{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "000259b0-c4c0-49e1-982c-51b9c58da63a",
   "metadata": {},
   "source": [
    "# 1. 认识张量"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fe1497a8-7903-41ab-b5cd-5ae04054a699",
   "metadata": {},
   "source": [
    "- 标量\n",
    "- 向量（矢量）\n",
    "- 张量（变化关系）"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "acf46b28-8e90-4414-bd1b-c2b243afea68",
   "metadata": {},
   "outputs": [],
   "source": [
    "!pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cpu"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f55cf0df-851f-4e1b-ab36-9710fe2dd0bc",
   "metadata": {},
   "source": [
    "- 张量的使用与numpy的ndarray基本上差不多！（与线性代数相关的部分）"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "52b332d5-c081-4cfe-867b-de30dea8ae33",
   "metadata": {},
   "source": [
    "## 1.1. 创建张量"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "41c77bcd-8730-459f-9040-150ae67a55f5",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch import LongTensor, FloatTensor, Tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "0a45c3e5-725d-4271-916f-808891a08f73",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([1., 2., 3., 4.])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t = Tensor((1, 2, 3, 4))\n",
    "t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "c7d21c70-aa9a-42e8-804b-f82ca38a9639",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy \n",
    "\n",
    "m = numpy.array([3,4,5,6])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "828cf0d2-2474-402d-a460-855a04f29aef",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([3., 4., 5., 6.])"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t = Tensor(m)\n",
    "t"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bcd56473-184d-478c-9d3d-f4d5a4ae2f40",
   "metadata": {},
   "source": [
    "> 1. 构造器（Storage：tuple,list, numpy.ndarray, Tensor， 整数:维度，）\n",
    "> 2. 工厂模式：函数生成对象"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "b70bb183-c993-491d-aa00-041ea0e75217",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[5.2857e-18, 1.0243e-42, 0.0000e+00, 0.0000e+00],\n",
       "        [0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00],\n",
       "        [0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00]])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t = Tensor(3, 4)   # Tensor == FloatTensor  # 不稳定的\n",
    "t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "7f1745e4-38f4-4034-8aef-a11ac45fe92a",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch import tensor   # 通过类的构造器创建\n",
    "t = tensor(data=[1,2,3], dtype=float, device=\"cuda:0\", requires_grad=True)   # 参数是稳定 : cpu, cuda\n",
    "# t.backward()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "a8842c37-3ed4-45c9-a940-ca1c5126c680",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(1)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import torch\n",
    "t = torch.tensor(1)  # 0维张量\n",
    "t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "fc31fbcf-31d3-4e9f-ad42-5647c4b0f5bc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(1, dtype=int64), 1)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = t.detach().numpy() # numpy向量 t.cpu().detach().numpy()\n",
    "b = t.detach().item()  # 标量\n",
    "a, b"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82295a52-583d-4c67-9e61-56f099671490",
   "metadata": {},
   "source": [
    "## 1.2. 张量的操作"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "0f5e12ee-b0ed-4d3f-b29a-130679a000af",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([3, 3, 4])"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t = torch.tensor([\n",
    "    [[1, 2, 3, 4],\n",
    "    [5, 6, 7, 8],\n",
    "    [9, 0, 1, 2]],\n",
    "    [[1, 2, 3, 4],\n",
    "    [5, 6, 7, 8],\n",
    "    [9, 0, 1, 2]],\n",
    "    [[1, 2, 3, 4],\n",
    "    [5, 6, 7, 8],\n",
    "    [9, 0, 1, 2]]\n",
    "], dtype=float)\n",
    "# 与numpy完全一样\n",
    "t.dtype\n",
    "t.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "6b0df8a4-6a35-4fff-9cb1-2082e9c6c360",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([2., 6., 0.], dtype=torch.float64)"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t[0, ..., 1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "13152bee-ab2b-4251-ba6a-ad3862c95f5f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1., 2., 3., 4.],\n",
       "        [5., 6., 7., 8.]], dtype=torch.float64)"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "t[0, 0:2, ...]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e606f7f2-f90a-4d23-8e52-a4f666256a79",
   "metadata": {},
   "source": [
    "## 1.3. 张量的运算"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "be38db75-1940-4421-ae48-22f3dde14949",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 1,  4,  9, 16],\n",
       "        [25, 36, 49, 64]])"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = tensor([\n",
    "    [1, 2, 3, 4],\n",
    "    [5, 6, 7, 8]    \n",
    "])\n",
    "\n",
    "b = tensor([\n",
    "    [1, 2, 3, 4],\n",
    "    [5, 6, 7, 8]    \n",
    "])\n",
    "\n",
    "a * b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "98062436-e5cf-4db5-8b73-da3e4dc7cb4c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 5, 10, 15, 20],\n",
       "        [25, 30, 35, 40]])"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a * 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "79661746-de23-47e2-9290-de5be3097335",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 1,  4,  9, 16],\n",
       "        [25, 36, 49, 64]])"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a ** 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "d4115674-f5b8-47a7-a332-9e86b6648e28",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 30,  70],\n",
       "        [ 70, 174]])"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a.matmul(b.T)   # mul（元素乘）  matmul(内积)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "da49a3a7-4676-4a12-9de0-3537bbc30f95",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\u001b[1;31mSignature:\u001b[0m \u001b[0mabs\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mx\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m/\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
       "\u001b[1;31mDocstring:\u001b[0m Return the absolute value of the argument.\n",
       "\u001b[1;31mType:\u001b[0m      builtin_function_or_method"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "abs?"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a2cd297e-9138-4c57-87d2-2d0dbc8d4372",
   "metadata": {},
   "source": [
    "## 1.4. 张量的求导数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "95aa038a-2f04-40cd-aede-1b71ee8d5454",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n",
      "None\n",
      "None\n",
      "False\n"
     ]
    }
   ],
   "source": [
    "x = torch.tensor([1.0, 2.0, 3.3])\n",
    "print(x.is_leaf)\n",
    "print(x.grad)\n",
    "print(x.grad_fn)\n",
    "print(x.requires_grad)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "05f96a6a-7539-4ad2-bef6-2ad59add1bf1",
   "metadata": {},
   "outputs": [],
   "source": [
    "x.requires_grad=True\n",
    "y = 3* (x ** 2) + 4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "d9d7cf2d-6773-44c5-a651-65ad4dfa561c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n",
      "None\n",
      "None\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "print(x.is_leaf)\n",
    "print(x.grad)\n",
    "print(x.grad_fn)\n",
    "print(x.requires_grad)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "382613b1-a129-4a19-85a3-776a6e1fa2e9",
   "metadata": {},
   "outputs": [],
   "source": [
    "y.sum().backward()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "638c2090-0bad-4255-8e04-fc758c79de2f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n",
      "tensor([ 6.0000, 12.0000, 19.8000])\n",
      "None\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "print(x.is_leaf)\n",
    "print(x.grad)\n",
    "print(x.grad_fn)\n",
    "print(x.requires_grad)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6bc1b84b-15fb-459f-acef-49d260f36138",
   "metadata": {},
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}