{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "2579fa00-0dc6-4672-8a96-5a6bb315881f",
   "metadata": {},
   "source": [
    "# 张量视图\n",
    "\n",
    "[![查看源文件](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/website-images/master/resource/_static/logo_source.svg)](https://gitee.com/mindspore/docs/blob/master/docs/mindspore/source_zh_cn/design/tensor_view.ipynb)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "d2f9816d-fa7a-4524-accb-4fa3c30a2dc4",
   "metadata": {},
   "source": [
    "张量视图（Tensor Views）是指一个Tensor经过[view类算子](#view类算子)的返回值，与该Tensor共享内存数据，避免了数据复制，从而可以进行快速且内存高效的重塑、切片和逐元素操作。\n",
    "\n",
    "例如，要获取Tensor t的视图，可以用t.view(...)。\n",
    "\n",
    "```python\n",
    "from mindspore import Tensor\n",
    "import numpy as np\n",
    "t = Tensor(np.array([[1, 2, 3], [2, 3, 4]], dtype=np.float32))\n",
    "b = t.view((3, 2))\n",
    "# Modifying view Tensor changes base Tensor as well.\n",
    "b[0][0] = 100\n",
    "print(t[0][0])\n",
    "# 100\n",
    "```\n",
    "\n",
    "由于视图与其原来的Tensor共享底层数据，因此如果在视图中修改数据，它也会反映在原来的Tensor中。\n",
    "\n",
    "通常，MindSpore算子操作会返回一个新的Tensor作为输出，例如[add()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.add.html)。\n",
    "但在视图操作的情况下，输出是输入Tensor的视图，以避免不必要的数据复制。创建视图时不会发生数据移动，视图Tensor只是改变它解析相同数据的方式。\n",
    "使用Tensor Views可能会使内存储存连续的Tensor产生内存存储非连续的Tensor。用户应格外注意，因为连续性可能会对性能产生隐式影响。[transpose()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.transpose.html)是一个常见的例子。\n",
    "\n",
    "```python\n",
    "from mindspore import Tensor\n",
    "import numpy as np\n",
    "base = Tensor([[0, 1], [2, 3]])\n",
    "base.is_contiguous()\n",
    "# True\n",
    "t = base.transpose(1, 0) # t is a view of base. No data movement happened here.\n",
    "t.is_contiguous()\n",
    "# False\n",
    "# To get a contiguous Tensor, call `.contiguous()` to enforce\n",
    "# copying data when `t` is not contiguous.\n",
    "c = t.contiguous()\n",
    "c.is_contiguous()\n",
    "# True\n",
    "```\n",
    "\n",
    "## view类算子\n",
    "\n",
    "作为参考，以下是MindSpore中支持view特性算子的完整列表：\n",
    "\n",
    "[broadcast_to()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.broadcast_to.html)\n",
    "\n",
    "[diagonal()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.diagonal.html)\n",
    "\n",
    "[expand_as()](https://www.mindspore.cn/docs/zh-CN/master/api_python/mindspore/Tensor/mindspore.Tensor.expand_as.html)\n",
    "\n",
    "[movedim()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.movedim.html)\n",
    "\n",
    "[narrow()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.narrow.html)\n",
    "\n",
    "[permute()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.permute.html)\n",
    "\n",
    "[squeeze()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.squeeze.html)\n",
    "\n",
    "[transpose()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.transpose.html)\n",
    "\n",
    "[t()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.t.html)\n",
    "\n",
    "[T](https://www.mindspore.cn/docs/zh-CN/master/api_python/mindspore/Tensor/mindspore.Tensor.T.html)\n",
    "\n",
    "[unsqueeze()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.unsqueeze.html)\n",
    "\n",
    "[view()](https://www.mindspore.cn/docs/zh-CN/master/api_python/mindspore/Tensor/mindspore.Tensor.view.html)\n",
    "\n",
    "[view_as()](https://www.mindspore.cn/docs/zh-CN/master/api_python/mindspore/Tensor/mindspore.Tensor.view_as.html)\n",
    "\n",
    "[unbind()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.unbind.html)\n",
    "\n",
    "[split()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.split.html)\n",
    "\n",
    "[hsplit()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.hsplit.html)\n",
    "\n",
    "[vsplit()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.vsplit.html)\n",
    "\n",
    "[tensor_split()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.tensor_split.html)\n",
    "\n",
    "[swapaxes()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.swapaxes.html)\n",
    "\n",
    "[swapdims()](https://www.mindspore.cn/docs/zh-CN/master/api_python/ops/mindspore.ops.swapdims.html)"
   ]
  }
 ],
 "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"
  },
  "vscode": {
   "interpreter": {
    "hash": "8c9da313289c39257cb28b126d2dadd33153d4da4d524f730c81a4aaccbd2ca7"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}