Skip to content

Wbk/1026#526

Open
wangbingke0 wants to merge 346 commits intoEmerge-Lab:mainfrom
wangbingke0:wbk/1026
Open

Wbk/1026#526
wangbingke0 wants to merge 346 commits intoEmerge-Lab:mainfrom
wangbingke0:wbk/1026

Conversation

@wangbingke0
Copy link

No description provided.

SamanKazemkhani and others added 30 commits April 23, 2024 14:43
* Working tests

* Github workflows

* Cmake

* Make it compatible with valid_files
…es (Emerge-Lab#89)

* Add collision behavior.

* Add video method to callback.

* Restructure

* Make SB3 wrapper compatible with rendering options.

* Environment updates.

* Environment updates.

* Stable MAPPO with all observation types.

* WIP: Stable MAPPO when max_cont_agents are the first N agents

* Updated base_env and wrapper to support worlds with a different number of agents.
* Ignore Pedestrians param

* Use better naming
* Rotate vectors for egocentric POV

* cleaner version
* Export the entity type of the agent in Info Tensor

* Export Entity Type bindings
* Update mgr.cpp

* Fix viewer

* Fix padding agent values on reset

* Render padding agents too

* Export valid state

* Fix mean calculations (Emerge-Lab#84)

* Local rendering (Emerge-Lab#79)

* Add scale to road observations (Emerge-Lab#83)

* Fix scale road issue

* save progress

* Looking good

* Add json for viewer

* Remove M_PI sub

* Reintroduce means

* Remove floor plane

---------

Co-authored-by: SamanKazemkhani <skazemkhani@gmail.com>
* Export vehicle size

* Working pygame renderer

* Working renderer

* Draw agents on top and map only once
Add missing files to requirements, add paths to install
* init

* Make vel zero on agent stop

* Make sure info is updated

* Dont allow negative z vel
* PyGame Renderer (Emerge-Lab#104)

* Export vehicle size

* Working pygame renderer

* Working renderer

* Draw agents on top and map only once

* Cleanup and integrate info_tensor.

* Improve logging and environment.

* Basic FFN

* Update default params.

* Speed benchmark

* Improve naming and minor changes.

* Basic policy / expert eval script

* Minor

* Exper eval script.

* Fix

* Clean up gym env

* speed nb

* wip

* Important bug fix: auto_reset must be False for RL training

* Minor

* Set default settings to full obs

* Name changes.

* separate out storage and computation devices in rollout buffer (Emerge-Lab#110)

* Remove cache and binding changes.

---------

Co-authored-by: Aarav Pandya <ap7641@nyu.edu>
Co-authored-by: Eugene Vinitsky <eugenevinitsky@users.noreply.github.com>
* Check controlled state safely

* Revert headless

* Use lambda function and inclusive or

* Type def the lambda

* Invert the booleans
* Init static agents

* Fix the compile error

* Formatting
daphne-cornelisse and others added 22 commits April 3, 2025 16:30
* Optional vbd imports

* use_vdb = False by default

* Add optional dependencies pytorch-lightning and jaxlib
Currently it is possible to unknowngly use two different seeds for the
PPO and dataloader. Normally if everyone was using 42 for the seed there
would be no effect, but we tested with varying seeds and had unintuitive
results, so I think matching the seeds here makes sense.

Co-authored-by: Ellington Kirby <ellingtonkirby@gmail.com>
* [FEATURE] cherry pick the json parsing update

* [TEST] adding a simple json parsing test

---------

Co-authored-by: Ellington Kirby <ellingtonkirby@gmail.com>
Co-authored-by: Ellington Kirby <ellingtonkirby@gmail.com>
* Update madrona to commit 4bda334 (from fork), which is uv compatible

* Use uv compatible build backend of madrona

* Switch to uv

* Update readme, include pyenv again
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Comments