Skip to content

ReinforcementLearningZoo-v0.9.0
Compare
Choose a tag to compare
@github-actions github-actions released this 07 Mar 09:36
· 36 commits to main since this release
ReinforcementLearningZoo-v0.9.0
a6623a8
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

ReinforcementLearningZoo ReinforcementLearningZoo-v0.9.0

Merged pull requests:

Closed issues:

  • A3C (#133)
  • Box2D environment (#2)
  • Visualize episodes (#3)
  • Implement TRPO/ACER (#134)
  • ERROR: UndefVarError: AtariPreprocessor not defined (#6)
  • bullet3 environment (#7)
  • vizdoom environment (#8)
  • Error tagging new release (#11)
  • loadenvironment error (#19)
  • Support alternative deep learning libraries (#20)
  • Random Thoughts on v0.3.0 (#24)
  • ViZDoom is broken (#130)
  • Document basic environments (#129)
  • bullet3 environment (#128)
  • Box2D environment (#127)
  • Prioritized DQN (#27)
  • Improve interfaces for model exploration and hyperparameter optimization (#28)
  • A2C (#32)
  • Add built-in support for TensorBoard (#35)
  • Add checkpoints (#36)
  • Improve code coverage (#40)
  • AbstractActionSelector not exported (#41)
  • Params empty - no tracking (#43)
  • Add reproducible examples for Atari environments (#44)
  • Add procgen (#126)
  • StopAfterEpisode with progress meter (#51)
  • Add MAgent (#125)
  • Regret Policy Gradients (#131)
  • Support SEED RL (SCALABLE AND EFFICIENT DEEP-RL ) (#62)
  • Support julia 1.4 (#63)
  • How to define a new environment? (#64)
  • Roadmap of v0.9 (#65)
  • Add MCTS related algorithms (#132)
  • Question about AbstractEnv API (#68)
  • bsuite (#124)
  • Classic environments in separate package? (#123)
  • Implement Fully Parameterized Quantile Function for Distributional Reinforcement Learning. (#135)
  • Failed to precompile ReinforcementLearning (#71)
  • Experimental support of Torch.jl (#136)
  • depends on HDF5? (#72)
  • warning and error (#73)
  • Compatibility issue in ReinforcementLearning & Flux (#74)
  • ERROR: KeyError: key "ArnoldiMethod" not found (#79)
  • Add dueling DQN (#137)
  • In DDPG: Add support for vector actions (#138)
  • Rename AbstractAgent to AbstractPolicy (#111)
  • How should ReinforcementLearning.jl be cited ? (#80)
  • Add a stop condition to terminate the experiment after reaching reward threashold (#112)
  • Add Game 2048 (#122)
  • add CUDA accelerated Env (#121)
  • Unify common network architectures and patterns (#139)
  • Alternative handling of max steps in environment (#140)
  • I get NNlib error when trying to load a model (#82)
  • "convert" warning (#83)
  • Seg fault on macbook pro (#84)
  • ACME RL lib by deepmind (#85)
  • Add Highway env (#120)
  • Definition of a policy (#86)
  • Add remote trajectories (#87)
  • Add experiments based on offline RL data (#141)
  • Asynchronous Methods for Deep Reinforcement Learning (#142)
  • Base.convert method for DiscreteSpace (#104)
  • Action Space Meaning (#88)
  • Base.in method for EmptySpace (#105)
  • Renaming get_terminal to isterminated (#106)
  • Requesting more informative field names for SharedTrajectory (#113)
  • R2D2 (#143)
  • Suggestion: More informative name for FullActionSet & MinimalActionSet (#107)
  • Returning an AbstractSpace object using get_actions (#108)
  • Recurrent Models (#144)
  • Split experiments into separate files (#145)
  • Add project.toml for tests (#146)
  • Migrate to Pluto (#148)
  • Update dependency to the latest version of ReinforcementLearning.jl (#149)
  • Docs build error (#91)
  • Split out Trajectory & CircularArrayBuffer as independent packages (#114)
  • Requesting explanation for better performance at ... (#115)
  • Add an extra mode when evaluating agent (#116)
  • Why are wrapper environments a part of RLBase instead of RLCore (say)? (#109)
  • Purpose and scope of sub-packages (#93)
  • The names of keyword arguments in Trajectory is kind of misunderstanding (#117)
  • Add experiments with GymEnv (#147)
  • Cross language support (#103)
  • Check compatibility between agent and environments (#118)
  • Behaviour for hooks for RewardOverridenEnv (#119)
  • StopAfterEpisode with custom DQNL errors beyond a particular Episode Count (#96)
  • ERROR: UndefVarError: NNlib not defined while loading agent (#110)
  • Support compression? (#102)
  • State monitoring and fault tolerance (#101)
  • Use JLSO for (de)serialization? (#97)
  • Add an example running in K8S (#100)
  • Setup github actions (#98)
  • Fails to load trajectory (#150)
  • Replace Travis with github actions (#151)
  • Test error in ReinforcementLearningEnvironments.jl (#152)
  • Move preallocations in MultiThreadEnv from global to local (#153)
  • Flux as service (#154)
  • remove @views (#155)
  • error in save & load ElasticCompactSARTSATrajectory (#156)
  • add early stopping in src\core\stop_conditions.jl (#157)
  • add time stamp in load & save function, in file src\components\agents\agent.jl (#158)
  • policies in GPU can not be saved || loaded (#159)
  • AbstractStage docstring doesn't render correctly in docs. (#160)
  • List of contributors (#161)
  • code formatting (#165)
  • Purpose of CommonRLInterface (#166)
  • Moving example environments from RLBase to RLEnvs? (#167)
  • Keeping prefix get_ in method names like get_reward (#168)
  • Currently getting an ambiguous method error in ReinforcementLearningCore v0.5.1 (#171)
  • Return experiment instead of hook only (#173)
  • TD3 Implementation (#174)
  • Train policy with GymEnv (#175)
  • Travis CI Credits (#178)
  • Training mode and testing mode (#179)
  • Unrecognized symbols (#180)
  • AbstractEnv (#181)
  • SARTTrajectory for SAC (#182)
  • define environment of FULL_ACTION_SET (#184)
  • CircularArraySLARTTrajectory instance is not of type CircularArraySLARTTrajectory (#185)
  • TagBot trigger issue (#186)
  • Is hook the same thing as "callback"? (#190)
  • Use @threads instead of @sync + @Spawn in MultiThreadEnv? (#191)
  • Blog custom env link typo (#192)
  • Change clip_by_global_norm! into a Optimizer (#193)
  • PPO related algorithms are broken (#194)
  • Add card game environments (#196)
  • Separate envs from algos in Zoo? (#197)
  • Why "examples"? (#198)
  • WandB integration? (#201)
  • Add default implementations for AbstractEnvWrapper (#202)
  • Add configuration in DQNLearner to enable double-dqn by default (#205)
  • Derivative-Free Reinforcement Learning (#206)
  • ERROR: type RandomPolicy has no field policy (#208)
  • Why split repos? (#209)
  • "Getting Started" too long imo (#210)
  • PreActStage clarification (#212)
  • What's a "trace"? (#213)
  • Continuous time supported? (#215)
  • Docs looks ugly in dark mode (#217)
  • Julia 1.6.0 dependency problem with ReinforcementLearningBase/RLBase (#221)
  • Add Discrete Batch-Constrained Deep Q-learning (#226)
  • Docstring of DoEveryNStep (#225)
  • Documentation of environment; actions seems not work. (#222)
  • Documentation of "How to use Tensorboard?": with_logger not defined (#223)
  • Getting figure object; How to get an animation using GR.plot in CartPolEnv (#246)
  • Update dependency to [email protected] and resolve type piracy of findmax (#227)
  • IQN is broken with [email protected] (#228)
  • The components of Rainbow (#229)
  • Source links in documentation directs to julia repo (#231)
  • Support Tables.jl and PrettyTables.jl for Trajectories (#232)
  • code in get_started seems to be broken (#233)
  • PPO strange behaviour from having actions as one element arrays instead of scalar (#234)
  • SAC and GaussianNetwork (#236)
  • Precompilation prohibitively long (#237)
  • Document how to save/load parameters (#238)
  • Workflow of saving trajectory data (#239)
  • An explanation of "how to train policy (agent)" such as Basic_DQN would be valuable (#240)
  • How to guarantee the environment's reproducibility? (#241)
  • [Call for Contributors] Summer 2021 of Open Source Promotion Plan (#242)
  • Cannot use RLBase.action_space etc. when writing my own environment (#244)
  • ReinforcementLearningZoo.jl experiments (#245)
  • Next Release Plan (v0.9) (#247)
  • How about making this package compatible with DifferentialEquations.jl? (#249)
  • Reinforcement Learning and Combinatorial Optimization (#250)
  • PPO and multi dimensional actions spaces (#251)
  • Add ReinforcementLearningDatasets (#253)
  • Incompatibility with CSVFiles.jl (#256)
  • [RLEnvs] easy access of the length of an action vector (dimension of action space) (#257)
  • Cannot add LinearAlgebra (#258)
  • What's the checkpoints? (#261)
  • Model based reinforcement learning (#262)
  • Add a dedicated multi-dimensional space type (#268)
  • PPO is broken when using CUDA (#280)
  • Lack of reproducibility of QRDQN CartPole Experiment. (#281)
  • Reinforcement Learning.jl in a RTS (#291)
  • StopAfterNoImprovement hook test fails occasionally (#297)
  • Get error when using ReinforcementLearning (#298)
  • Problems with PGFPlotsX during the install (#301)
  • Plotting CartPole environment in Jupyter (#306)
  • Support CircularVectorSARTTrajectory RLZoo (#316)
  • Local development environment setup tips causing error (#320)
  • Rename some functions to help beginners navigate source code (#326)
  • Question about PER (#328)
  • Docs error in code output (#332)
  • Setup a CI for typo (#336)
  • double code & dysfunctional master branch when downloading package (#341)
  • Support multiple discrete action space (#347)
  • Precompilation error; using Plots makes a conflict (#349)
  • Problem with running initial tutorial. Using TabularPolicy() generates an UndefinedKeyword error for n_action (#354)
  • Question: Clarification on the RL plots generated by the run() function (#357)
  • prob question for QBasedPolicy (#360)
  • Can evaluate function be used as a component of RLcore? (#369)
  • problem about precompiling the forked package (#377)
  • Question: Can we use packages like DifferentialEquations.jl to evolve or model the environment in ReinforcementLearning.jl (#378)
  • Combine transformers and RL (#392)
  • MultiAgentManager does not select correct action space for RockPaperScissorsEnv (#393)
  • Add ReinforcementLearningDatasets.jl (#397)
  • error: dimension mismatch "cannot broadcast array to have fewer dimensions" (#400)
  • SAC policy problems? (#410)
  • Add pre-training hook (#411)
  • Dead links in documentation (#418)
  • Links of show nbview badges in RLExperiments are incorrect (#421)
  • Problem accessing public google cloud storage bucket for RLDatasets.jl (#424)
  • Function to access base env through multiple wrapper layers (#425)
  • The problem of using GaussianNetwork in gpu (#455)
  • Next Release Plan (v0.10) (#460)
  • Error in experiment "JuliaRL_DDPG_Pendulum" (#471)
  • In Windows, ReinforcementLearningDataset.jl encounter a bug (#485)
  • Conditional Testing (#493)
  • Inconsistency of the EpsilonGreedyExplorer selection function (#520)
  • PyCall.getindex in module ReinforcementLearningEnvironments conflict warning (#527)
  • device method definition overwritten (#530)
  • How to display/render AtariEnv? (#546)
  • StackFrames bug? (#551)
  • Refactor of DQN Algorithms (#557)
  • Small performance improvement (#558)
  • Infinite-recursion bug in function is_discrete_space when an object of type ClosedInterval is passed (#566)
  • JuliaRL_BasicDQN_CartPole example fails (#568)
  • action_space not defined in tutorial (#569)
  • CI fails with [email protected] (#572)
  • Warning while precompiling RLCore due to kwargs (#575)
  • Gain in VPGPolicy does not account for terminal states? (#578)
  • Strange Bug with examples CartPoleEnv and RLBase.test_runnable!(RandomWalk1D) (#579)
  • Missing docs for TDLearner (#580)
  • Difficulty Creating a Custom Environment (#581)
  • Missing docs for how to implement a new algorithm (#582)
  • Donation (#595)
  • MultiThreadEnv with custom (continuous) action spaces fails (#596)
  • PPOCartPole fails, source of error included (#605)
  • Question: Can ReinforcementLearning.jl handle Partially Observed Markov Processes (POMDPs)? (#608)
  • Add an environment wrapper to IsaacGym (#619)
  • Explain current implementation of PPO in detail (#620)
  • How to run this source code in vscode? (#623)
  • Bug: Issue with TD3 for multi-dimensional action spaces (#624)
  • Make documentation on trace normalization (#633)
  • ActionTransformedEnv doesn't transform legal_action_space_mask (#642)
  • Bug: Previous example from RLZoo now has a bug (#643)
  • TDLearner time step parameter (#648)
  • Examples of multidimensional continous actions (#676)
  • estimate v.s. basis in policies (#677)
  • Base.copy not implemented for the TicTacToe environment (#678)
  • Broken link to src (#693)
  • Support Brax (#696)
  • Q-learning update timing (#702)
  • PPO on environments with multiple action dimensions? (#703)
  • Can't checkout RLCore for development (#704)
  • various eligibility trace-equipped TD methods (#709)
  • Improve the logging mechanism during training (#725)
  • questions while looking at implementation of VPG (#729)
  • Setup sponsor related info (#730)
  • new _run() (#731)
  • SAC example experiment does not work (#736)
  • Custom environment action and state space explanation (#738)
  • PPOPolicy training: ERROR: DomainError with NaN: Normal: the condition σ >= zero(σ) is not satisfied. (#739)
  • Code Readability (#740)
  • MultiThreadEnv not available in ReinforcementLearningZoo (#741)
  • ReinforcementLearningExperiment dependencies fail to precompile (#744)
  • tanh normalization destabilizes learning with GaussianNetwork (#745)
  • how to load a saved model and test it? (#755)
  • Custom Environment Passes RLBase.test_runnable!(env) but infinite hangs and crashes when run. (#757)
  • Bounds Error at UPDATE_FREQ Step (#758)
  • StopAfterEpisode returns 1 more episode using StepsPerEpisode() hook (#759)
  • Move basic definition of environment wrapper into RLBase (#760)
  • Precompilation error - DomainSets not in dependencies (#761)
  • How to set RLBase.state_space() if the number of state space is uncertain (#762)
  • Collect both number of steps and rewards in a single hook (#763)
  • how to use MultiAgentManager on different algorithms? (#764)
  • Example run of Offline RL that totally depends on dataset without online environment (#765)
  • Every single environment / experiment crashes with following error: (#766)
  • Neural Network Approximator based policies not working (#770)
  • Deep RL example for LSTM (#772)
  • "params not defined," "JuliaRL_BasicDQN_CartPole" (#778)
  • MonteCarloLearner incorrect PreActStage behavior (#779)
  • Prioritised Experience Replay (#780)
  • Outdated dependencies (#781)
  • Running experiments throws a "UndefVarError: params not defined" message (#784)
  • Failing MPOCovariance experiment (#791)
  • Logo image was not found (#836)
  • Reactivate docs on CI/CD (#838)
  • update docs: Tips for developers (#844)
  • Package dependencies not compatible (#860)
  • need help from an expert (#862)
  • Installing ReinforcementLearning.jl downgrades Flux.jl (#864)
  • Fix Experiment type setup (#881)
  • LoadError: UndefVarError: params not defined (#882)
  • Rename update! to push! (#883)
  • Contribute Neural Fitted Q-iteration algorithm (#895)
  • PPo policy experiments failing (#910)
  • Executing RLBase.plan! after end of experiment (#913)
  • EpisodeSampler in Trajectories (#927)
  • Hook RewardsPerEpisode broken (#945)
  • Can implement this ARZ algorithm ? (#965)
  • AssertionError: action in env.action_space (#967)
  • Fixing SAC Policy (#970)
  • Prioritized DQN experiment nonfunctional (#971)
  • Prioritised DQN failing on GPU (#973)
  • An error (#983)
  • params() is no longer supported in Flux (#996)
  • GPU Compile error on PPO with MaskedPPOTrajectory (#1007)
  • RL Core tests fail sporadically (#1010)
  • RL Env tests fail with latest OpenSpiel patches (#1011)
  • Tutorial OpenSpiel KuhnOpenNSFP fails (#1024)
  • CI: Should spell check be dropped or fixed? (#1026)

Contributors

joelreymont, threads, and 45 other contributors
Assets 2
Loading