Skip to content

Releases: NVlabs/COMPASS

v1.6.0

18 May 19:22

Choose a tag to compare

Added

  • OSMO cloud submission: osmo/run_osmo.py + four workflow YAMLs (train / eval / record / distillation). Assets (X-Mobility base checkpoint, COMPASS USDs) download from HuggingFace inside the workflow; W&B / HF tokens routed via env vars or interactive --prompt. Includes --embodiment / --environment overrides for multi-embodiment training sweeps without YAML hand-edits.
  • Multi-GPU residual-RL training with --distributed: torchrun-launched, per-rank Isaac Sim instance, manual gradient / KL / metric all-reduce in PPO.update, rank-0-only logger / checkpoint / video / episode-log writes. Pairs with osmo/run_osmo.py --num-gpus N and the parameterized osmo/workflows/rl_es_train_workflow.yaml (handles 1-N GPUs via a single {{num_gpus}} template variable).
  • Auto-generated occupancy maps from USD: scripts/generate_omap_from_usd.py wraps Isaac Sim's isaacsim.asset.gen.omap. OccupancyMapCollisionChecker falls back to discovering <usd_dir>/omap/occupancy_map.yaml when OMAP_PATHS has no entry, so SAGE-driven training and new scenes ramp up without hand-tuned omaps.
  • Docker-as-venv dev environment: docker/run.sh (build / assets / up / down / exec / shell / status subcommands) plus a sourceable docker/activate that transparently routes python / pip / tensorboard through the container. Cuts first-time setup from ~30-60 minutes to ~3 minutes.
  • COMPASS docs site: Sphinx handbook (docs/handbook/) deployed at nvlabs.github.io/COMPASS/docs/; academic project page consolidated under docs/project_page/. Auto-deployed via .github/workflows/docs.yml.
  • Agentic skills for Claude Code under .claude/skills/: /compass (umbrella front-door for training / eval / SAGE / OSMO / deployment) plus specialty siblings /compass-doctor (8-check diagnostic with scripts/compass_status.sh) and /compass-newembodiment (interactive robot-onboarding flow).
  • CI: .github/workflows/pre-commit.yml runs yapf / pylint / nbstripout / clang-format / large-files / trailing-whitespace / EOF / requirements-txt-fixer on every PR.
  • No-regression benchmark gate: osmo/run_benchmark.py fires one OSMO eval workflow per scene for a given embodiment (default sweep across 5 scenes); results land in W&B for offline regression assessment.

Changed

  • Isaac Lab 2.1 → 3.0.0-beta1. Breaking for downstream consumers of mobility_es APIs: AdditiveUniformNoiseCfgUniformNoiseCfg; restructured physics via isaaclab_physx.physics.PhysxCfg; rerender_on_resetnum_rerenders_on_reset; ActionTerm import moves from isaaclab.envs.mdp.actions to isaaclab.managers; asset data fields (root_pos_w, root_quat_w, etc.) require wp.to_torch() wrapping; write_root_state_to_sim() split into write_root_pose_to_sim_index() + write_root_velocity_to_sim_index(); quaternion convention flipped wxyz → xyzw across environments.py and NonHolonomicPerfectControlAction yaw extraction; velocity_limitvelocity_limit_sim on carter caster joints; commands.UniformPose2dCommand used directly (module reorg).
  • --wandb-project-name is now required in run.py, record.py, and distillation_train.py (previously defaulted to internal project names).
  • requirements.txt: 17 previously-unpinned dependencies pinned to versions verified inside compass-rl:latest.
  • README.md leads with the Docker quick-start; bare-metal install moved under "Manual install".
  • Pre-commit yapf bumped to v0.43.0 for Python 3.12 compatibility (the previous v0.31.0 pin imports lib2to3, removed from the stdlib in 3.13).

v1.5.2

16 Oct 17:25

Choose a tag to compare

[1.5.1] - 2025-10-16

Added

  • Gr00t client service to run benchmark on Gr00t post-trained policy

Changed

  • Updated the README for Gr00t fine tuning instruction

v1.5.1

13 Oct 22:10

Choose a tag to compare

[1.5.1] - 2025-10-13

Added

  • Script to convert distillation dataset from HDF5 to Gr00t Lerobot format

Changed

  • Updated the README for Gr00t fine tuning instruction
  • Updated the data recording function to include more fields

v1.5.0

29 Jul 23:20

Choose a tag to compare

[1.5.0] - 2025-07-29

Added

  • Enabled goal heading alignment for object navigation integration, improving navigation accuracy to targets.
  • Added ROS2 deployment examples for COMPASS Navigator, including containerized workflows and Isaac Sim integration.
  • Added target_pose_generator node to interface with object localization modules to enable object navigation using COMPASS

Changed

  • Upgraded Isaac Lab to version 2.1 for improved simulation capabilities and compatibility with the latest Isaac Sim and ROS2 releases.

Fixed

  • Disable weights only for generalist policy loading