent DRL.

Full Changelog: v.1.2.0...v1.1.1

XuanCe (1.0.11)

11 Apr 10:28
Compare
Choose a tag to compare

Updates for XuanCe v1.0.11:

  1. Support and finish test for gym-pybullet-drones environments;
  2. Fix some issues for installation of xuance. Now it is more easy to install and use;
  3. Improve the compatibility for MacOS with Apple's M chips;
  4. Fix some other bugs.

Welcome to install and try it!

XuanCe (1.0.10)

05 Mar 13:30
Compare
Choose a tag to compare

Updates for XuanCe v1.0.10:

  1. Fix some bugs for MARL;
  2. Support Minigrid and gym-pybullet-drones environments;
  3. Provide examples of customized environments for MARL;
  4. Some other issues are solved.

Welcome to install and try it!

XuanCe (1.0.6)

25 Dec 10:23
Compare
Choose a tag to compare

A stable version of XuanCe.

XuanCe (1.0.1)

27 Nov 02:23
Compare
Choose a tag to compare

Beta version for MARL with SMAC and GFootball.