{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 网络内构造常量\n",
    "\n",
    "[![下载Notebook](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/r1.9/resource/_static/logo_notebook.png)](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/r1.9/tutorials/experts/zh_cn/network/mindspore_constexpr.ipynb) [![下载样例代码](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/r1.9/resource/_static/logo_download_code.png)](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/r1.9/tutorials/experts/zh_cn/network/mindspore_constexpr.py) [![查看源文件](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/r1.9/resource/_static/logo_source.png)](https://gitee.com/mindspore/docs/blob/r1.9/tutorials/experts/source_zh_cn/network/constexpr.ipynb)\n",
    "\n",
    "`mindspore.ops.constexpr`中提供了一个@constexpr的Python\n",
    "装饰器，该装饰器可以用于修饰一个函数，该函数在编译阶段将会通过Python解释器执行，最终在MindSpore的类型推导阶段被常量折叠成为ANF图的一个常量节点(ValueNode)。\n",
    "\n",
    "由于该函数在MindSpore编译时期进行，所以使用@constexpr函数时，要求输入函数的入参必须为一个编译时刻就能够确定的常量值，否则如果该函数入参为一个编译时刻无法确定的值，那么入参将会为None，从而可能导致函数输出与预期不符。\n",
    "\n",
    "当@constexpr的入参为提前明确的参数时可以实现一些在construct函数中不支持的操作。比如根据shape创建Tensor等。\n",
    "\n",
    "为了避免出现@constexpr输入为编译时无法确定的值，可以在内部进行对None的判断处理，避免一些未知错误。\n",
    "\n",
    "代码样例如下:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-04T02:52:51.150509Z",
     "start_time": "2022-01-04T02:52:48.980586Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[7. 6. 3.]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from mindspore.ops import constexpr\n",
    "import mindspore.ops as ops\n",
    "import mindspore.nn as nn\n",
    "from mindspore import Tensor\n",
    "import mindspore\n",
    "\n",
    "\n",
    "@constexpr\n",
    "def construct_tensor(x):\n",
    "    if x is None:\n",
    "        raise ValueError(\"input is an unknown value\")\n",
    "    return Tensor(np.array(x), dtype=mindspore.float32)\n",
    "\n",
    "\n",
    "class Net(nn.Cell):\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        self.relu = ops.ReLU()\n",
    "\n",
    "    def construct(self, x):\n",
    "        return self.relu(construct_tensor(ops.shape(x)))\n",
    "\n",
    "\n",
    "net = Net()\n",
    "x = Tensor(np.random.random([7, 6, 3]))\n",
    "out = net(x)\n",
    "print(out)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "如下所示，如果我们将Net改成输入为编译时无法确定的值时，则会抛出异常。由于construct_tensor输入为运行ReLU时才能确定的值。在constexpr中会抛出ValueError。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "```python\n",
    "@constexpr\n",
    "def construct_tensor(x):\n",
    "    if x is None:\n",
    "        raise ValueError(\"input is an unknown value\")\n",
    "    return Tensor(np.array(x), dtype=mindspore.float32)\n",
    "\n",
    "\n",
    "class Net(nn.Cell):\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        self.relu = ops.ReLU()\n",
    "\n",
    "    def construct(self, x):\n",
    "        return self.relu(construct_tensor(self.relu(x)))\n",
    "\n",
    "\n",
    "net = Net()\n",
    "x = Tensor(np.random.random([7, 6, 3]))\n",
    "out = net(x)\n",
    "print(out)\n",
    "```\n",
    "\n",
    "运行结果如下:\n",
    "\n",
    "```text\n",
    "ValueError: input is an unknown value\n",
    "```"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "MindSpore",
   "language": "python",
   "name": "mindspore"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}