feat(dr)!: unify capability paths with unisim 2.2.3

docs/sphinx/source/adr/ADR-0021-genesis-device-resident-dr-lifecycle.md

@@ -13,7 +13,8 @@ orphan: true
 - Superseded by: None
 - Roadmap: [Issue #2520](https://github.com/Motphys/UniLab/issues/2520);
   FlashMimic subset extension:
-  [Issue #2521](https://github.com/Motphys/UniLab/issues/2521)
+  [Issue #2521](https://github.com/Motphys/UniLab/issues/2521); unified capability sets:
+  [Issue #2550](https://github.com/Motphys/UniLab/issues/2550)
 
 ## Context
 
@@ -32,19 +33,19 @@ before this integration; the authoritative adapter ADR and implementation live i
 
 UniLab consumes the unified contract as device-only Genesis runtime DR:
 
-- Genesis declares reset terms only in `supported_device_reset_terms`, and interval terms only
-  in `supported_device_interval_terms`. The host term sets are empty, including for portable
-  scenes and sessions without the exact CUDA/AMDGPU lifecycle.
-- Manager capability negotiation selects the plane from the backend-bound tensor device. It never
-  treats default host capability wording as a fallback for a device-resident runtime.
-- The packed reset device is declared before cold event-term binds, so owner terms negotiate
-  against the same plane that will commit after backend materialization.
+- Genesis declares reset terms in `supported_reset_terms` and interval terms in
+  `supported_interval_terms`. These single sets enumerate supported terms without splitting them
+  into host/device public capability planes. They do not imply that Genesis accepts host payloads.
+- Manager capability negotiation is plane-independent. The backend-bound tensor device still
+  determines operand placement, while the backend owns execution and any adapter-internal
+  conversion for an advertised term.
+- The packed reset device is declared before cold event-term binds, so owner terms construct and
+  commit operands on the same device after backend materialization.
 - Reset DR is limited to the audited device lifecycle (`body_mass`, `base_mass_delta`,
   `body_ipos`, `base_com_offset`, `kp`, `kd`); interval DR is limited to `body_force`,
   `body_torque`, world-frame root linear velocity delta, and root-local angular velocity delta.
-  Unsupported, host-plane, mixed-plane, and CPU/host payloads fail closed with the term and plane
-  named.
-- Genesis owner YAML enables `pd_gains` only because that device-plane packed lane declares
+  Unsupported terms and CPU/host payloads fail closed with explicit backend errors.
+- Genesis owner YAML enables `pd_gains` only because that device-resident packed lane declares
   `kp`/`kd`. It does not re-enable broad portable host setters, geometry, gravity, DOF, or the
   opaque legacy push descriptor. Startup body mass and the six-axis velocity-delta event use
   their audited public device lifecycle; absolute `geom_friction` remains intentionally
@@ -55,21 +56,21 @@ UniLab consumes the unified contract as device-only Genesis runtime DR:
 
 ## Stable Contracts
 
-- `DomainRandomizationCapabilities` plane sets are authoritative; aggregate projections keep
-  `dr.reset.*`/`dr.interval.*` and `dr.device.reset.*`/`dr.device.interval.*` independent.
+- The single `DomainRandomizationCapabilities.supported_reset_terms` and
+  `supported_interval_terms` sets are authoritative; execution topology is not an alternate DR
+  term set.
 - The Manager packed reset transaction declares its device before reset event construction.
 - Genesis interval dispatch uses public `IntervalTermOp` values only; backend IDs are Torch int64
   on the backend device and wrench payloads are float32 on that same device.
 - Genesis owners cannot declare backend capability by comments or backend-name branches; owner
-  YAML may only enable terms whose plane the current backend declares.
+  YAML may only enable terms that the current backend declares.
 
 ## Alternatives Considered
 
-- **Default to the historical host plane before packed-device declaration:** rejected. It made a
-  portable host setter look valid during cold binds and could accept a term that the eventual
-  device commit must reject.
-- **Convert host operands inside each backend adapter:** rejected as the Manager contract. It
-  reintroduces hidden hot-path H2D detours and plane ambiguity at every adapter boundary.
+- **Keep separate host/device capability sets:** rejected. It made a portable host setter look
+  valid during cold binds and leaked adapter execution topology into every Manager consumer.
+- **Select capability sets from the Manager payload device:** rejected. Term support is a public
+  backend capability; payload placement and adapter-internal conversion are execution details.
 - **Keep `events.pd_gains: null` until every Genesis reset term is device-capable:** rejected.
   `kp`/`kd` have one audited lifecycle; unrelated unsupported terms already fail closed and need
   no owner-level blanket disable.
@@ -82,16 +83,17 @@ UniLab consumes the unified contract as device-only Genesis runtime DR:
   reset writes, and six-axis velocity kicks. Its foot-friction event stays disabled.
 - Portable Genesis scenes have a smaller reset contract, and removed portable host runtime paths
   have no compatibility alias in UniLab.
-- Owners must sample device-resident operands on the Manager RNG/backend device. CPU-only Genesis
-  sessions do not advertise runtime DR.
+- Owners must sample device-resident operands on the Manager RNG/backend device. Unified term
+  sets do not add CPU/host execution support to Genesis.
 - Adding a Genesis reset or interval term requires first extending the audited device lifecycle
   and capability declaration in `unisim-core`, then updating owners/docs here.
 
 ## Evidence In Repo
 
-- `src/unilab/base/entity.py` — interval plane negotiation and device-resident interval ops.
+- `src/unilab/base/entity.py` — plane-independent interval capability checks and device-resident
+  interval ops.
 - `src/unilab/envs/manager_based_rl_env.py` — packed reset device declaration before event binds.
-- `src/unilab/base/reset_state.py` — plane-aware capability checks.
+- `src/unilab/base/reset_state.py` — unified reset-term capability checks.
 - `src/unilab/conf/{ppo,sac}/task/g1_walk_flat/genesis.yaml` — walk owner parity.
 - `src/unilab/conf/sac/task/g1_motion_tracking/genesis.yaml` — FlashMimic DR subset owner.
 - `tests/base/test_genesis_backend.py`; `tests/envs/mdp/test_events.py`;

docs/sphinx/source/en/2-user_guide/3-backends/5-genesis.md

@@ -52,10 +52,11 @@ compensates on the materialize cold path — hot paths never parse XML:
   reproduction of it.
 - **Actuators**: `<position kp kv>` actuators import losslessly and are
   driven with `control_dofs_position`; kp/kd reset randomization is supported
-  on the device plane (the DR capability sets declare only the measured device
-  reset terms: body mass, base-mass delta, body CoM / base-CoM delta, kp,
-  kd; the device interval set additionally covers body force, body torque,
-  and world-frame root linear plus root-local angular velocity deltas).
+  through the device-resident lifecycle. The unified DR capability sets declare
+  only the measured reset terms: body mass, base-mass delta, body CoM /
+  base-CoM delta, kp, kd; the interval set additionally covers body force,
+  body torque, and world-frame root linear plus root-local angular velocity
+  deltas.
 
 ## Prerequisites
 
@@ -159,9 +160,10 @@ degrading:
   construction/plan time. Use the six-axis velocity-delta event, which commits
   world-frame linear and root-local angular deltas through Genesis public
   state without touching qpos, controls, solver state, or unrelated rows.
-- **Host-plane runtime DR**: Genesis accepts reset and interval payloads only
-  on its backend-bound CUDA/AMDGPU device. Host payloads fail closed with the
-  term and plane named; there is no NumPy or host-bridge fallback.
+- **Host-payload runtime DR**: Genesis accepts reset and interval payloads only
+  on its backend-bound CUDA/AMDGPU device. Host payloads fail closed with an
+  explicit backend error; the unified term sets do not imply host execution
+  support, and there is no NumPy or host-bridge fallback.
 - **Repeated `gs.init`** in one process (see above).
 
 ## Cross-Backend Migration (sim2sim)

docs/sphinx/source/en/2-user_guide/5-domain_randomization/0-index.md

@@ -23,7 +23,7 @@ These three paths correspond to three lifecycle classes:
 
 1. Reset/interval randomization consists of `events:` manager terms in the owner YAML, executed uniformly by the manager lifecycle.
 2. Manager-Based owners declare reset behavior through Hydra command/event terms. G1 motion reset perturbations belong to the device-resident `MotionCommandCfg`, while task owners add `EventTermCfg` reset and interval terms.
-3. `TensorResetRandomizationPayload` (the torch-first reset payload) expresses curated reset terms; a backend must advertise every requested term on the matching host/device plane and own its derived-quantity obligation.
+3. `TensorResetRandomizationPayload` (the torch-first reset payload) expresses curated reset terms; a backend must advertise every requested term in its unified capability set and own its derived-quantity and device-execution obligations.
 4. `MotrixBackend` currently supports `base_mass_delta`, `base_com_offset`, `kp`, `kd`, and interval push; and it requires all model actuators to be position actuators during initialization.
 5. Fixed mesh/tool identity is declared by `env.fixed_model_variants`; reset-time geometry fields remain behind backend capability declarations and never change that identity.
 

docs/sphinx/source/en/4-developer_guide/2-contracts/4-dr_contract.md

@@ -33,23 +33,24 @@ provider protocol and the environment does not carry a DR manager.
 
 Backend differences are explicit capabilities, not task-side branches:
 
-- `DomainRandomizationCapabilities.supported_reset_terms` — host plane
-  (`torch.cpu` payloads)
-- `supported_interval_terms` — host plane
-- `supported_device_reset_terms` / `supported_device_interval_terms` — device
-  plane (CUDA payloads); an empty set fails closed
+- `DomainRandomizationCapabilities.supported_reset_terms` — every reset term the
+  backend supports through the public DR path
+- `supported_interval_terms` — every interval term the backend supports through
+  the public DR path
 - fixed-variant layouts
 - per-environment playback support
 - `TensorLifecycleCapabilities.packed_reset` with
   `SimBackend.bind_packed_reset()`: required for the Manager-Based reset and
   startup DR lifecycle; a missing declaration fails closed at env
   construction.
 
-The payload tensors' device selects the plane: `torch.cpu` payloads negotiate
-the host term sets and CUDA payloads the device term sets. An unadvertised
-requested term fails closed with the backend, term, and plane named.
-Manager code never imports MuJoCo or mjbatch and never accesses a backend model
-or pool.
+These are single, plane-independent term sets. Manager capability negotiation
+does not inspect the payload device or pass a plane. CPU/GPU execution and any
+required internal conversion belong to the backend adapter, not to separate
+public capability sets. An unadvertised requested term fails closed with the
+backend and term named; a backend may also reject payloads that violate its
+device or lifecycle contract. Manager code never imports MuJoCo or mjbatch and
+never accesses a backend model or pool.
 
 ## Reset Payload And Defaults
 
@@ -89,10 +90,10 @@ the former UniLab-side MuJoCo recompilation used to obtain inertia defaults.
 Interval plans are term-descriptor based: `IntervalRandomizationPlan.ops`
 carries `IntervalTermOp` entries from `unisim.dr.interval`. Payloads are
 torch-first: `IntervalTermOp.payload` is a torch tensor or a
-`TensorIntervalPayload` whose device selects the plane, and push is
-producer-sampled on every backend — backend-internal global-RNG push sampling
-no longer exists. Builtin payload contracts are enforced by
-`IntervalTermOp.validate`; unknown backend-owned custom terms pass through to
+`TensorIntervalPayload`; push is producer-sampled on every backend —
+backend-internal global-RNG push sampling no longer exists. Builtin payload
+contracts are enforced by `IntervalTermOp.validate`; unknown backend-owned
+custom terms pass through to
 that backend's handler table. Plans that cross spawn collectors carry sampling
 ranges rather than tensors; device payloads are process-local.
 

docs/sphinx/source/zh_CN/2-user_guide/3-backends/5-genesis.md

@@ -47,11 +47,11 @@ materialize 冷路径补偿——热路径从不解析 XML：
   contact sensor 变为 per-link net contact force 阈值（1 N）——这是对
   geom 对 `found` 语义的近似，不是复现。
 - **执行器**：`<position kp kv>` 执行器无损导入，用
-  `control_dofs_position` 驱动；kp/kd reset 随机化在 device plane 受支持
-  （DR 能力集只声明实测的 device reset 项：body mass、kp、kd，以及
-  base-mass delta、body CoM / base-CoM delta、kp、kd；device interval 集
-  另覆盖 body force、body torque，以及世界系 root 线速度与 root 本体系
-  角速度 delta）。
+  `control_dofs_position` 驱动；kp/kd reset 随机化通过 device-resident
+  生命周期受支持。统一 DR 能力集只声明实测的 reset 项：body mass、
+  base-mass delta、body CoM / base-CoM delta、kp、kd；interval 集另覆盖
+  body force、body torque，以及世界系 root 线速度与 root 本体系角速度
+  delta。
 
 ## 前置条件
 

docs/sphinx/source/zh_CN/4-developer_guide/2-contracts/4-dr_contract.md

@@ -29,21 +29,20 @@ Manager-Based event term 是 UniLab 唯一的 DR lifecycle。任务 provider 协
 
 Backend 差异是显式 capability，不是 task-side 分支：
 
-- `DomainRandomizationCapabilities.supported_reset_terms`——host 平面
-  （`torch.cpu` payload）
-- `supported_interval_terms`——host 平面
-- `supported_device_reset_terms` / `supported_device_interval_terms`——device
-  平面（CUDA payload）；空集合 fail closed
+- `DomainRandomizationCapabilities.supported_reset_terms`——backend 通过公开 DR
+  路径支持的全部 reset term
+- `supported_interval_terms`——backend 通过公开 DR 路径支持的全部 interval term
 - fixed-variant layout
 - per-environment playback 支持
 - `TensorLifecycleCapabilities.packed_reset` 与 `SimBackend.bind_packed_reset()`：
   Manager-Based reset 与 startup DR 生命周期的必需 capability；缺失声明会在 env
   构造时 fail closed。
 
-payload tensor 的设备决定平面：`torch.cpu` payload 走 host term 集合协商，CUDA
-payload 走 device term 集合。未在对应平面声明的请求 term 会携带 backend、term
-与平面名称 fail closed。Manager code 不 import MuJoCo 或 mjbatch，也不访问
-backend model/pool。
+这两个集合是单一且与 plane 无关的 term 集合。Manager capability 协商不检查
+payload 所在设备，也不传递 plane。CPU/GPU 执行及所需的内部转换由 backend
+adapter 负责，不再拆成不同的公开 capability 集合。未声明的请求 term 会携带
+backend 与 term 名称 fail closed；backend 仍会拒绝违反其设备或生命周期契约的
+payload。Manager code 不 import MuJoCo 或 mjbatch，也不访问 backend model/pool。
 
 ## Reset Payload 与默认值
 
@@ -76,8 +75,8 @@ closed。该边界移除了 UniLab 侧为 inertia 默认值重新 compile MuJoCo
 
 Interval plan 基于 term descriptor：`IntervalRandomizationPlan.ops` 携带来自
 `unisim.dr.interval` 的 `IntervalTermOp`。payload 为 torch-first：
-`IntervalTermOp.payload` 是 torch tensor 或 `TensorIntervalPayload`，其设备决定
-平面；push 在所有后端统一为生产方采样——后端内部的全局 RNG push 采样已不存在。
+`IntervalTermOp.payload` 是 torch tensor 或 `TensorIntervalPayload`；push 在所有
+后端统一为生产方采样——后端内部的全局 RNG push 采样已不存在。
 内置 payload contract 由 `IntervalTermOp.validate` 强制；未知 backend-owned
 custom term 传给该 backend handler table。跨 spawn collector 的 plan 携带采样
 范围而不是 tensor；device payload 是进程本地的。

pyproject.rocm.toml

@@ -23,7 +23,7 @@ classifiers = [
 requires-python = ">=3.10,<3.14"
 dependencies = [
     "numpy",
-    "unisim-core==2.2.0",
+    "unisim-core==2.2.3",
     "prettytable>=3.10",
     "torch==2.14.0",
     "triton-rocm==3.8.0 ; sys_platform == 'linux' and platform_machine == 'x86_64'",
@@ -68,7 +68,7 @@ mujoco = [
     "mujoco~=3.11.0",
     "mjbatch-uni~=0.2.5",
 ]
-motrix = ["unisim-core[motrix]==2.2.0"]
+motrix = ["unisim-core[motrix]==2.2.3"]
 genesis = ["genesis-world==1.3.3"]
 # Keep the optional tensor-ray extra aligned with pyproject.toml.
 ray = ["uniray[warp]==0.1.0"]

pyproject.toml

@@ -33,7 +33,7 @@ classifiers = [
 requires-python = ">=3.10,<3.14"
 dependencies = [
     "numpy",
-    "unisim-core==2.2.0",
+    "unisim-core==2.2.3",
     "prettytable>=3.10",
     # A range (not an exact pin) lets ROCm users substitute a ROCm torch
     # build; uv.lock pins the tested CUDA builds via tool.uv.sources.
@@ -102,12 +102,12 @@ newton = [
     "pyglet>=2.1.6,<3",
     "imgui-bundle>=1.92.0",
 ]
-motrix = ["unisim-core[motrix]==2.2.0"]
+motrix = ["unisim-core[motrix]==2.2.3"]
 genesis = ["genesis-world==1.3.3"]
 # Wheels are CPython 3.12/3.13 Linux x86_64 only; the marker keeps uv lock
 # resolvable for the other required-environments.
 superdex = [
-    "unisim-core[superdex]==2.2.0 ; python_version >= '3.12' and sys_platform == 'linux' and platform_machine == 'x86_64'",
+    "unisim-core[superdex]==2.2.3 ; python_version >= '3.12' and sys_platform == 'linux' and platform_machine == 'x86_64'",
 ]
 # Tensor ray perception is optional and is discovered through UniSim's public
 # caster factory. The PyPI distribution is `uniray`; the import namespace is

src/unilab/base/entity.py

@@ -42,7 +42,6 @@
     RESET_TERM_GEOM_FRICTION,
     RESET_TERM_KD,
     RESET_TERM_KP,
-    DataPlaneName,
     IntervalRandomizationPlan,
     IntervalTermOp,
 )
@@ -3807,9 +3806,7 @@ def _bind_root_velocity_delta(self, *, angular: bool, term_name: str) -> None:
             if angular
             else INTERVAL_TERM_BODY_LINEAR_VELOCITY_DELTA
         )
-        supported = capabilities.supports_interval_term(
-            term, plane=self._interval_capability_plane()
-        )
+        supported = capabilities.supports_interval_term(term)
         if not supported:
             raise self._capability_error(
                 "interval root velocity delta",
@@ -3965,15 +3962,12 @@ def bind_body_wrench(
             capabilities = self._backend.get_dr_capabilities()
         except (AttributeError, NotImplementedError) as exc:
             raise self._capability_error("interval body wrench", str(exc)) from exc
-        plane = self._interval_capability_plane()
-        if not capabilities.supports_interval_term(INTERVAL_TERM_BODY_FORCE, plane=plane):
+        if not capabilities.supports_interval_term(INTERVAL_TERM_BODY_FORCE):
             raise self._capability_error(
                 "interval body wrench",
                 f"EventManager term '{term_name}' requested an unsupported backend capability",
             )
-        if torque and not capabilities.supports_interval_term(
-            INTERVAL_TERM_BODY_TORQUE, plane=plane
-        ):
+        if torque and not capabilities.supports_interval_term(INTERVAL_TERM_BODY_TORQUE):
             raise self._capability_error(
                 "interval body wrench",
                 f"EventManager term '{term_name}' requested an unsupported backend "
@@ -3993,11 +3987,6 @@ def _interval_operand_device(self) -> torch.device:
         self._interval_operand_device_value = device
         return device
 
-    def _interval_capability_plane(self) -> DataPlaneName:
-        """Return the capability plane used by this entity's interval payloads."""
-
-        return "device" if self._interval_operand_device().type != "cpu" else "host"
-
     def apply_body_wrench_to_sim(
         self,
         forces: np.ndarray | torch.Tensor,