Releases: opendilab/DI-drive
Releases · opendilab/DI-drive
DI-drive (beta) v0.3.4
DI-drive (beta) v0.3.3
- Update readme, reorganize info
- Add MetaDriveTrajEnv and doc
- Add utils for MetaDrive
- Modify utils for macro env
DI-drive (beta) v0.3.2
- Update banner logo
- Update to DI-engine 0.3, modify env properties
- Fix Basic policy bug
- Fix CILRS model bug
- Fix bugs in evaluator
- Fix bug in ppo config
- Fix bug in simple rl pipeline
- Add pre-train weights path in foc
DI-drive (beta) v0.3.0
- Add MetaDrive support (env, entry, doc)
- Add simulator selection when import
- Rename all common modules without 'carla'
- Add MetaDrive basic env training
- Add MetaDrive Macro env training and test
- Add all env registry
- Update DriveEnvWrapper
- Update logo
- Update doc, add metadrive guide and tutorial
- Reformat simple rl, all policies now have common entry(train eval test)
- Fix LBC planner bug
- Delete config helper
DI-drive (beta) v0.2.2
- Add fail count and retry for benchmark collector
- Add LBC Image training phase 0 & 1
- Add comments for commonly used default config
- Add LBC Image train doc; scenario doc; save replay doc
- Update ci and setup
- Update simple rl env wrappers
- Update args in all Carla policy
- Modify all eps-greedy input into learn iterations
- Fix bug in CILRS model
- Fix typos in doc and comments
DI-drive (beta) v0.2.1
- Add NoCrash Carla099
- Add bev dataset
- Add Carla AD Challenge scenarios and routes
- Add LBC data collection and BeV training doc
- Add Latent RL BeV and RL training doc
- Add validation in LBC training
- Update Benchmark datasaver
- Update VAE model
- Update LBC Image policy learn mode
- Fix bug in benchmark collector
- Fix visualizer not closed after done
- Fix not alived sensor killed bug
DI-drive (beta) v0.2.0
- Update DI-engine to version 0.2.0
- Reformat CIL to CILRS
- Add CILRS dataset, model, policy and entry
- Add LBC data collection and dataset
- Add LBC BeV training entry
- Add simple RL demo doc
- Add logger and should eval for all evaluators
- Add enable save replay in DI-engine demos
- Update IL tutorial and CILRS demo doc
- Update Steer Noise wrapper
- Update Benchmark datasaver and collector
- Fix import bug in sensor helper
- Fix visualization bug
DI-drive (beta) v0.1.3
- Update DI-engine to version 0.1.1
- Add serial evaluator for evaluation in training. It has tensorboard
- Update single and benchmark evaluator
- Update simple_rl entries. Adapt ding 0.1.1
- Modify models in simple_rl. Use modules in ding
- Modify env wrapper in simple_rl. Add inputs for discrete wrapper
- Update config system in RL. Move all without server into env and policy
- Update comple_config to adapt new system
- Add CICT eval
- Fix bug in BasicPlanner
- Fix bug in implicit train_sl
- Fix figure in doc
- Add auto doc in github pages
DI-drive (beta) v0.1.2
- Rename 'orientation' to 'forward_vector'
- Redefine train, test, and eval in demo
- Move 'utils' in simple rl into core
- Add Latent RL demo, including env, DQN train, test and eval
- Add MPC controller with waypoints input
- Add AutoMPCPolicy, rename AutoPolicy to AutoPIDPolicy
- Add Vanilla VAE model
- Add API doc for MPC controller
- Add API doc for LPC Policy
- Update evaluator. It now returns success / success rate
- Fix a bug that walker controller may be identified as walker in CarlaDataProvider
- Fix birdview visualizer to show less channels
DI-drive (beta) v0.1.1
- Update AutoPolicy: better performance in multi-lane maps
- Add bike avoidance and agent state in behavior planner
- Add CICT IL method: data, train, eval and test
- Add CICT doc
- Add LBC: eval and test
- Add LBC: doc
- Add common PID controller
- Add LBC planner
- Add transform obs for evaluator
- Add velocity, angular velocity and rotation in simulator's navigation and data provider
- Modify initialization of AutoPolicy
- Fix bugs in default config of all modules
- Add overview for Implicit Affordance doc