v0.3.1

Major features and improvements

• Human in the Loop (HITL) Application: client/server architecture for real human interaction with simulated robots and environments

  • Reorganize the HITL framework as a new habitat-hitl package by @eundersander in #1754
  • Remote VR HITL Evaluation by @0mdc in #1731
  • Clean-up and fixes by @eundersander and @0mdc in #1738, #1766, #1771, #1777, #1788, #1789, #1795, #1802, #1849, #1848, #1800, #1837, #1852
  • Migrate HITL apps away from command-line args and to Hydra config yaml files by @eundersander in #1745
  • Support for headless HITL apps by @eundersander and @0mdc in #1768, #1794, #1792
  • server-side helpers to measure client latency by @eundersander in #1796
  • rearrange_v2 example app and other features/fixes for Feb 12 demo by @eundersander in #1810, #1830
  • Add camera sync support. by @0mdc in #1811
  • Make skinning support optional. by @0mdc in #1791
  • GUI-controlled Spot and local multi-user support by @eundersander in #1831
  • Remote keyboard input redirection. by @0mdc in #1853
  • Add GUI rendering wrapper. by @0mdc in #1838

• New Simulator utility functions: quality of life wrappers and prepositional logic functions

  • Object "getter" wrapper utils to sim_utilities by @aclegg3 in #1816
  • "Above" set query util by @aclegg3 in #1817
  • "Within" set query util by @aclegg3 in #1819
  • "Ontop" utility by @aclegg3 in #1835
  • "on_floor" query util by @aclegg3 in #1841
  • Region containment query utils by @aclegg3 in #1820

• Improved DebugVisualizer Util: get images of the world from within breakpoint() and other debugging scenarios

  • DebugVisualizer Update by @aclegg3 in #1775
  • DebugVisualizer generic 'peek' function by @aclegg3 in #1834

• Spot robot support changes for sim2real deployment:

  • Spot Mobile Gaze Training Code by @jimmytyyang in #1647, #1691
  • Spot social nav by @jimmytyyang in #1658
  • add jaw camera by @jimmytyyang in #1856

• Documentation updates:

  • Expose all habitat-lab API docs by @henrysamer in #1779, #1736
  • Add Humanoid + ArticulatedAgent tutorials by @xavierpuigf and @0mdc in #1765
  • Social nav README by @jimmytyyang in #1699, #1860
  • Humanoid details (convert SMPL-X to humanoid motions) by @xavierpuigf in #1703, #1720
  • Troubleshooting Common Problems readme by @0mdc in #1734
  • Improving PDDL Docs by @ASzot in #1784
  • Deprecate colabs in favor of Jupyter notebooks by @aclegg3 in #1742
  • Update README.md by @dhruvbatra in #1690, #1693

Other changes and bugfixes

• Update indices humanoid by @xavierpuigf in #1657

• Fix Spot-floating bug by @rpartsey in #1661

• query attr filepath if available to find AO rec meshes by @aclegg3 in #1669

• [BE] Fix the env pause issue by @jimmytyyang in #1744

• [BE] Make key as string by @jimmytyyang in #1764

• Add habitat_hitl.disable_agents_and_stepping config param by @eundersander in #1769

• Fixed config habitat_hitl.episodes_filter='4' and episodes_filter episode order by @henrysamer in #1772

• (1) Fixed error when using config habitat_hitl.episodes_filter='4' (2) Fixed episodes_filter episode order not respected (3) added check for data folder in current working directory by @henrysamer in #1774

• Adapt to Magnum Python bindings update by @mosra in #1776

• correction2 of type-cast when self._hitl_config.episodes_filter==None by @henrysamer in #1780

• Remove RTLD_GLOBAL flag from apps. by @0mdc in #1783

• Use receptacle unique name for dictionary key by @aclegg3 in #1782

• -- Change ints to float in sensor spec position and orientation by @jturner65 in #1793

• Correct object initialization by @xavierpuigf in #1806

• [bugfix] - minor generator fixes by @aclegg3 in #1814

• --Test fixes to support semantic sensor updates by @jturner65 in #1801

• --Address regression and missing stage_id checks by @jturner65 in #1822

• [3.1] Fix missing org name changes. by @0mdc in #1850

• [3.1] Fix assorted typos in user-facing files. by @0mdc in #1851

• Fix agent GPS compass key in multi-agent README by @gchhablani in #1844

New Contributors

• @henrysamer made their first contribution in #1736
• @gchhablani made their first contribution in #1844

Full Changelog: v0.3.0...v0.3.1

v0.3.0

Habitat 3.0: A Co-Habitat for Humans, Avatars and Robots (https://aihabitat.org/habitat3/)

• Multi-agent training
• Humanoid agents
• HSSD dataset
• Human-in-the-loop evaluation app for rearrangement tasks

What's Changed

• save velocity of objects in simstate by @xavierpuigf in #1499
• Core Trainer Refactors by @ASzot in #1482
• Fix Articulated Object Drift by @ASzot in #1523
• [hab3_merge] remove agent index by @xavierpuigf in #1521
• [hab3_merge] Add Skills for the robot to move forward/backward and turn by @xavierpuigf in #1508
• [hab3_merge] Define different policies for different agents by @xavierpuigf in #1530
• Revert "[hab3_merge] Define different policies for different agents" by @xavierpuigf in #1535
• [hab3_merge] Main different policies per agent by @xavierpuigf in #1536
• [hab3_merge] RearrangeEpisodeGenerator changes by @aclegg3 in #1517
• [hab3_merge] - RearrangeDataset pickle by @aclegg3 in #1545
• Docs change: Correct the URL for objectnav_mp3d_v1.zip by @xwy-bit in #1559
• Rearrange Task Config Cleanup by @ASzot in #1543
• [bugfix]: handle the case where a scene has no indoor navmesh islands by @aclegg3 in #1567
• [bugfix]: remove joint motors when converting objects to MotionType.KINEMATIC by @aclegg3 in #1566
• [hab3_merge] Updates in humanoid controller by @xavierpuigf in #1582
• [hab3_merge] - navmesh related refactor from SIRo to existing files by @aclegg3 in #1562
• [CI] lower hab2 interact bench thresholds after bug fix by @aclegg3 in #1590
• Flag to enable hbao in experiments by @xavierpuigf in #1598
• [CI] - update resource class by @aclegg3 in #1600
• Unoccluded navmesh snap by @aclegg3 in #1591
• [hab3_merge] Multi-Agent Trainers by @ASzot in #1524
• Fix gym.obs key name by @rpartsey in #1616
• Replace deprecated torch.norm by @Kuzmarg in #1611
• SIRo HITL Tool by @rpartsey in #1615
• Humanoid pick action by @xavierpuigf in #1619
• [Hab3] social nav in main by @jimmytyyang in #1605
• [Hab3] Planner Agent by @ASzot in #1610
• update version tags to v0.3.0 by @aclegg3 in #1613
• Fix Social Nav/Rearrange Task Integration by @ASzot in #1629
• [bugfix] - pass correct object to function by @aclegg3 in #1597
• Readme for hab3 by @xavierpuigf in #1633
• Update README.md by @xavierpuigf in #1636
• updates readme by @xavierpuigf in #1640
• Updated HITL tool README.md by @rpartsey in #1641
• Update EpisodeHelper to account for endless episode iterators by @rpartsey in #1642
• CI for Social Rearrange/Nav by @ASzot in #1634
• update readme with note by @xavierpuigf in #1643
• Agent configs refactor by @rpartsey in #1635
• Hab3 benchmarking scripts by @aclegg3 in #1639
• check for MotionType.DYNAMIC before trying to set joint motors by @aclegg3 in #1648

New Contributors

• @xwy-bit made their first contribution in #1559
• @Kuzmarg made their first contribution in #1611

Full Changelog: v0.2.5...v0.3.0

v0.2.5

Major features and improvements:

• RearrangeGenerator improvements:
  • cleanup and unstable object culling by @aclegg3 in #1248
  • Add support for glTF multi-receptacle assets by @aclegg3 in #1249
  • Receptacle unique_name filtering by @aclegg3 in #1298
  • Better stable rearrange sampling by @aclegg3 in #1301
  • [Bug Fix] change the snap point to be attached the largest island id by @jimmytyyang in #1371
• Removing Habitat v1 style actions and move to v2 style by @vincentpierre in #1251
• [breaking] - NavMeshSettings refactor by @aclegg3 in #1351
• Support for CLIP visual encoder by @mukulkhanna in #1339
• Huggingface data migration by @aclegg3 in #1450
• Add HSSD (Habitat Synthetic Scene Dataset) by @mukulkhanna in #1332
• Environment Factory+Habitat 2.0 Code Cleanup by @ASzot in #1401

Other improvements and bug fixes:

• [Development] Batch renderer integration by @0mdc in #1200
• [bugfix] - revert erroneous snap-down "fix" for stage mesh margin by @aclegg3 in #1240
• [Bug Fix] fix the angle between the agent and the object by @jimmytyyang in #1237
• Habitat 2.0 Tasks Cleanup by @ASzot in #1265
• General Hab2 Bug Fixes by @ASzot in #1331
• Fixing regression of performance (~7x slowdown) on CI by @vincentpierre in #1319
• [BUG FIX] Hydra migration reordered the obs_transforms erroneously by @vincentpierre in #1372
• [refactor] - update for Magnum breaking API change by @aclegg3 in #1356
• Apply breaking change in replay renderer API. by @0mdc in #1408
• add default navmesh construction config options to SimulatorConfig by @aclegg3 in #1411
• [Development] Batch renderer: Depth sensor support by @0mdc in #1416
• using config instead of len for counting envs by @naokiyokoyama in #1422
• [Development] Batch renderer PPO training integration by @0mdc in #1436
• [Docs change] Add bench renderer documentation. by @0mdc in #1443
• Remove typing_extensions. by @0mdc in #1446
• [CI] - pin doxygen version at 1.9.5 by @aclegg3 in #1452
• [CI] - disable hab2 multi-proc benchmark assertions by @aclegg3 in #1455
• [CI] - add --no-prune to automatic downloader calls by @aclegg3 in #1461
• misc perf optmizations by @Skylion007 in #1447
• Bug fix for loading weights when resuming training by @naokiyokoyama in #1472
• version bump v0.2.5 by @aclegg3 in #1475
• [Bug Fix] Fix bug in method to find action range by @sidpatki in #1476

New Contributors

• @sidpatki made their first contribution in #1476

Full Changelog: v0.2.4...v0.2.5

v0.2.4

Major features and improvements:

• Deprecation of OMPL and SLAM by @vincentpierre and @rpartsey in #1075, #1097 and #1193
• Added several omegaconf configurations to our public API and documented them by @vincentpierre in #1101
• Changes to high level policies training by @ASzot in #1053
• Added Humanoids to Habitat by @xavierpuigf in #1165
• Added Spot robot configuration by @jimmytyyang in #1177
• Changed our contribution guidelines and pull request templates by @vincentpierre in #1175
• Added Objectnav dataset for HM3DSem v0.2 by @srama2512 in #1185
• Added instanceImageNav HM3D dataset v0.3 by @jacobkrantz in #1186
• pip installable habitat-lab changes by @rpartsey in #1187, #1221, #1224, #1229
• New Agent abstraction in trainers by @ASzot in #1169
• Upgrade to python version 3.9 by @vincentpierre in #1198

Other improvements and bug fixes:

• Update a dataset URL in the Readme by @dhruvbatra in #1039
• Fix field of view issue by @jimmytyyang in #1048
• Update instructions of interactive testing by @akshararai in #1054
• Conda update on CI by @0mdc in #1052
• Fix version of toml in CI by @vincentpierre in #1063
• New visualization tutorial by @rpartsey in #1037
• Bugfix on topdown map drawing by @jimmytyyang in #1060
• Refactor of debug visualization by @vincentpierre in #1064
• Allow topdown map for multi-agent environments by @vincentpierre in #1072
• Fix speed regression in rearrangement tasks by @ASzot in #1085
• Small updates post hydra migration by @rpartsey in #1088
• Making read_write on omegaconf part of the public API by @vincentpierre in #1090
• Adding documentation on DatasetConfig omegaconf by @vincentpierre in #1091
• Better error handling when loading checkpoints by @vincentpierre in #1094
• Moving all gym files into a dedicated gym folder for better separation by @vincentpierre in #1095
• Fixing import bug for torch by @vincentpierre in #1106
• Bump isort version in CI by @0mdc in #1107
• Bump black version in CI and pre-commit by @aclegg3 in #1110 and #1113
• Removed unused line by @jimmytyyang in #1104
• Performance improvements (~20%) on habitat-lab by @vincentpierre in #1120
• Documenting HABITAT_ENV_DEBUG environment variable by @vincentpierre in #1127
• Refactoring code to unlock all of hydra with submitit functionalities by @vincentpierre in #1128
• Add flaky to some pytests to make them more robust by @aclegg3 in #1131
• Documenting some of the most useful configuration keys for navigation by @vincentpierre in #1111
• Added velocity control to the default configurations by @luisenp in #1119
• CI fix with pytest version by @vincentpierre in #1141
• Documenting some of the most useful configuration keys for rearrangement by @vincentpierre in #1126
• Fix tutorials configuration paths by @dhruvbatra in #1065
• Fix on depth image normalization by @vincentpierre in #1152
• Faster loading of habitat module by @vincentpierre in #1144 and #1157
• Test for speed performance of simulation on CI by @vincentpierre in #1114
• Gym API support enhancements by @rpartsey in #1145
• Renaming of agents to be more general by @xavierpuigf in #1150
• Bug fix on instanceImageNav distance viewpoint threshold by @Ram81 in #1130
• Optimized a list comprehension by @Skylion007 in #1168
• Documentation change by @rpartsey in #1167
• CI change replace assets if already present by @jimmytyyang in #1179
• Rename Robots to Articulated Agents by @xavierpuigf in #1176
• Moved from attr.s to dataclass by @rpartsey in #1182
• Bug fix on accumulated force metric by @jimmytyyang in #1194
• Fix Humanoid armature by @xavierpuigf in #1195
• Update torch dependency version by @rpartsey in #1197
• Fix loading of DD-PPO checkpoints by @rpartsey in #1116
• TriangleMeshReceptacles and RearrangeGenerator improvements by @aclegg3 in #1108
• Fix typing issue in receptacles.py by @0mdc in #1203
• PDDL system upgrade by @ASzot in #1190
• Fix of humanoid controller by @xavierpuigf in #1202
• New grasp action for Spot and Stretch robot by @jimmytyyang in #1208
• Bug fix in rearrangement generator by @aclegg3 in #1220
• Fixes for HRL trainer by @ASzot in #1209

New Contributors

• @luisenp made their first contribution in #1119

Full Changelog: v0.2.3...v0.2.4

v0.2.3

Major features and improvements:

• Major speed up of ObjectNav evaluation by @ykarmesh in #867
• VER trainer implementation by @erikwijmans in #905, #906 and #907
• New tasks for the habitat challenge by @ASzot in #883
• New InstanceImageNav task by @jacobkrantz in #936 and #940
• Splitting habitat-lab and habitat-baselines into separate packages by @vincentpierre in #922
• Configuration system was moved from YACS to Hydra by @rpartsey and @vincentpierre in #977, #983 and #1000
• Fix to the MoveObjectReward measure by @SergioArnaud in #992 and #1010
• Port Stretch and Spot Robots to habitat by @jimmytyyang in #982, #1014 and #1018

Other improvements and bug fixes:

• Robot scripts were moved from habitat-sim to habitat-lab by @zacavis in #888
• Add support for multiple continuous or discrete actions in the gym wrapper by @vincentpierre in #891
• Make habitat-baselines use the gym API for training by @vincentpierre in #892
• Make LSTM recurrent state contiguous to fix errors by @vincentpierre in #901
• Render the last frame of evaluation videos by @vincentpierre in #904 and #924
• Replace RearrangeRLEnv and NavRLEnv with a single RLTaskEnv class for better readability by @vincentpierre in #902
• Add Polymetis usage example by @laikhtewari in #897
• CI test fixes by @ASzot in #915
• CI test fixes by @vincentpierre in #913, #912 and #921
• Bug fix for the observation transformer by @vincentpierre in #903
• Creating Static Manipulator Interface by @laikhtewari in #909
• Documenting habitat test scenes and pointnav episodes by @srama2512 in #916
• Improvements to the README by @dhruvbatra in #927
• Allow DDPPO to train any gym environment with compatible observation and action spaces by @vincentpierre in #930
• Fix mypy CI issue by @vincentpierre in #932 and #1023
• Fix Franka robot CI by @laikhtewari in #935
• Refactor the way rgb and depth observations are processed by the ResnetEncoder by @vincentpierre in #933
• New Franka robot block stacking task by @laikhtewari in #925
• Bugfix in ResNetEncoder by @vincentpierre in #938
• Enable all tasks configs for the auto gym tests by @vincentpierre in #939
• Change link to pretrained VQA encoders by @mukulkhanna in #942
• Multi-agent integration by @ASzot in #941
• Add human render option to the gym interface by @vincentpierre in #944
• Refactor mobile_manipulator and static_manipulator by @jimmytyyang in #948
• Fix the first observation after reset being from the wrong position by @vincentpierre in #952
• Switch CirecleCI versions by @mathfac in #958
• Fix an issue in which the simulation would not be properly reconfigured by @vincentpierre in #955
• Fix typos in the README by @0mdc in #962
• Added a habitat namespace in the configuration by @vincentpierre in #969
• Fix indentation error in gym wrapper code by @vincentpierre in #972
• Update CircleCi for Spot robot by @jimmytyyang in #967
• VER updates by @erikwijmans in #979
• Fixing merge config issue with VER by @vincentpierre in #984
• Fix links to documentation in example colab by @xavierpuigf in #987
• Change pytest-sugar version on CI to avoid a dependency issue by @0mdc in #994
• Add Stretch robot urdf to CircleCI by @jimmytyyang in #995
• Moving flake8 pre-commit dependency from gitlab to github by @vincentpierre in #1002
• Fixing circular import dependency for objectnav by @vincentpierre in #998
• Re-adding --add-ik to example script by @vincentpierre in #1004
• Changing copyright in files by @NakuraMino in #986 and #988
• Fix documentation path in gym example by @vincentpierre in #1016
• Fix manifest issues by @0mdc in #991
• Moving habitat2 documentation page to habitat-lab by @rpartsey in #1025
• Adding instance image nav paper reference by @jacobkrantz in #1020
• Bumping version numbers by @aclegg3 in #1026
• Routing new users to the new discussion page by @aclegg3 in #1027
• Updating CI tests for datasets by @vincentpierre in #1019
• Update the Readme to remove the Slack channel link by @dhruvbatra in #1031
• Large bug fixes for habitat 2.0 tasks by @ASzot in #959
• Bug fix for navigation task by @mukulkhanna in #1032

New Contributors

• @zachavis made their first contribution in #888
• @laikhtewari made their first contribution in #897
• @jacobkrantz made their first contribution in #936
• @jimmytyyang made their first contribution in #948
• @0mdc made their first contribution in #962
• @xavierpuigf made their first contribution in #987
• @SergioArnaud made their first contribution in #992
• @NakuraMino made their first contribution in #986

Full Changelog: v0.2.2...v0.2.3

v0.2.2

Major features and improvements:

• Complete Habitat 2.0 tasks, episode generation, and benchmark migration (#837, #862, #764)
• Upgrade minimum python version to 3.7 by @aclegg3 in #801
• [ObsTransforms] Add support for semantic sensor observation transforms by @Ram81 in #847
• Move gym from habitat_baselines to habitat by @vincentpierre in #864
• HM3D Semantics update by @srama2512 in #822

Other improvements and bug fixes:

• PointNav datasets for Gibson 0+ and HM3D by @srama2512 in #711
• Fix failing gym 2.0 test by @erikwijmans in #725
• [TechDebtWeek] made sucess -> success by @ykarmesh in #724
• Remove unused agent dynamic properties by @aclegg3 in #720
• --Update field name to match Habitat-Sim refactor. by @jturner65 in #731
• Fixed Dockerfile conda installation by @NickLucche in #730
• Fixed flake suggestions and combined logical clauses. by @mathfac in #732
• Remove tensorflow dependency by @erikwijmans in #740
• Update README.md by @dhruvbatra in #741
• Fix set episode(s) by @erikwijmans in #739
• Fix batch_obs for CPU training. by @erikwijmans in #744
• Update references to habitat-sim master => main by @eundersander in #733
• Add pre-commit hook to ban blanket noqa by @Skylion007 in #746
• Added pointer to ROS-X-Habitat by @ericchen321 in #743
• Renamed references from master to main in docs and for DDPPO master address by @mathfac in #748
• Update README.md by @dhruvbatra in #762
• Citation update by @ASzot in #750
• Update pre-commit and apply change for new black version by @aclegg3 in #767
• Updated README to Clarify MacOS Installation by @ASzot in #770
• Fix set goal bug by @erikwijmans in #772
• Bugfix negative force penalty by @vincentpierre in #783
• Deprecating TASK.SUCCESS_DISTANCE since TASK.SUCESS.SUCCESS_DISTANCE replaces it by @vincentpierre in #784
• Made tensorboard metrics and losses each appear in their own graph by @vincentpierre in #790
• Mypy precommit by @vincentpierre in #792
• [CI] Showing the mypy version on circleci tests by @vincentpierre in #798
• Adapt to m.css updates in the sim repository by @mosra in #806
• Upgrade CI gpu resource to gpu.nvidia.medium by @aclegg3 in #807
• docs: update README by @Aakash-Roy in #789
• Fixing typing annotation on gym.Env by @vincentpierre in #819
• setting gym min version to 0.22.0 by @vincentpierre in #820
• [bugfix] - Apply AO instance uniform scale to Receptacles by @aclegg3 in #815
• Adds missing PointNav HM3D config file by @srama2512 in #823
• [Fix] Fixing mypy by fixing mypy version by @vincentpierre in #836
• Require config keys have default by @vincentpierre in #776
• removing unused configs by @vincentpierre in #840
• updating click version on main by @vincentpierre in #858
• (bugfix): Fix mypy typing error + precommit issue by @Skylion007 in #865
• Tensorboard Version Fix by @ASzot in #874
• Use find free port by @erikwijmans in #877
• Fix backwards compatability for Hab1 tasks and Hab2 gym-style tasks by @erikwijmans in #878
• Fix a minor bug about generate_video by @Jiayuan-Gu in #879
• Update all references to YCB configs directory for v1.2 by @aclegg3 in #884
• [ci-bugfix] - remove opencv conda by @aclegg3 in #886

New Contributors

• @ykarmesh made their first contribution in #724
• @NickLucche made their first contribution in #730
• @ericchen321 made their first contribution in #743
• @Aakash-Roy made their first contribution in #789
• @Ram81 made their first contribution in #847

Full Changelog: v0.2.1...v0.2.2

v0.2.2 rc1

v0.2.2 rc1

v0.2.1

Major features and improvements:

• Addition of velocity_control actions and support for Gaussian action (#655)

House Assistant Benchmarks Task Changes:

• Collision force measurements fixes. (#684)
• Contact measuring API changes with properly return contacts differentiating between object-on-scene, robot-on-scene, and robot-on-object types of contacts.
• Starting state and the robot arm proper location fix.
• Snap to grasp manager implemented and the grasp manager takes care of the collision groups.
• Refactored observations_to_image to allow variable-sized sensor images. The different image sizes are now tiled side by side. Image and text flip fix. (#687)
• Added actions for setting absolute kinematic state and target joint states. (#685)
• Added separate IK config for the interactive play script.
  Refactored measurements code to break down the reward signal and reuse measurements for the reward and success signal.

Habitat Baselines Changes:

• Support rendering images from observation with unequal dimensions.
• Added benchmark_render.py, a new version of benchmark.py which supports rendering videos and saving trajectories to a file.
• Added Gym wrapper with config for which parts of the observation to get.
• Added Sense Plan Act pipeline from Habitat 2.0 paper.

Other improvements and bug fixes:

• Add detail of headless vs. not to the readme (#679)
• Fix to use resume config when training (#695)
• Fix bug with size of the joint observation space.
• Fixed the base transform was not being properly calculated.
• Added interactive play test utility script. This is an important way to test the functionality. (#684)
• Added collision specific measurements.
• Refactored the rearrange simulator class to have less responsibility.
• Record trajectory from the interactive play script for automated testing.
• Added episode generation script for rearrangement tasks under examples folder.
• Fixed properly load in scene config for new ReplicaCAD format.

Corresponding Habitat-Sim v0.2.1 release.

Kudos to our community contributors: @naokiyokoyama.

v0.1.7

Major features and improvements:

• Added fisheye camera to observation transformations by @matsuren (#486, #496)
  Implementation equirectangular into perspective images of observation transformation and performance improvement on cube map to equirectangular transformation by @matsuren (#517, #485)
• Collapsed PPO and DD-PPO trainers, added faster RNN code, and double-buffered sampling, switched to optimal default models like ResNet18 (#538, #557)
• Added performance profiling capabilities for PPO/DDPPO training (#477, #476)
• Added Visual Question Answering baseline and NAV-PACMAN baseline implementation for EQA by @mukulkhanna (#487, #539)
• Added mypy and static mandatory type checking and significantly improved typing of the existing codebase (#492, #527, #535, #535)
• +5% speedup for all model training with zero-copy pickling method used for the vector environments (#582)

Other improvements and bug fixes:

• Support of modern gym API (#498)
• Enhance the display of the top-down map for a floor the agent is currently on by @Skylark0924 (#580)
• Support properties and members with a call method for vector environment (#548)
• [Backwards Incompatible Change] Address out of sync with Habitat Sim config file extensions and refactor for removal of Habitat SimSceneConfiguration (#508, #516, #525)
• Episode generator bug fix by @Jbwasse2 (#560)
• SLAM baselines install script update by @Jbwasse2 (#552)
• Allow PointNav GPS and Compass to support 3D (#553)
• Fix DDP Reduction in PyTorch 1.7 when the model has unused parameters #586
• Compatibility with refactored CameraSensor and SceneGraph in Habitat Sim (#575)
• Bugfix for the situation when the child pipe was not closed (#585)
• Updates of code style guide enforcement tooling (#512)
• Continue integrations improvements: using pre-installed CUDA and better pytest results handling (#510, #511, #524)
• Added support of running evaluation on multiple single checkpoints in the same folder (#493)
• Added shellcheck to pre-commit hook (#509)

Corresponding Habitat-Sim v0.1.7 release.

Kudos to our community contributors: @matsuren, @mukulkhanna, @Jbwasse2, @Skylark0924

v0.1.7 rc1

v0.1.7 rc1