CHANGELOG.md

[v0.0.8] - 21/04/2024

Fix import errors for matplotlib and seaborn visualization.

[v0.0.7] - 04/04/2024

1. Fixed most of the typing issues under pylint --strict. 1.1 This required unifying the interface of step, step_env, get_obs, 'is_terminal` for proper inheritance.

2. spaces now depends on gymnasium instead of gym, adding an additional dependency. Maybe later on gym dependency can be completely removed since it's not maintained anymore? haven't looked into that.

3. Removed _DeviceArray from the tests as it's deprecated. Uses jax version 0.4.24

4. Ran everything in python 3.10 so technically you could support that too now.

5. In wrapper/brax.py, GymnaxtoBraxWrapper has two new methods backend and observation_size, as its required to be defined under newest brax version. Used brax version 0.10.0.

Thanks to @Aidandos !

[v0.0.6] - 12/04/2023

Added

• Gym, EvoJAX, Brax, DM env wrappers
• Reacher environment inspired by Lenton et al. (2021)
• Swimmer environment inspired by Lenton et al. (2021)
• Basic Pong environment inspired by Kirsch (2018)

Fixed

• Fixed Minatar tests for jax arrays.
• Fixed reward setting in DiscountingChain-v0
• Fixed Tuple space check via enumerate.

Changed

• Refactored wrappers into separate sub-directory.

[v0.0.5] - 24/08/2022

Fixed

• Fix deprecated tree_multimap.
• Fix device grabbing when using an jnp.array to set default in Acrobot env.

[v0.0.3] - 15/06/2022

Fixed

• Fix import structure.

[v0.0.2] - 15/06/2022

Added

• Fully functional API using flax-style EnvState and EnvParams for calling env.step.
• MinAtar environments are not operational yet.
• Release to ensure that evosax 0.0.9 release can work with GymFitness backend in gymnax.

Changed

• Basically everything :)

[v0.0.1] - 22/11/2021

Added

• Adds main gym-like infrastructure that is jit, vmap & pmap compatible:
  • Jittable environment base class: environment.py
  • Spaces: Discrete, Continuous, Box
  • Translation of numpy state to state dictionary for JAX step.
• Adds base set of environments:
  • OpenAI's classic_control: Pendulum-v0, CartPole-v0, MountainCar-v0, ContinuousMountainCar-v0, Acrobot-v0
  • DeepMind's bsuite: Catch-bsuite, DeepSea-bsuite, DiscountingChain-bsuite, MemoryChain-bsuite, UmbrellaChain-bsuite, MNISTBandit-bsuite, SimpleBandit-bsuite
  • MinAtar: Asterix-MinAtar, Breakout-MinAtar, Freeway-MinAtar, Seaquest-MinAtar, SpaceInvaders-MinAtar
• Adds tests for comparing gym/numpy reset + step with JAX version.
  • tests/classic_control/test_gym_env.py
  • tests/bsuite/test_bsuite_env.py
• Adds set of notebooks walking through the individual environments.
• Adds set of examples incorporating gymnax into JAX-based RL experiments.
  • Anakin agent - port DM Colab for gymnax
  • CMA-ES policy evolution - port blogpost with experimental minimal agent

Todo

• Adds benchmark infrastructure and numbers on different devices (CPU/GPU/TPU):
  • Transitions/Second & Specific rollout types vs Torch setup
    • CPU: Intel Xeon 2.4 GHz
    • GPU: V100, A100, RTX 2080Ti
    • TPU: V2, V3 - vmap + pmap
• Adds set of experimental utilities:
  • dojos: Multi-transition rollout wrapper via lax.scan + jit for sequential step.
  • agents: Minimal and evaluation agent wrappers.
• Adds miscellaneous environments: Bandit-misc, Rooms-misc