default:
			panic("unexpected node type in if")
		}
		ast.Walk(f, n.Else)
		return nil
	case *ast.SelectStmt:
		// Don't annotate an empty select - creates a syntax error.
		if n.Body == nil || len(n.Body.List) == 0 {
			return nil
		}
	case *ast.SwitchStmt:
		// Don't annotate an empty switch - creates a syntax error.
		if n.Body == nil || len(n.Body.List) == 0 {
			if n.Init != nil {
				ast.Walk(f, n.Init)

push-manifest:
	docker manifest create --amend $(MANIFEST_IMAGE):$(IMAGE_TAG) $(shell echo $(ALL_OS_ARCH) | sed -e "s~[^ ]*~$(MANIFEST_IMAGE):$(IMAGE_TAG)\-&~g")
	set -x; for arch in $(ALL_ARCH.linux); do docker manifest annotate --os linux --arch $${arch} ${MANIFEST_IMAGE}:${IMAGE_TAG} ${MANIFEST_IMAGE}:${IMAGE_TAG}-linux-$${arch}; done

👆 💪 🌀 `dataclasses` ⏮️ 🎏 🆎 ✍ ⚒ 🐦 📊 📊.

💼, 👆 💪 ✔️ ⚙️ Pydantic ⏬ `dataclasses`. 🖼, 🚥 👆 ✔️ ❌ ⏮️ 🔁 🏗 🛠️ 🧾.

👈 💼, 👆 💪 🎯 💱 🐩 `dataclasses` ⏮️ `pydantic.dataclasses`, ❔ 💧-♻:

```{ .python .annotate hl_lines="1  5  8-11  14-17  23-25  28" }
{!../../../docs_src/dataclasses/tutorial003.py!}
```

1️⃣. 👥 🗄 `field` ⚪️➡️ 🐩 `dataclasses`.

2️⃣. `pydantic.dataclasses` 💧-♻ `dataclasses`.

    }
  }
}

project.applyOsgi(
  "Export-Package: okhttp3,okhttp3.internal.*;okhttpinternal=true;mandatory:=okhttpinternal",
  "Import-Package: " +
    "android.*;resolution:=optional," +
    "com.oracle.svm.core.annotate;resolution:=optional," +
    "com.oracle.svm.core.configure;resolution:=optional," +
    "dalvik.system;resolution:=optional," +
    "org.conscrypt;resolution:=optional," +
    "org.bouncycastle.*;resolution:=optional," +

    "for the main function";
  let constructor = "TFDevice::CreateResourceOpLiftingForMainFunctionPass()";
}

def AnnotateParameterReplicationPass :
    Pass<"tf-annotate-parameter-replication", "ModuleOp"> {
  let summary = "Annotate whether a ClusterFuncOp's parameters have the same data across replicas.";
  let constructor = "TFDevice::CreateAnnotateParameterReplicationPass()";
}

the device capabilites. For example, If the user specified the desired targets
are "GPU", "CPU", `conv2d` can run on both "GPU" and "CPU", we will annotate
the op `conv2d` with "GPU" since it's preferred; `pack` can only run on "CPU",
so we will annotate the op with "CPU" since "GPU" does not support this op.

#### Raise Target Subgraphs Pass

In this pass, ops will be broken down into subgraph. Those ops have the same

## 在嵌套数据结构中使用数据类

您还可以把 `dataclasses` 与其它类型注解组合在一起，创建嵌套数据结构。

还有一些情况也可以使用 Pydantic 的 `dataclasses`。例如，在 API 文档中显示错误。

本例把标准的 `dataclasses` 直接替换为 `pydantic.dataclasses`：

```{ .python .annotate hl_lines="1  5  8-11  14-17  23-25  28" }
{!../../../docs_src/dataclasses/tutorial003.py!}
```

1. 本例依然要从标准的 `dataclasses` 中导入 `field`；

2. 使用 `pydantic.dataclasses` 直接替换 `dataclasses`；

#include "tensorflow/compiler/mlir/tensorflow/ir/tf_ops.h"
#include "tensorflow/compiler/mlir/tensorflow/utils/attribute_utils.h"
#include "xla/mlir_hlo/mhlo/IR/hlo_ops.h"

#define DEBUG_TYPE "tf-tpu-annotate-dynamic-shape-inputs"

namespace mlir {
namespace TFTPU {

namespace {

#define GEN_PASS_DEF_TPUANNOTATEDYNAMICSHAPEINPUTSPASS
#include "tensorflow/compiler/mlir/tensorflow/transforms/tf_passes.h.inc"

In that case, you can simply swap the standard `dataclasses` with `pydantic.dataclasses`, which is a drop-in replacement:

```{ .python .annotate hl_lines="1  5  8-11  14-17  23-25  28" }
{!../../../docs_src/dataclasses/tutorial003.py!}
```

1. We still import `field` from standard `dataclasses`.

In diesem Fall können Sie einfach die Standard-`dataclasses` durch `pydantic.dataclasses` ersetzen, was einen direkten Ersatz darstellt:

```{ .python .annotate hl_lines="1  5  8-11  14-17  23-25  28" }
{!../../../docs_src/dataclasses/tutorial003.py!}
```

1. Wir importieren `field` weiterhin von Standard-`dataclasses`.