[TATP-233]

Summary

• Version bump 1.5.2 → 1.5.3 in both package profiles and both root lock entries.
• Add the bilingual 1.5.3 changelog section for:
  • opt-in state-feedback physics substeps and mapped body-state row binding;
  • the self-contained env contract/package split and unirl extra;
  • unified DR capability consumption with UniSim 2.2.3;
  • reset seeding/API corrections and playback sim2sim preflight.
• Release-only diff; no runtime code is modified by this PR. The published package contains the already-merged behavior documented above.
• Version decision: this release uses the maintainer-requested exact version 1.5.3; compatibility effects are documented explicitly in the changelog rather than hidden.

Linked Work

• Issue: n/a (release process; source issues are linked in the changelog)
• Parent roadmap (when applicable): n/a
• Roadmap declared base (when applicable): n/a
• Milestone: n/a
• Base branch: main

Validation

• make test-all passed on the final local head before this PR was created or updated
• Additional task-specific validation listed below

Commands actually run:

uv lock
# Updated root unilab entry v1.5.2 -> v1.5.3

# ROCm profile, temporarily activated and restored
uv lock
# Updated ROCm root unilab entry v1.5.2 -> v1.5.3

uv lock --check
# PASS, default profile (240 packages)

uv lock --check
# PASS, ROCm profile (228 packages)

uv run pytest tests/scripts/test_check_docs.py -q
# 20 passed

UNILAB_DOCS_SKIP_AUTODOC=1 uv run --no-project --with-requirements docs/sphinx/requirements.txt \
  sphinx-build -b html -n docs/sphinx/source docs/sphinx/build/html
# build succeeded; 9 pre-existing warnings

make check
# PASS: ruff format/lint, mypy (212 files), pyright (0 errors, 1 pre-existing warning), test lint

make test-all
# PASS on release head bf9de51f
# test-cov: 2851 passed, 61 skipped, 375 deselected, 18 warnings
# benchmark smoke: 35/35 module mode and 36/36 script mode

uv build --out-dir /tmp/unilab-dist-1.5.3.2FuEFD
uvx --from twine twine check /tmp/unilab-dist-1.5.3.2FuEFD/*
# sdist and wheel PASSED

uv run python - <<'PY'
# Verified wheel contains unilab/rl/py.typed,
# unilab/rl/runtime/ipc/replay_pipelines/native_h2d_ext.cpp, and
# unilab/base/contracts/env.py.
PY

uv venv /tmp/unilab-release-venv-1.5.3.lwg4Aj
uv pip install --python /tmp/unilab-release-venv-1.5.3.lwg4Aj/bin/python \
  --reinstall --no-deps --no-binary unilab --no-cache \
  --find-links /tmp/unilab-dist-1.5.3.2FuEFD --no-index unilab
uv run --no-project --python /tmp/unilab-release-venv-1.5.3.lwg4Aj/bin/python python -c \
  'import unilab, unilab.rl.runtime.distributed.context; assert unilab.__version__ == "1.5.3"'
# installed unilab 1.5.3; ROOT_PATH/conf smoke assertions passed

Remote CI route:

• Base main: current-head CI passed on head bf9de51f: CI and Docs.
• Exact merge commit ffb7760f: dispatched CI passed.
• Tagged release workflow v1.5.3 passed; GitHub Release: https://github.com/Motphys/UniLab/releases/tag/v1.5.3.

Impact

• Backend impact: none from the release-only diff; published 1.5.3 includes the documented mapped-entity, reset, DR, and playback behavior across supported backends
• Platform impact: both
• Training effect expected: no from this release-only diff; published 1.5.3 includes the documented opt-in state-feedback and playback/reset changes

Artifacts

• W&B: n/a
• benchmark result: n/a for release-only diff; make test-all benchmark import smoke passed
• video / screenshot: n/a
• ONNX / checkpoint: n/a

Checklist

• Added or updated tests where needed (release-only change; existing source tests are unchanged and covered by make test-all)
• Updated docs if behavior or workflow changed (bilingual changelog)
• Linked the driving issue (n/a — release process)
• Noted any follow-up work explicitly (current-head CI; exact merge-commit CI; then annotated v1.5.3 tag)