v0.0.2

What's Changed

• fix: clip mpo actions used in q function to avoid extrapolation by @EdanToledo in #55
• chore: remove self-implemented code in favour of jumanji wrapper by @EdanToledo in #56
• fix: use of truncation in GAE calc by @EdanToledo in #57
• fix: add option to use GAE as value targets by @EdanToledo in #58
• feat: add running statistics utils modified from acme by @EdanToledo in #60
• feat: add beta distribution policy head by @EdanToledo in #63
• Chore/refactor loss metrics by @EdanToledo in #61
• Feat/add ppo penalty by @EdanToledo in #64
• chore: slight change to configs by @EdanToledo in #65
• chore: Make Update Batch Size not affect num envs, buffer size and batch size by @EdanToledo in #68
• fix: double critic being initialised to same network by @EdanToledo in #73
• Chore/refactor type by @EdanToledo in #74
• Feat/add vmpo by @EdanToledo in #75
• fix: recurrent ppo by @EdanToledo in #76
• Chore/change mpo loss by @EdanToledo in #80
• feat: add notebook to plot stoix algorithms by @EdanToledo in #87
• chore: edit readme by @EdanToledo in #88
• feat: add a weights and biases logger by @EdanToledo in #89
• fix: add nstep transitions to d4pg by @EdanToledo in #92
• Feat/rainbow by @RPegoud in #86
• Chore/change muzero networks by @EdanToledo in #93
• chore: move input of distributional network args into config by @EdanToledo in #94
• chore: edit wrappers to have a separate flatten obs wrapper by @EdanToledo in #95
• feat: generalise win rate to be solve rate by @EdanToledo in #96
• Feat/add popjym by @EdanToledo in #97
• fix: typing issues causing double compilation by @EdanToledo in #100
• Feat/add navix by @EdanToledo in #101
• Feat/Add Sebulba by @EdanToledo in #105

New Contributors

• @RPegoud made their first contribution in #86

Full Changelog: v0.0.1...v0.0.2