Releases: yardenas/safe-learning
Releases · yardenas/safe-learning
release/0.1.25
What's Changed
- Add fingers to Franka by @yardenas in #211
- Rccar hardware baseline by @yardenas in #212
- Use red box by @yardenas in #213
- add lbsgd penalizer and example config for cartpole swingup by @yardenas in #216
- Add implementation of pure exploration by @yardenas in #217
- Second attempt on offline rccar + hardware by @yardenas in #214
- minor changes to log barriers implementation and finish tuning by @acpoll in #218
- Walker offline by @ManuelWendl in #219
Full Changelog: release/0.1.24...release/0.1.25
release/0.1.24
What's Changed
- Implementation of sooper with vision by @yardenas in #205
- Add breaking controller by @yardenas in #207
- Lagrangian baselines by @acpoll in #210
- Sooper go-to-goal v2 by @ManuelWendl in #209
Final parameters for SOOPER paper
Full Changelog: release/0.1.23...release/0.1.24
release/0.1.23
What's Changed
- MBPO + Vision by @yardenas in #200
- SOOPER + Vision by @yardenas in #202
- Fix bugs in sooper by @yardenas in #204
- Updates to Franka Online by @yardenas in #203
- Variable goal size for go-to-goal environment by @acpoll in #206
Online learning on franka works
Full Changelog: release/0.1.22...release/0.1.23
release/0.1.22
What's Changed
- Load auxiliaries by @yardenas in #189
- Add PointGoal1 by @yardenas in #190
- Sooper humanoid by @ManuelWendl in #191
- Take other fix by @yardenas in #192
- Handle safe penalizer without sooper by @yardenas in #193
- Take other fix by @yardenas in #194
- Fix normalizer in offline by @yardenas in #195
- Sooper gtg easy offline by @ManuelWendl in #196
- Add offline data collection by @yardenas in #197
- Logging updates by @yardenas in #198
- SPiDR from Vision by @yardenas in #177
- Fix MBPO vision wrapper by @yardenas in #199
- Sooper go to goal v1 improved version by @ManuelWendl in #201
Offline on PointGoal1
Full Changelog: release/0.1.21...release/0.1.22
release/0.1.21
What's Changed
- Observation wrapper safety gym by @ManuelWendl in #185
- Observation wrapper walker by @ManuelWendl in #187
- Preprocess RBG images by @yardenas in #172
- Sooper walker walk by @ManuelWendl in #188
Walker baseline and franka arm sim-to-real
Full Changelog: release/0.1.20...release/0.1.21
release/0.1.20
What's Changed
- Sooper cartpoleswingup by @ManuelWendl in #169
- Franka hardware debug image by @yardenas in #171
- Get offline training working by @ManuelWendl in #173
- Collect offline data by @yardenas in #174
- Offline config by @ManuelWendl in #175
- Sooper rccar by @ManuelWendl in #176
- MBPO no pessimistic sample by @yardenas in #178
- if case dict by @yardenas in #179
- Minor utility things by @acpoll in #180
- Eval policy by @acpoll in #182
- Fix wrapping of eval env in safety gym by @ManuelWendl in #183
- Rccar Hardware Offline Training by @ManuelWendl in #184
Offline to online works
Full Changelog: release/0.1.19...release/0.1.20
release/0.1.19
What's Changed
Full Changelog: release/0.1.18...release/0.1.19
release/0.1.18
What's Changed
- Configure sooper filtering as parameter by @yardenas in #133
- Aloha by @yardenas in #128
- rccar environment with integrated FrameActionStack and ActionObservationDelayWrapper by @ManuelWendl in #134
- Sooper benchmarks by @yardenas in #136
- Initial commit by @yardenas in #135
- Sooper rccar real by @yardenas in #138
- Fix bug in sac training replay buffer by @yardenas in #140
- Add image observations by @yardenas in #137
- Fix init alpha by @yardenas in #142
- Sac vision improvement by @yardenas in #141
- Add scheduler to RAE by @yardenas in #143
- Minor changes to make Euler guide easier to follow by @acpoll in #144
- Wagener implementation by @ManuelWendl in #146
- Add franka by @yardenas in #147
- Update steps on euler by @yardenas in #148
- Use pytree replay buffer in rae by @yardenas in #149
- Don't collect data if offline flag is on by @yardenas in #150
- Add panda policy loading by @yardenas in #151
- Add vision to mbpo by @yardenas in #152
- Actor does the same grad steps as critic, just updating the params less by @yardenas in #154
- Fix bug in checkpoint loading by @yardenas in #155
- Save and load alpha by @yardenas in #156
- Add sim-to-sim experiments for franka by @yardenas in #153
- Fix FPS bug by @yardenas in #160
- Fix target initialization by @ManuelWendl in #161
- Add matplotlib by @yardenas in #162
- Add ball by @yardenas in #159
- On policy sac buffer by @ManuelWendl in #166
- Sooper rccar hardware by @yardenas in #168
Stable version on RaceCar hardware (one seed)
New Contributors
Full Changelog: release/0.1.17...release/0.1.18
release/0.1.17
What's Changed
- Update and clean deps to support madrona by @yardenas in #129
- Safe mbpo by @yardenas in #121
- Generalize saute by @yardenas in #130
- Sooper by @ManuelWendl in #132
- Add penalizer to mbpo by @yardenas in #131
Implementation of SOOPER works on cartpole swingup
Full Changelog: release/0.1.16...release/0.1.17
release/0.1.16
What's Changed
- Dump buffer by @yardenas in #104
- Bump tornado from 6.4.2 to 6.5.1 by @dependabot in #106
- Rea by @yardenas in #105
- Implementation of Model-based PPO by @ManuelWendl in #107
- Make RAE experiment easier by @yardenas in #108
- Minor updates by @yardenas in #109
- Split by @yardenas in #110
- Make mb ppo work by @yardenas in #111
- Implementation of MBPO by @yardenas in #112
- Mbpo no replay by @yardenas in #113
- Use done for model learning by @yardenas in #116
- Add hard autoreset wrapper by @yardenas in #115
- Add safety and terminations to the planning MDP by @yardenas in #114
- Updates from the sample efficiency branch by @yardenas in #117
- Cost-Critic Implementation by @ManuelWendl in #118
- Restore to the best working version by @yardenas in #119
- Fix bugs in how transitions are parserd by @yardenas in #120
- Take updates from get-back-good-perf by @yardenas in #122
- Make sure that main is OK by @yardenas in #123
- Parameterize the schedules by @yardenas in #124
- Preparing for RAE analysis by @yardenas in #125
- Add go2 by @yardenas in #126
- Smarter rae by @yardenas in #127
New Contributors
- @ManuelWendl made their first contribution in #107
Full Changelog: release/0.1.15...release/0.1.16