ch2.2.4 模型定义.ipynb · 王志伟/人工智能导论2023年

forked from 人工智能导论/人工智能导论2023年

加入 Gitee

与超过 1200万开发者一起发现、参与优秀开源项目，私有仓库也完全免费：）

该仓库未声明开源许可证文件（LICENSE），使用请关注具体项目描述及其代码上游依赖。

克隆/下载

ch2.2.4 模型定义.ipynb 7.02 KB

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 课堂练习2-6\n",
    "\n",
    "修改前述代码，使用pytorch的线性模型。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([100, 1]) torch.Size([100])\n",
      "torch.Size([100, 1]) torch.Size([100, 1])\n",
      "100 1\n",
      "100 1\n"
     ]
    }
   ],
   "source": [
    "import torch \n",
    "import numpy as np\n",
    "from sklearn import datasets\n",
    "\n",
    "X_numpy,Y_numpy = datasets.make_regression(\n",
    "        n_samples=100,\n",
    "        n_features=1,\n",
    "        noise=20,\n",
    "        random_state=12\n",
    ")\n",
    "\n",
    "# 将numpy 的数据类型转成tensor\n",
    "X = torch.from_numpy(X_numpy.astype(np.float32))\n",
    "Y = torch.from_numpy(Y_numpy.astype(np.float32))\n",
    "print(X.shape,Y.shape)\n",
    "\n",
    "# 对Y进行格式统一\n",
    "Y = Y.view(100,1)\n",
    "print(X.shape, Y.shape)\n",
    "\n",
    "n_samples, in_features = X.shape   #行、列分别作为样本个数，特征个数\n",
    "n_lables, out_features = Y.shape   #行、列为标记个数，1或多个值（类别）\n",
    "\n",
    "print(n_samples, in_features )\n",
    "print(n_lables, out_features )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Linear(in_features=1, out_features=1, bias=True)\n",
      "SGD (\n",
      "Parameter Group 0\n",
      "    dampening: 0\n",
      "    differentiable: False\n",
      "    foreach: None\n",
      "    lr: 0.1\n",
      "    maximize: False\n",
      "    momentum: 0\n",
      "    nesterov: False\n",
      "    weight_decay: 0\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "# def model(x):\n",
    "#     return m * x + b\n",
    "# 定义模型\n",
    "model = torch.nn.Linear(in_features,out_features) #构造函数的参数\n",
    "print(model)\n",
    "\n",
    "# 定义超参数（学习率、训练次数）\n",
    "learning_rate = 0.1\n",
    "n_iters = 30\n",
    "\n",
    "# 定义损失函数\n",
    "lossFunction = torch.nn.MSELoss()\n",
    "\n",
    "# 定义优化器\n",
    "optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)\n",
    "print(optimizer)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loss=2146.69873047,w=7.986851692199707,b=-1.1643909215927124\n",
      "loss=1020.88751221,w=20.057565689086914,b=-2.688673257827759\n",
      "loss=624.92114258,w=27.286291122436523,b=-3.1160788536071777\n",
      "loss=483.12850952,w=31.637418746948242,b=-3.060955047607422\n",
      "loss=431.28756714,w=34.27067565917969,b=-2.8275864124298096\n",
      "loss=411.89178467,w=35.87339782714844,b=-2.5581817626953125\n",
      "loss=404.45550537,w=36.85468292236328,b=-2.312589645385742\n",
      "loss=401.53305054,w=37.45915985107422,b=-2.1105422973632812\n",
      "loss=400.35696411,w=37.833831787109375,b=-1.9535506963729858\n",
      "loss=399.87319946,w=38.067508697509766,b=-1.8358919620513916\n",
      "loss=399.67044067,w=38.21414566040039,b=-1.7498562335968018\n",
      "loss=399.58398438,w=38.30672073364258,b=-1.6880477666854858\n",
      "loss=399.54663086,w=38.365509033203125,b=-1.644225001335144\n",
      "loss=399.53030396,w=38.403045654296875,b=-1.6134661436080933\n",
      "loss=399.52319336,w=38.42714309692383,b=-1.5920464992523193\n"
     ]
    }
   ],
   "source": [
    "#模型开始训练\n",
    "for epoch in range(n_iters):\n",
    "    # 1.正向传播\n",
    "    y_pre = model(X)  #调用的是 nn.Linear类中的forward(x)方法\n",
    "    # 2.定义损失函数\n",
    "    loss = lossFunction(Y, y_pre)\n",
    "    # 3.向后传播\n",
    "    loss.backward()\n",
    "    # 4.更新权重\n",
    "    optimizer.step()\n",
    "    # 5.清空梯度\n",
    "    optimizer.zero_grad()\n",
    "\n",
    "    if epoch % 2 == 0:\n",
    "        w, b = model.parameters()\n",
    "        print(f\"loss={loss.item():.8f},w={w[0][0].item()},b={b[0].item()}\")\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Parameter containing:\n",
       "tensor([[38.4358]], requires_grad=True)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Parameter containing:\n",
       "tensor([-1.5839], requires_grad=True)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "b"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 课堂练习2-7\n",
    "\n",
    "利用pytorch定义神经网络类"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 线性函数模型的定义和训练：\n",
    "# 1. 定义模型（包括正向传播方法forward）\n",
    "n_samples,in_features = X.shape\n",
    "n_labels, out_features = Y.shape\n",
    "\n",
    "# 搭建自己的神经网络,并创建模型对象\n",
    "class MyModel(torch.nn.Module):\n",
    "    # 初始化函数的定义，定义神经网络和参数\n",
    "    def __init__(self, in_features_len, out_features_len):\n",
    "        super(MyModel, self).__init__()\n",
    "        # 构建线性层\n",
    "        self.linear = torch.nn.Linear(\n",
    "                           in_features_len, \n",
    "                           out_features_len\n",
    "        )\n",
    "\n",
    "    # 向前传播，构建计算图\n",
    "    def forward(self, x):\n",
    "        \"\"\"重写了父类的forward函数，正向传播\"\"\"\n",
    "        out = self.linear(x)\n",
    "        return out\n",
    "\n",
    "model = MyModel(in_features,out_features)\n",
    "\n",
    "# 2.定义损失（代价）函数loss\n",
    "lossF = torch.nn.MSELoss()\n",
    "\n",
    "# 3. 定义优化器optimizer,利用其管理模型参数\n",
    "optimizer = torch.optim.SGD(model.parameters(),lr=0.1)\n",
    "print(list(model.parameters()))\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "py38",
   "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": "3.8.13"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}

一键复制编辑原始数据按行查看历史

提交于 2023-10-27 10:51 . first commit

马建仓 AI 助手

尝试更多

代码解读

代码找茬

代码优化

https://gitee.com/wang--zhiwei/AI2023.git

git@gitee.com:wang--zhiwei/AI2023.git

wang--zhiwei

AI2023

人工智能导论2023年

main

王志伟/人工智能导论2023年

简介

发行版

贡献者

近期动态

王志伟/人工智能导论2023年 .gitee-modal { width: 500px !important; }

简介

发行版

贡献者

近期动态

搜索帮助

王志伟/人工智能导论2023年