physicalai-runtime-loading-exported-policies

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Loading Exported Policies

加载导出的策略

Runtime loads Studio export directories (or Hub snapshots that mirror them) through

InferenceModel

src/physicalai/inference/model.py

. Manifest parsing lives in

src/physicalai/inference/manifest.py

; backends register in

src/physicalai/inference/adapters/registry.py

(

onnx

→

.onnx

openvino

→

.xml

). Hub downloads use

src/physicalai/inference/utils/_hub.py

Runtime通过

src/physicalai/inference/model.py

中的

InferenceModel

加载Studio导出目录（或镜像这些目录的Hub快照）。清单解析逻辑位于

src/physicalai/inference/manifest.py

；后端在

src/physicalai/inference/adapters/registry.py

中注册（

onnx

对应

.onnx

，

openvino

对应

.xml

）。Hub下载功能使用

src/physicalai/inference/utils/_hub.py

。

Workflow

工作流程

Identify the artifact: local export directory or Hub
```
repo_id
```
, expected backend, and whether the user needs
```
select_action
```
vs
```
predict_action_chunk
```
.
- Done when: load path and API entry point are chosen before editing code.
Load with auto-detection first (local):
python
```
from physicalai.inference import InferenceModel
model = InferenceModel("./exports/act_policy")
```
- Done when:
```
model
```
  constructs without explicit
```
backend=
```
  when artifacts match a registered extension.
Hub load when the package is published:
python
```
model = InferenceModel.from_pretrained("OpenVINO/act-fp16-ov", revision="<commit-sha>")
```
- Done when: revision is pinned for reproducibility when security or CI matters.

Explicit backend only when auto-detection is ambiguous:

python

model = InferenceModel("./exports/act_policy", backend="openvino", device="CPU")

Validate structure against
```
references/export-load-contract.md
```
and backend notes (
```
references/onnx.md
```
,
```
references/openvino.md
```
).
- Done when: manifest, model file, and processor artifacts resolve under the export directory.
Smoke inference without owning robot timing:
python
```
model.reset()
action = model.select_action(observation)
```
- Done when: one forward pass succeeds on representative observation keys/shapes. For hardware loops, hand off to
```
physicalai-runtime-running-policy-on-robot
```
  .

识别工件：本地导出目录或Hub
```
repo_id
```
、预期后端，以及用户是否需要
```
select_action
```
或
```
predict_action_chunk
```
。
- 完成标志：在编辑代码前确定加载路径和API入口点。
先使用自动检测加载（本地）：
python
```
from physicalai.inference import InferenceModel
model = InferenceModel("./exports/act_policy")
```
- 完成标志：当工件匹配已注册的扩展名时，无需显式指定
```
backend=
```
  即可成功构建
```
model
```
  。
当包已发布时从Hub加载：
python
```
model = InferenceModel.from_pretrained("OpenVINO/act-fp16-ov", revision="<commit-sha>")
```
- 完成标志：在安全性或CI相关场景中，固定版本号以确保可复现性。

仅在自动检测存在歧义时指定显式后端：

python

model = InferenceModel("./exports/act_policy", backend="openvino", device="CPU")

对照
references/export-load-contract.md
和后端说明（
references/onnx.md
、
references/openvino.md
）验证结构。
- 完成标志：清单、模型文件和处理器工件可在导出目录下正常解析。
脱离机器人时序限制进行快速推理测试：
python
```
model.reset()
action = model.select_action(observation)
```
- 完成标志：针对代表性观测键/形状的一次前向传递成功。对于硬件循环场景，移交至
```
physicalai-runtime-running-policy-on-robot
```
  处理。

Validation loop

验证循环

bash

uv run pytest tests/unit/inference/test_model.py tests/unit/inference/test_manifest.py -q

For adapter changes, add or run targeted tests under

tests/unit/inference/

bash

uv run pytest tests/unit/inference/test_model.py tests/unit/inference/test_manifest.py -q

对于适配器变更，请在

tests/unit/inference/

下添加或运行针对性测试。

Required checks

必要检查

Export directory contains backend model file(s) and
```
manifest.json
```
consistent with
```
docs/reference/manifest-schema.md
```
.
Metadata input/output/feature names align with preprocessors in the manifest.
Optional backends fail with clear install guidance (
```
onnxruntime
```
, OpenVINO).
Do not document a backend unless an adapter is registered in
```
src/physicalai/inference/adapters/
```
.
Hub loads must not log tokens; prefer pinned
```
revision=
```
for production docs.

导出目录包含后端模型文件和符合
```
docs/reference/manifest-schema.md
```
规范的
```
manifest.json
```
。
元数据中的输入/输出/特征名称与清单中的预处理器对齐。
可选后端加载失败时需提供清晰的安装指引（
```
onnxruntime
```
、OpenVINO）。
除非适配器已在
```
src/physicalai/inference/adapters/
```
中注册，否则不得记录该后端。
Hub加载不得记录令牌；生产文档中优先使用固定的
```
revision=
```
参数。

References

参考资料

```
references/export-load-contract.md
```
— consumer-side contract (coordinate with Studio export skill).

references/onnx.md

references/openvino.md

— adapter constraints.

```
references/export-load-contract.md
```
—— 消费端契约（与Studio导出功能协同）。

references/onnx.md

、

references/openvino.md

—— 适配器约束条件。