diff --git a/.ipynb_checkpoints/dnn-fdtm-checkpoint.ipynb b/.ipynb_checkpoints/dnn-fdtm-checkpoint.ipynb new file mode 100644 index 0000000..b0746f2 --- /dev/null +++ b/.ipynb_checkpoints/dnn-fdtm-checkpoint.ipynb @@ -0,0 +1,608 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Prerequisites" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "!pip install numpy matplotlib pandas sklearn" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Python 3.7.9\n" + ] + } + ], + "source": [ + "!python -V" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Data Preprocessing" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "metadata": {}, + "outputs": [], + "source": [ + "import pandas as pd\n", + "import numpy as np\n", + "np.set_printoptions(suppress=True)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "导入原始 csv 数据" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [], + "source": [ + "X = pd.read_csv('./data.csv', header=None)\n", + "X = X.to_numpy()" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "原始数据大小 100 00 00 x 300" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "(1000000, 300)" + ] + }, + "execution_count": 7, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "X.shape" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "metadata": {}, + "outputs": [], + "source": [ + "layer_1 = np.linspace(10, 250, 100, endpoint=True)\n", + "layer_2 = np.linspace(0.5, 10, 100, endpoint=True)\n", + "layer_3 = np.linspace(710, 1700, 100, endpoint=True)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "训练集一共有 100 00 00 条数据，预测 3 个输出 (实际每个模型每次只预测1个值，训练3个模型)" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "metadata": {}, + "outputs": [], + "source": [ + "y = np.zeros((X.shape[0], 3))" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": {}, + "outputs": [], + "source": [ + "i = 0\n", + "for l3 in layer_3:\n", + " for l2 in layer_2:\n", + " for l1 in layer_1:\n", + " y[i] = [l3, l2, l1]\n", + " i = i + 1" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "(1000000, 3)" + ] + }, + "execution_count": 11, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "y.shape" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Train test split" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "划分训练集，测试集" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": {}, + "outputs": [], + "source": [ + "from sklearn.model_selection import train_test_split" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "metadata": {}, + "outputs": [], + "source": [ + "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.05, random_state=0)" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "metadata": {}, + "outputs": [], + "source": [ + "import torch\n", + "from torch.utils.data import Dataset, TensorDataset, DataLoader\n", + "from torch.utils.data.dataset import random_split" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "指定当前模型预测的值" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "metadata": {}, + "outputs": [], + "source": [ + "property = 1" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "转换数据为 pyrotch 类型" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "metadata": {}, + "outputs": [], + "source": [ + "x_train_tensor = torch.from_numpy(X_train).float()\n", + "y_train_tensor = torch.from_numpy(y_train[:, property]).float()\n", + "\n", + "x_test_tensor = torch.from_numpy(X_test).float()\n", + "y_test_tensor = torch.from_numpy(y_test[:, property]).float()" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "metadata": {}, + "outputs": [], + "source": [ + "# Builds dataset with ALL data\n", + "origin_train_dataset = TensorDataset(x_train_tensor, y_train_tensor)\n", + "\n", + "# Splits randomly into train and validation datasets\n", + "train_dataset, val_dataset = random_split(origin_train_dataset, [int(x_train_tensor.shape[0] * 0.9), int(x_train_tensor.shape[0] * 0.1)])\n", + "\n", + "# Builds a loader for each dataset to perform mini-batch gradient descent\n", + "train_loader = DataLoader(dataset=train_dataset, batch_size=2000)\n", + "val_loader = DataLoader(dataset=val_dataset, batch_size=2000)\n", + "\n", + "test_dataset = TensorDataset(x_test_tensor, y_test_tensor)\n", + "test_loader = DataLoader(dataset=test_dataset, batch_size=2000)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Define the Model" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "metadata": {}, + "outputs": [], + "source": [ + "import torch.nn as nn" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "metadata": {}, + "outputs": [], + "source": [ + "class Net(nn.Module):\n", + "\n", + " def __init__(self):\n", + " super(Net, self).__init__()\n", + " self.bn1 = nn.BatchNorm1d(X.shape[1])\n", + " self.fc1 = nn.Linear(X.shape[1], 100)\n", + " self.bn2 = nn.BatchNorm1d(100)\n", + " self.fc2 = nn.Linear(100, 50)\n", + " self.fc3 = nn.Linear(50, 10)\n", + " self.fc4 = nn.Linear(10, 1)\n", + "\n", + " def forward(self, x):\n", + " x = self.bn1(x)\n", + " x = self.fc1(x)\n", + " x = torch.tanh(x)\n", + " x = self.bn2(x)\n", + " x = self.fc2(x)\n", + " x = torch.tanh(x)\n", + " x = self.fc3(x)\n", + " x = torch.relu(x)\n", + " x = self.fc4(x)\n", + " x = torch.relu(x)\n", + " return x" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Training" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "metadata": {}, + "outputs": [], + "source": [ + "import torch.optim as optim" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "如果有 GPU 优先用 GPU" + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "metadata": {}, + "outputs": [], + "source": [ + "device = 'cuda' if torch.cuda.is_available() else 'cpu'" + ] + }, + { + "cell_type": "code", + "execution_count": 22, + "metadata": {}, + "outputs": [], + "source": [ + "n_epochs = 50" + ] + }, + { + "cell_type": "code", + "execution_count": 23, + "metadata": {}, + "outputs": [], + "source": [ + "def make_train_step(model, loss_fn, optimizer):\n", + " def train_step(x, y):\n", + " model.train()\n", + " yh = model(x)\n", + " yh = torch.reshape(yh, (-1,))\n", + " loss = loss_fn(y, yh)\n", + " loss.backward()\n", + " torch.nn.utils.clip_grad_norm_(model.parameters(), 0.25)\n", + " optimizer.step()\n", + " optimizer.zero_grad()\n", + " return loss.item()\n", + " return train_step" + ] + }, + { + "cell_type": "code", + "execution_count": 24, + "metadata": {}, + "outputs": [], + "source": [ + "model = Net().to(device)\n", + "\n", + "loss_fn = nn.MSELoss(reduction='mean')\n", + "\n", + "# optimizer = optim.SGD(model.parameters(), lr=0.01)\n", + "optimizer = optim.Adam(model.parameters(), lr=0.001, weight_decay=0.001)\n", + "\n", + "train_step = make_train_step(model, loss_fn, optimizer)" + ] + }, + { + "cell_type": "code", + "execution_count": 25, + "metadata": { + "scrolled": true + }, + "outputs": [ + { + "data": { + "text/plain": [ + "Net(\n", + " (bn1): BatchNorm1d(300, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", + " (fc1): Linear(in_features=300, out_features=100, bias=True)\n", + " (bn2): BatchNorm1d(100, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", + " (fc2): Linear(in_features=100, out_features=50, bias=True)\n", + " (fc3): Linear(in_features=50, out_features=10, bias=True)\n", + " (fc4): Linear(in_features=10, out_features=1, bias=True)\n", + ")" + ] + }, + "execution_count": 25, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "model.eval()" + ] + }, + { + "cell_type": "code", + "execution_count": 51, + "metadata": { + "scrolled": true + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[1] Training loss: 3.309\t Validation loss: 0.564\n", + "[2] Training loss: 0.227\t Validation loss: 0.171\n", + "[3] Training loss: 0.170\t Validation loss: 0.160\n", + "[4] Training loss: 0.146\t Validation loss: 0.228\n", + "[5] Training loss: 0.129\t Validation loss: 0.154\n", + "[6] Training loss: 0.124\t Validation loss: 0.130\n", + "[7] Training loss: 0.121\t Validation loss: 0.112\n", + "[8] Training loss: 0.119\t Validation loss: 0.106\n", + "[9] Training loss: 0.116\t Validation loss: 0.105\n", + "[10] Training loss: 0.115\t Validation loss: 0.118\n", + "[11] Training loss: 0.113\t Validation loss: 0.104\n", + "[12] Training loss: 0.114\t Validation loss: 0.104\n", + "[13] Training loss: 0.113\t Validation loss: 0.105\n", + "[14] Training loss: 0.112\t Validation loss: 0.101\n", + "[15] Training loss: 0.113\t Validation loss: 0.109\n", + "[16] Training loss: 0.112\t Validation loss: 0.104\n", + "[17] Training loss: 0.112\t Validation loss: 0.128\n", + "[18] Training loss: 0.111\t Validation loss: 0.118\n", + "[19] Training loss: 0.112\t Validation loss: 0.110\n", + "[20] Training loss: 0.111\t Validation loss: 0.099\n" + ] + }, + { + "ename": "KeyboardInterrupt", + "evalue": "", + "output_type": "error", + "traceback": [ + "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m", + "\u001b[1;31mKeyboardInterrupt\u001b[0m Traceback (most recent call last)", + "\u001b[1;32m\u001b[0m in \u001b[0;36m\u001b[1;34m\u001b[0m\n\u001b[0;32m 7\u001b[0m \u001b[0mx_batch\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mx_batch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mto\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 8\u001b[0m \u001b[0my_batch\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0my_batch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mto\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 9\u001b[1;33m \u001b[0mloss\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtrain_step\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mx_batch\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0my_batch\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 10\u001b[0m \u001b[0mbatch_losses\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mloss\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 11\u001b[0m \u001b[0mtraining_loss\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmean\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mbatch_losses\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;32m\u001b[0m in \u001b[0;36mtrain_step\u001b[1;34m(x, y)\u001b[0m\n\u001b[0;32m 2\u001b[0m \u001b[1;32mdef\u001b[0m \u001b[0mtrain_step\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mx\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0my\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[0;32m 3\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtrain\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;32m----> 4\u001b[1;33m \u001b[0myh\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\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[0m\n\u001b[0m\u001b[0;32m 5\u001b[0m \u001b[0myh\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mreshape\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0myh\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m(\u001b[0m\u001b[1;33m-\u001b[0m\u001b[1;36m1\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[0;32m 6\u001b[0m \u001b[0mloss\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mloss_fn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0my\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0myh\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;32md:\\anaconda3\\envs\\pytorch\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[1;34m(self, *input, **kwargs)\u001b[0m\n\u001b[0;32m 725\u001b[0m \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 726\u001b[0m \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 727\u001b[1;33m \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 728\u001b[0m for hook in itertools.chain(\n\u001b[0;32m 729\u001b[0m \u001b[0m_global_forward_hooks\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalues\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;32m\u001b[0m in \u001b[0;36mforward\u001b[1;34m(self, x)\u001b[0m\n\u001b[0;32m 12\u001b[0m \u001b[1;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\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[0m\n\u001b[0;32m 13\u001b[0m \u001b[0mx\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbn1\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[0m\n\u001b[1;32m---> 14\u001b[1;33m \u001b[0mx\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfc1\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[0m\n\u001b[0m\u001b[0;32m 15\u001b[0m \u001b[0mx\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtanh\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[0m\n\u001b[0;32m 16\u001b[0m \u001b[0mx\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbn2\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[0m\n", + "\u001b[1;32md:\\anaconda3\\envs\\pytorch\\lib\\site-packages\\torch\\nn\\modules\\module.py\u001b[0m in \u001b[0;36m_call_impl\u001b[1;34m(self, *input, **kwargs)\u001b[0m\n\u001b[0;32m 725\u001b[0m \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 726\u001b[0m \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 727\u001b[1;33m \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 728\u001b[0m for hook in itertools.chain(\n\u001b[0;32m 729\u001b[0m \u001b[0m_global_forward_hooks\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalues\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;32md:\\anaconda3\\envs\\pytorch\\lib\\site-packages\\torch\\nn\\modules\\linear.py\u001b[0m in \u001b[0;36mforward\u001b[1;34m(self, input)\u001b[0m\n\u001b[0;32m 91\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 92\u001b[0m \u001b[1;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0minput\u001b[0m\u001b[1;33m:\u001b[0m \u001b[0mTensor\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m->\u001b[0m \u001b[0mTensor\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 93\u001b[1;33m \u001b[1;32mreturn\u001b[0m \u001b[0mF\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlinear\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mweight\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbias\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 94\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 95\u001b[0m \u001b[1;32mdef\u001b[0m \u001b[0mextra_repr\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m->\u001b[0m \u001b[0mstr\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;32md:\\anaconda3\\envs\\pytorch\\lib\\site-packages\\torch\\nn\\functional.py\u001b[0m in \u001b[0;36mlinear\u001b[1;34m(input, weight, bias)\u001b[0m\n\u001b[0;32m 1688\u001b[0m \u001b[1;32mif\u001b[0m \u001b[0minput\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdim\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;33m==\u001b[0m \u001b[1;36m2\u001b[0m \u001b[1;32mand\u001b[0m \u001b[0mbias\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 1689\u001b[0m \u001b[1;31m# fused op is marginally faster\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1690\u001b[1;33m \u001b[0mret\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0maddmm\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mbias\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0minput\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mweight\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mt\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[0m\u001b[0;32m 1691\u001b[0m \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 1692\u001b[0m \u001b[0moutput\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0minput\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmatmul\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mweight\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mt\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;31mKeyboardInterrupt\u001b[0m: " + ] + } + ], + "source": [ + "training_losses = []\n", + "validation_losses = []\n", + "\n", + "for epoch in range(n_epochs):\n", + " batch_losses = []\n", + " for x_batch, y_batch in train_loader:\n", + " x_batch = x_batch.to(device)\n", + " y_batch = y_batch.to(device)\n", + " loss = train_step(x_batch, y_batch)\n", + " batch_losses.append(loss)\n", + " training_loss = np.mean(batch_losses)\n", + " training_losses.append(training_loss)\n", + "\n", + " with torch.no_grad():\n", + " val_losses = []\n", + " for x_val, y_val in val_loader:\n", + " x_val = x_val.to(device)\n", + " y_val = y_val.to(device)\n", + " model.eval()\n", + " yh = model(x_val)\n", + " yh = torch.reshape(yh, (-1,))\n", + " val_loss = loss_fn(y_val, yh).item()\n", + " val_losses.append(val_loss)\n", + " validation_loss = np.mean(val_losses)\n", + " validation_losses.append(validation_loss)\n", + "\n", + " print(f\"[{epoch+1}] Training loss: {training_loss:.3f}\\t Validation loss: {validation_loss:.3f}\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Testing" + ] + }, + { + "cell_type": "code", + "execution_count": 53, + "metadata": {}, + "outputs": [], + "source": [ + "def mean_absolute_percentage_error(y_true, y_pred):\n", + " return torch.mean(torch.abs((y_true - y_pred) / y_true)) * 100" + ] + }, + { + "cell_type": "code", + "execution_count": 54, + "metadata": {}, + "outputs": [], + "source": [ + "x_test_tensor = x_test_tensor.to(device)\n", + "y_test_tensor = y_test_tensor.to(device)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "用训练好的模型进行预测" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "y_pred = model(x_test_tensor).squeeze()" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "计算 loss function (MSE)" + ] + }, + { + "cell_type": "code", + "execution_count": 55, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "tensor(0.0992, grad_fn=)\n" + ] + } + ], + "source": [ + "test_loss = loss_fn(y_test_tensor, y_pred)\n", + "print(test_loss)" + ] + }, + { + "cell_type": "code", + "execution_count": 56, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "The mean of absolute percentage error for C: 3.82%\n" + ] + } + ], + "source": [ + "print(f\"The mean of absolute percentage error for C: {mean_absolute_percentage_error(y_test_tensor.cpu(), y_pred.cpu()):.2f}%\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "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": "3.7.9" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/.ipynb_checkpoints/dnn-fdtm_ree-checkpoint.ipynb b/.ipynb_checkpoints/dnn-fdtm_ree-checkpoint.ipynb new file mode 100644 index 0000000..bf3d501 --- /dev/null +++ b/.ipynb_checkpoints/dnn-fdtm_ree-checkpoint.ipynb @@ -0,0 +1,1035 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Prerequisites" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "!pip install numpy matplotlib pandas sklearn" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "!python -V" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Data Preprocessing" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "metadata": {}, + "outputs": [], + "source": [ + "import pickle\n", + "import numpy as np" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "metadata": {}, + "outputs": [], + "source": [ + "np.set_printoptions(suppress=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "X = pickle.load( open( \"res.pickle\", \"rb\" ) )" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": {}, + "outputs": [], + "source": [ + "X = X.T" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "(1000000, 493)" + ] + }, + "execution_count": 12, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "X.shape" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "metadata": {}, + "outputs": [], + "source": [ + "layer_1 = np.linspace(10, 250, 100, endpoint=True)\n", + "layer_2 = np.linspace(0.5, 10, 100, endpoint=True)\n", + "layer_3 = np.linspace(710, 1700, 100, endpoint=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "metadata": {}, + "outputs": [], + "source": [ + "y = np.zeros((X.shape[0], 3))" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "metadata": {}, + "outputs": [], + "source": [ + "i = 0\n", + "for l3 in layer_3:\n", + " for l2 in layer_2:\n", + " for l1 in layer_1:\n", + " y[i] = [l3, l2, l1]\n", + " i = i + 1" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "(1000000, 3)" + ] + }, + "execution_count": 16, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "y.shape" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## HDF5" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "Write to HDF5" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "metadata": {}, + "outputs": [], + "source": [ + "import pandas as pd" + ] + }, + { + "cell_type": "code", + "execution_count": 25, + "metadata": {}, + "outputs": [], + "source": [ + "df_X = pd.DataFrame(X)\n", + "df_y = pd.DataFrame(y)" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "INFO:MainThread:numexpr.utils:NumExpr defaulting to 8 threads.\n" + ] + } + ], + "source": [ + "store = pd.HDFStore('res.h5')" + ] + }, + { + "cell_type": "code", + "execution_count": 26, + "metadata": {}, + "outputs": [], + "source": [ + "store['X'] = df_X\n", + "store['y'] = df_y" + ] + }, + { + "cell_type": "code", + "execution_count": 22, + "metadata": {}, + "outputs": [], + "source": [ + "store.close()" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "Read from HDF5" + ] + }, + { + "cell_type": "code", + "execution_count": 23, + "metadata": {}, + "outputs": [], + "source": [ + "res = pd.read_hdf('res.h5', 'X')" + ] + }, + { + "cell_type": "code", + "execution_count": 37, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "(1000000, 3)" + ] + }, + "execution_count": 37, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "res.shape" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "Vaex" + ] + }, + { + "cell_type": "code", + "execution_count": 26, + "metadata": {}, + "outputs": [], + "source": [ + "import vaex" + ] + }, + { + "cell_type": "code", + "execution_count": 45, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Python 3.8.3\n" + ] + } + ], + "source": [ + "!python -V" + ] + }, + { + "cell_type": "code", + "execution_count": 51, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "array([ 710., 710., 710., ..., 1700., 1700., 1700.])" + ] + }, + "execution_count": 51, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "y[:, 0]" + ] + }, + { + "cell_type": "code", + "execution_count": 59, + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "vaex_df = vaex.from_arrays(x=X, y=y)" + ] + }, + { + "cell_type": "code", + "execution_count": 63, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "(1000000, 493)" + ] + }, + "execution_count": 63, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "X.shape" + ] + }, + { + "cell_type": "code", + "execution_count": 64, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "(1000000, 3)" + ] + }, + "execution_count": 64, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "y.shape" + ] + }, + { + "cell_type": "code", + "execution_count": 62, + "metadata": {}, + "outputs": [ + { + "data": { + "text/html": [ + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "
# x y
0 '[0.93921708 1. 0.96758555 0.88402555 0....[710. 0.5 10. ]
1 '[0.9392044 1. 0.96759943 0.88406149 0....[710. 0.5 12.42424242]
2 '[0.93918573 1. 0.9676289 0.88417059 0....[710. 0.5 14.84848485]
3 '[0.93915796 1. 0.96770218 0.88444087 0....[710. 0.5 17.27272727]
4 '[0.93910895 1. 0.96785629 0.88496242 0....[710. 0.5 19.6969697]
... ... ...
999,995'[0.90888059 1. 0.99898726 0.9374508 0....[1700. 10. 240.3030303]
999,996'[0.90857156 1. 0.99928981 0.93796907 0....[1700. 10. 242.72727273]
999,997'[0.90826533 1. 0.99958941 0.93848211 0....[1700. 10. 245.15151515]
999,998'[0.90796191 1. 0.99988598 0.93898997 0....[1700. 10. 247.57575758]
999,999'[0.90749824 0.99982035 1. 0.93932393 0....[1700. 10. 250.]
" + ], + "text/plain": [ + "# x y\n", + "0 '[0.93921708 1. 0.96758555 0.88402555 0.... [710. 0.5 10. ]\n", + "1 '[0.9392044 1. 0.96759943 0.88406149 0.... [710. 0.5 12.42424242]\n", + "2 '[0.93918573 1. 0.9676289 0.88417059 0.... [710. 0.5 14.84848485]\n", + "3 '[0.93915796 1. 0.96770218 0.88444087 0.... [710. 0.5 17.27272727]\n", + "4 '[0.93910895 1. 0.96785629 0.88496242 0.... [710. 0.5 19.6969697]\n", + "... ... ...\n", + "999,995 '[0.90888059 1. 0.99898726 0.9374508 0.... [1700. 10. 240.3030303]\n", + "999,996 '[0.90857156 1. 0.99928981 0.93796907 0.... [1700. 10. 242.72727273]\n", + "999,997 '[0.90826533 1. 0.99958941 0.93848211 0.... [1700. 10. 245.15151515]\n", + "999,998 '[0.90796191 1. 0.99988598 0.93898997 0.... [1700. 10. 247.57575758]\n", + "999,999 '[0.90749824 0.99982035 1. 0.93932393 0.... [1700. 10. 250.]" + ] + }, + "execution_count": 62, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "vaex_df" + ] + }, + { + "cell_type": "code", + "execution_count": 61, + "metadata": { + "collapsed": true + }, + "outputs": [ + { + "ename": "ValueError", + "evalue": "could not broadcast input array from shape (100000,493) into shape (100000)", + "output_type": "error", + "traceback": [ + "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m", + "\u001b[1;31mValueError\u001b[0m Traceback (most recent call last)", + "\u001b[1;32m\u001b[0m in \u001b[0;36m\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mvaex_df\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexport_hdf5\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'res.hdf5'\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mprogress\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m", + "\u001b[1;32mD:\\Anaconda3\\lib\\site-packages\\vaex\\dataframe.py\u001b[0m in \u001b[0;36mexport_hdf5\u001b[1;34m(self, path, column_names, byteorder, shuffle, selection, progress, virtual, sort, ascending)\u001b[0m\n\u001b[0;32m 5802\u001b[0m \"\"\"\n\u001b[0;32m 5803\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mvaex\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexport\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 5804\u001b[1;33m \u001b[0mvaex\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexport\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexport_hdf5\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mpath\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcolumn_names\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mbyteorder\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mshuffle\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mselection\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mprogress\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mprogress\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvirtual\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mvirtual\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0msort\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0msort\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mascending\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mascending\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 5805\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 5806\u001b[0m \u001b[1;32mdef\u001b[0m \u001b[0mexport_fits\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mpath\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcolumn_names\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mNone\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mshuffle\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mselection\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mprogress\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mNone\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvirtual\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0msort\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mNone\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mascending\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\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;32mD:\\Anaconda3\\lib\\site-packages\\vaex\\export.py\u001b[0m in \u001b[0;36mexport_hdf5\u001b[1;34m(dataset, path, column_names, byteorder, shuffle, selection, progress, virtual, sort, ascending)\u001b[0m\n\u001b[0;32m 310\u001b[0m \u001b[0mkwargs\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mlocals\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[0;32m 311\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mvaex\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mhdf5\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexport\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 312\u001b[1;33m \u001b[0mvaex\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mhdf5\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexport\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexport_hdf5\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 313\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 314\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;32mD:\\Anaconda3\\lib\\site-packages\\vaex\\hdf5\\export.py\u001b[0m in \u001b[0;36mexport_hdf5\u001b[1;34m(dataset, path, column_names, byteorder, shuffle, selection, progress, virtual, sort, ascending)\u001b[0m\n\u001b[0;32m 243\u001b[0m \u001b[0mdataset_output\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mvaex\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mhdf5\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mHdf5MemoryMapped\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mwrite\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 244\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 245\u001b[1;33m column_names = vaex.export._export(dataset_input=dataset, dataset_output=dataset_output, path=path, random_index_column=random_index_name,\n\u001b[0m\u001b[0;32m 246\u001b[0m \u001b[0mcolumn_names\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mcolumn_names\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mselection\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mselection\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mshuffle\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mshuffle\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mbyteorder\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mbyteorder\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 247\u001b[0m progress=progress, sort=sort, ascending=ascending)\n", + "\u001b[1;32mD:\\Anaconda3\\lib\\site-packages\\vaex\\export.py\u001b[0m in \u001b[0;36m_export\u001b[1;34m(dataset_input, dataset_output, random_index_column, path, column_names, byteorder, shuffle, selection, progress, virtual, sort, ascending)\u001b[0m\n\u001b[0;32m 139\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0mfuture\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mfutures\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 140\u001b[0m \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 141\u001b[1;33m \u001b[0mfuture\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mresult\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m0.1\u001b[0m\u001b[1;33m/\u001b[0m\u001b[1;36m4\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 142\u001b[0m \u001b[1;32mexcept\u001b[0m \u001b[0mconcurrent\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfutures\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mTimeoutError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 143\u001b[0m \u001b[0mdone\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mFalse\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;32mD:\\Anaconda3\\lib\\concurrent\\futures\\_base.py\u001b[0m in \u001b[0;36mresult\u001b[1;34m(self, timeout)\u001b[0m\n\u001b[0;32m 430\u001b[0m \u001b[1;32mraise\u001b[0m \u001b[0mCancelledError\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[0;32m 431\u001b[0m \u001b[1;32melif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_state\u001b[0m \u001b[1;33m==\u001b[0m \u001b[0mFINISHED\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 432\u001b[1;33m \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m__get_result\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[0m\u001b[0;32m 433\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 434\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_condition\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mwait\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtimeout\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;32mD:\\Anaconda3\\lib\\concurrent\\futures\\_base.py\u001b[0m in \u001b[0;36m__get_result\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m 386\u001b[0m \u001b[1;32mdef\u001b[0m \u001b[0m__get_result\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\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[0;32m 387\u001b[0m \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_exception\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 388\u001b[1;33m \u001b[1;32mraise\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_exception\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 389\u001b[0m \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 390\u001b[0m \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_result\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;32mD:\\Anaconda3\\lib\\concurrent\\futures\\thread.py\u001b[0m in \u001b[0;36mrun\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m 55\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 56\u001b[0m \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 57\u001b[1;33m \u001b[0mresult\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m**\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 58\u001b[0m \u001b[1;32mexcept\u001b[0m \u001b[0mBaseException\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0mexc\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 59\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfuture\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mset_exception\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mexc\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;32mD:\\Anaconda3\\lib\\site-packages\\vaex\\export.py\u001b[0m in \u001b[0;36m_export_column\u001b[1;34m(dataset_input, dataset_output, column_name, shuffle, sort, selection, N, order_array, order_array_inverse, progress_status)\u001b[0m\n\u001b[0;32m 211\u001b[0m \u001b[0mto_array\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mtarget_set_item\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mvalues\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mfilled\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfill_value\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 212\u001b[0m \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 213\u001b[1;33m \u001b[0mto_array\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mtarget_set_item\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mvalues\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m 214\u001b[0m \u001b[0mto_offset\u001b[0m \u001b[1;33m+=\u001b[0m \u001b[0mno_values\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m 215\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n", + "\u001b[1;31mValueError\u001b[0m: could not broadcast input array from shape (100000,493) into shape (100000)" + ] + } + ], + "source": [ + "vaex_df.export_hdf5(path='res.hdf5', progress=False)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "### Train test split" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": {}, + "outputs": [], + "source": [ + "from sklearn.model_selection import train_test_split" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "metadata": {}, + "outputs": [], + "source": [ + "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.05, random_state=0)" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "metadata": {}, + "outputs": [], + "source": [ + "import torch\n", + "from torch.utils.data import Dataset, TensorDataset, DataLoader\n", + "from torch.utils.data.dataset import random_split" + ] + }, + { + "cell_type": "code", + "execution_count": 55, + "metadata": {}, + "outputs": [], + "source": [ + "property = 2" + ] + }, + { + "cell_type": "code", + "execution_count": 56, + "metadata": {}, + "outputs": [], + "source": [ + "x_train_tensor = torch.from_numpy(X_train).float()\n", + "y_train_tensor = torch.from_numpy(y_train[:, property]).float()\n", + "\n", + "x_test_tensor = torch.from_numpy(X_test).float()\n", + "y_test_tensor = torch.from_numpy(y_test[:, property]).float()" + ] + }, + { + "cell_type": "code", + "execution_count": 57, + "metadata": {}, + "outputs": [], + "source": [ + "# Builds dataset with ALL data\n", + "origin_train_dataset = TensorDataset(x_train_tensor, y_train_tensor)\n", + "\n", + "# Splits randomly into train and validation datasets\n", + "train_dataset, val_dataset = random_split(origin_train_dataset, [int(x_train_tensor.shape[0] * 0.9), int(x_train_tensor.shape[0] * 0.1)])\n", + "\n", + "# Builds a loader for each dataset to perform mini-batch gradient descent\n", + "train_loader = DataLoader(dataset=train_dataset, batch_size=2000)\n", + "val_loader = DataLoader(dataset=val_dataset, batch_size=2000)\n", + "\n", + "test_dataset = TensorDataset(x_test_tensor, y_test_tensor)\n", + "test_loader = DataLoader(dataset=test_dataset, batch_size=2000)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Model" + ] + }, + { + "cell_type": "code", + "execution_count": 58, + "metadata": {}, + "outputs": [], + "source": [ + "import torch.nn as nn" + ] + }, + { + "cell_type": "code", + "execution_count": 59, + "metadata": {}, + "outputs": [], + "source": [ + "class Net(nn.Module):\n", + "\n", + " def __init__(self):\n", + " super(Net, self).__init__()\n", + " self.bn1 = nn.BatchNorm1d(X.shape[1])\n", + " self.fc1 = nn.Linear(X.shape[1], 100)\n", + " self.bn2 = nn.BatchNorm1d(100)\n", + " self.fc2 = nn.Linear(100, 50)\n", + " self.fc3 = nn.Linear(50, 10)\n", + " self.fc4 = nn.Linear(10, 1)\n", + "\n", + " def forward(self, x):\n", + " x = self.bn1(x)\n", + " x = self.fc1(x)\n", + " x = torch.tanh(x)\n", + " x = self.bn2(x)\n", + " x = self.fc2(x)\n", + " x = torch.tanh(x)\n", + " x = self.fc3(x)\n", + " x = torch.relu(x)\n", + " x = self.fc4(x)\n", + " x = torch.relu(x)\n", + " return x" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Training" + ] + }, + { + "cell_type": "code", + "execution_count": 60, + "metadata": {}, + "outputs": [], + "source": [ + "import torch.optim as optim" + ] + }, + { + "cell_type": "code", + "execution_count": 61, + "metadata": {}, + "outputs": [], + "source": [ + "device = 'cuda' if torch.cuda.is_available() else 'cpu'" + ] + }, + { + "cell_type": "code", + "execution_count": 62, + "metadata": {}, + "outputs": [], + "source": [ + "n_epochs = 20" + ] + }, + { + "cell_type": "code", + "execution_count": 63, + "metadata": {}, + "outputs": [], + "source": [ + "def make_train_step(model, loss_fn, optimizer):\n", + " def train_step(x, y):\n", + " model.train()\n", + " yh = model(x)\n", + " yh = torch.reshape(yh, (-1,))\n", + " loss = loss_fn(y, yh)\n", + " loss.backward()\n", + " torch.nn.utils.clip_grad_norm_(model.parameters(), 0.25)\n", + " optimizer.step()\n", + " optimizer.zero_grad()\n", + " return loss.item()\n", + " return train_step" + ] + }, + { + "cell_type": "code", + "execution_count": 64, + "metadata": {}, + "outputs": [], + "source": [ + "model = Net().to(device)\n", + "\n", + "loss_fn = nn.MSELoss(reduction='mean')\n", + "\n", + "# optimizer = optim.SGD(model.parameters(), lr=0.01)\n", + "optimizer = optim.Adam(model.parameters(), lr=0.001, weight_decay=0.001)\n", + "\n", + "train_step = make_train_step(model, loss_fn, optimizer)" + ] + }, + { + "cell_type": "code", + "execution_count": 65, + "metadata": { + "scrolled": true + }, + "outputs": [ + { + "data": { + "text/plain": [ + "Net(\n", + " (bn1): BatchNorm1d(493, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", + " (fc1): Linear(in_features=493, out_features=100, bias=True)\n", + " (bn2): BatchNorm1d(100, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n", + " (fc2): Linear(in_features=100, out_features=50, bias=True)\n", + " (fc3): Linear(in_features=50, out_features=10, bias=True)\n", + " (fc4): Linear(in_features=10, out_features=1, bias=True)\n", + ")" + ] + }, + "execution_count": 65, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "model.eval()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "scrolled": true + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[1] Training loss: 17419.833\t Validation loss: 11374.671\n", + "[2] Training loss: 5437.492\t Validation loss: 981.141\n", + "[3] Training loss: 170.984\t Validation loss: 17.300\n", + "[4] Training loss: 16.513\t Validation loss: 14.237\n", + "[5] Training loss: 14.340\t Validation loss: 9.781\n", + "[6] Training loss: 12.084\t Validation loss: 9.245\n", + "[7] Training loss: 10.989\t Validation loss: 9.115\n", + "[8] Training loss: 10.344\t Validation loss: 9.222\n", + "[9] Training loss: 9.963\t Validation loss: 7.374\n", + "[10] Training loss: 9.699\t Validation loss: 9.329\n", + "[11] Training loss: 9.408\t Validation loss: 9.093\n", + "[12] Training loss: 9.157\t Validation loss: 10.942\n", + "[13] Training loss: 9.157\t Validation loss: 9.646\n", + "[14] Training loss: 9.015\t Validation loss: 8.723\n", + "[15] Training loss: 8.873\t Validation loss: 8.430\n", + "[16] Training loss: 8.702\t Validation loss: 8.981\n", + "[17] Training loss: 8.599\t Validation loss: 9.332\n" + ] + } + ], + "source": [ + "training_losses = []\n", + "validation_losses = []\n", + "\n", + "for epoch in range(n_epochs):\n", + " batch_losses = []\n", + " for x_batch, y_batch in train_loader:\n", + " x_batch = x_batch.to(device)\n", + " y_batch = y_batch.to(device)\n", + " loss = train_step(x_batch, y_batch)\n", + " batch_losses.append(loss)\n", + " training_loss = np.mean(batch_losses)\n", + " training_losses.append(training_loss)\n", + "\n", + " with torch.no_grad():\n", + " val_losses = []\n", + " for x_val, y_val in val_loader:\n", + " x_val = x_val.to(device)\n", + " y_val = y_val.to(device)\n", + " model.eval()\n", + " yh = model(x_val)\n", + " yh = torch.reshape(yh, (-1,))\n", + " val_loss = loss_fn(y_val, yh).item()\n", + " val_losses.append(val_loss)\n", + " validation_loss = np.mean(val_losses)\n", + " validation_losses.append(validation_loss)\n", + "\n", + " print(f\"[{epoch+1}] Training loss: {training_loss:.3f}\\t Validation loss: {validation_loss:.3f}\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "# model.state_dict()" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Testing" + ] + }, + { + "cell_type": "code", + "execution_count": 26, + "metadata": {}, + "outputs": [], + "source": [ + "def mean_absolute_percentage_error(y_true, y_pred):\n", + " return torch.mean(torch.abs((y_true - y_pred) / y_true)) * 100" + ] + }, + { + "cell_type": "code", + "execution_count": 27, + "metadata": {}, + "outputs": [], + "source": [ + "x_test_tensor = x_test_tensor.to(device)\n", + "y_test_tensor = y_test_tensor.to(device)\n", + "y_pred = model(x_test_tensor).squeeze()" + ] + }, + { + "cell_type": "code", + "execution_count": 28, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "tensor(3279.8179, grad_fn=)\n" + ] + } + ], + "source": [ + "test_loss = loss_fn(y_test_tensor, y_pred)\n", + "print(test_loss)" + ] + }, + { + "cell_type": "code", + "execution_count": 92, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "The mean of absolute percentage error for L1: 2.51%\n" + ] + } + ], + "source": [ + "print(f\"The mean of absolute percentage error for L1: {mean_absolute_percentage_error(y_test_tensor.cpu(), y_pred.cpu()):.2f}%\")" + ] + }, + { + "cell_type": "code", + "execution_count": 31, + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "The mean of absolute percentage error for L2: 3.12%\n" + ] + } + ], + "source": [ + "print(f\"The mean of absolute percentage error for L2: {mean_absolute_percentage_error(y_test_tensor.cpu(), y_pred.cpu()):.2f}%\")" + ] + }, + { + "cell_type": "code", + "execution_count": 29, + "metadata": { + "scrolled": true + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "The mean of absolute percentage error for L3: 3.09%\n" + ] + } + ], + "source": [ + "print(f\"The mean of absolute percentage error for L3: {mean_absolute_percentage_error(y_test_tensor.cpu(), y_pred.cpu()):.2f}%\")" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "## Real-world case" + ] + }, + { + "cell_type": "code", + "execution_count": 31, + "metadata": {}, + "outputs": [], + "source": [ + "import pandas as pd" + ] + }, + { + "cell_type": "code", + "execution_count": 44, + "metadata": {}, + "outputs": [], + "source": [ + "case1 = pd.read_csv('case1.txt', header=None)" + ] + }, + { + "cell_type": "code", + "execution_count": 45, + "metadata": {}, + "outputs": [], + "source": [ + "case1 = np.reshape(np.array(case1), [1, -1])" + ] + }, + { + "cell_type": "code", + "execution_count": 46, + "metadata": {}, + "outputs": [], + "source": [ + "case1_tensor = torch.from_numpy(case1).float()" + ] + }, + { + "cell_type": "code", + "execution_count": 94, + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "model.eval()\n", + "pred_1 = model(case1_tensor)" + ] + }, + { + "cell_type": "code", + "execution_count": 95, + "metadata": { + "scrolled": false + }, + "outputs": [ + { + "data": { + "text/plain": [ + "tensor([[150.0873]], grad_fn=)" + ] + }, + "execution_count": 95, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "pred_1" + ] + }, + { + "cell_type": "code", + "execution_count": 53, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "tensor([[1.1310]], grad_fn=)" + ] + }, + "execution_count": 53, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "pred_1" + ] + }, + { + "cell_type": "code", + "execution_count": 75, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "tensor([[1472.1879]], grad_fn=)" + ] + }, + "execution_count": 75, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "pred_1" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "case1 = pd.read_csv('case2.txt', header=None)\n", + "case1 = np.reshape(np.array(case1), [1, -1])\n", + "case1_tensor = torch.from_numpy(case1).float()\n", + "model.eval()\n", + "pred_1 = model(case1_tensor)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "pred_1" + ] + }, + { + "cell_type": "code", + "execution_count": 54, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "tensor([[2.4505]], grad_fn=)" + ] + }, + "execution_count": 54, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "pred_1" + ] + }, + { + "cell_type": "code", + "execution_count": 35, + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "tensor([[1467.8439]], grad_fn=)" + ] + }, + "execution_count": 35, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "pred_1" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3", + "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": "3.7.9" + } + }, + "nbformat": 4, + "nbformat_minor": 4 +} diff --git a/README.md b/README.md index fb8ae15..cf0a0a3 100644 --- a/README.md +++ b/README.md @@ -2,4 +2,26 @@ The first try: data.csv (2.35GB) + dnn-fdtm.ipynb -The second try: ree1217.mat --> res.pickle --> res.h5 + dnn-fdtm_ree.ipynb \ No newline at end of file +The second try: ree1217.mat --> res.pickle --> res.h5 + dnn-fdtm_ree.ipynb + + + +#### Prerequisites + +- python 3.8 or anaconda + +在当前项目路径打开 cmd，使用 jupyter notebook 命令打开源码： + +``` +$ jupyter notebook +``` + +![](docs/jupyter.png) + +上面的命令会自动弹出，在浏览器里打开项目路径，如果有缺失的软件包，可以先用 pip install 安装。 + +![](docs/code.png) + + + +快捷键：使用 shift + enter 执行代码块，或者菜单栏也有按钮可以执行。 \ No newline at end of file diff --git a/dnn-fdtm.ipynb b/dnn-fdtm.ipynb index 59c7389..b0746f2 100644 --- a/dnn-fdtm.ipynb +++ b/dnn-fdtm.ipynb @@ -20,9 +20,17 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 1, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Python 3.7.9\n" + ] + } + ], "source": [ "!python -V" ] @@ -34,46 +42,44 @@ "## Data Preprocessing" ] }, - { - "cell_type": "code", - "execution_count": 1, - "metadata": {}, - "outputs": [], - "source": [ - "import pandas as pd\n", - "import numpy as np" - ] - }, - { - "cell_type": "code", - "execution_count": 2, - "metadata": {}, - "outputs": [], - "source": [ - "np.set_printoptions(suppress=True)" - ] - }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [], "source": [ - "X = pd.read_csv('./data.csv', header=None)" + "import pandas as pd\n", + "import numpy as np\n", + "np.set_printoptions(suppress=True)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "导入原始 csv 数据" ] }, { "cell_type": "code", - "execution_count": 4, + "execution_count": 6, "metadata": {}, "outputs": [], "source": [ + "X = pd.read_csv('./data.csv', header=None)\n", "X = X.to_numpy()" ] }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "原始数据大小 100 00 00 x 300" + ] + }, { "cell_type": "code", - "execution_count": 5, + "execution_count": 7, "metadata": {}, "outputs": [ { @@ -82,7 +88,7 @@ "(1000000, 300)" ] }, - "execution_count": 5, + "execution_count": 7, "metadata": {}, "output_type": "execute_result" } @@ -93,7 +99,7 @@ }, { "cell_type": "code", - "execution_count": 6, + "execution_count": 8, "metadata": {}, "outputs": [], "source": [ @@ -102,9 +108,16 @@ "layer_3 = np.linspace(710, 1700, 100, endpoint=True)" ] }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "训练集一共有 100 00 00 条数据，预测 3 个输出 (实际每个模型每次只预测1个值，训练3个模型)" + ] + }, { "cell_type": "code", - "execution_count": 7, + "execution_count": 9, "metadata": {}, "outputs": [], "source": [ @@ -113,7 +126,7 @@ }, { "cell_type": "code", - "execution_count": 8, + "execution_count": 10, "metadata": {}, "outputs": [], "source": [ @@ -127,7 +140,7 @@ }, { "cell_type": "code", - "execution_count": 9, + "execution_count": 11, "metadata": {}, "outputs": [ { @@ -136,7 +149,7 @@ "(1000000, 3)" ] }, - "execution_count": 9, + "execution_count": 11, "metadata": {}, "output_type": "execute_result" } @@ -152,9 +165,16 @@ "### Train test split" ] }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "划分训练集，测试集" + ] + }, { "cell_type": "code", - "execution_count": 35, + "execution_count": 12, "metadata": {}, "outputs": [], "source": [ @@ -163,7 +183,7 @@ }, { "cell_type": "code", - "execution_count": 36, + "execution_count": 13, "metadata": {}, "outputs": [], "source": [ @@ -172,7 +192,7 @@ }, { "cell_type": "code", - "execution_count": 37, + "execution_count": 14, "metadata": {}, "outputs": [], "source": [ @@ -181,18 +201,32 @@ "from torch.utils.data.dataset import random_split" ] }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "指定当前模型预测的值" + ] + }, { "cell_type": "code", - "execution_count": 38, + "execution_count": 15, "metadata": {}, "outputs": [], "source": [ "property = 1" ] }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "转换数据为 pyrotch 类型" + ] + }, { "cell_type": "code", - "execution_count": 39, + "execution_count": 16, "metadata": {}, "outputs": [], "source": [ @@ -205,7 +239,7 @@ }, { "cell_type": "code", - "execution_count": 40, + "execution_count": 17, "metadata": {}, "outputs": [], "source": [ @@ -227,12 +261,12 @@ "cell_type": "markdown", "metadata": {}, "source": [ - "## Model" + "## Define the Model" ] }, { "cell_type": "code", - "execution_count": 41, + "execution_count": 18, "metadata": {}, "outputs": [], "source": [ @@ -241,7 +275,7 @@ }, { "cell_type": "code", - "execution_count": 42, + "execution_count": 19, "metadata": {}, "outputs": [], "source": [ @@ -279,16 +313,23 @@ }, { "cell_type": "code", - "execution_count": 43, + "execution_count": 20, "metadata": {}, "outputs": [], "source": [ "import torch.optim as optim" ] }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "如果有 GPU 优先用 GPU" + ] + }, { "cell_type": "code", - "execution_count": 44, + "execution_count": 21, "metadata": {}, "outputs": [], "source": [ @@ -297,7 +338,7 @@ }, { "cell_type": "code", - "execution_count": 47, + "execution_count": 22, "metadata": {}, "outputs": [], "source": [ @@ -306,7 +347,7 @@ }, { "cell_type": "code", - "execution_count": 48, + "execution_count": 23, "metadata": {}, "outputs": [], "source": [ @@ -326,7 +367,7 @@ }, { "cell_type": "code", - "execution_count": 49, + "execution_count": 24, "metadata": {}, "outputs": [], "source": [ @@ -342,7 +383,7 @@ }, { "cell_type": "code", - "execution_count": 50, + "execution_count": 25, "metadata": { "scrolled": true }, @@ -360,7 +401,7 @@ ")" ] }, - "execution_count": 50, + "execution_count": 25, "metadata": {}, "output_type": "execute_result" } @@ -450,15 +491,6 @@ " print(f\"[{epoch+1}] Training loss: {training_loss:.3f}\\t Validation loss: {validation_loss:.3f}\")" ] }, - { - "cell_type": "code", - "execution_count": 52, - "metadata": {}, - "outputs": [], - "source": [ - "# model.state_dict()" - ] - }, { "cell_type": "markdown", "metadata": {}, @@ -483,10 +515,32 @@ "outputs": [], "source": [ "x_test_tensor = x_test_tensor.to(device)\n", - "y_test_tensor = y_test_tensor.to(device)\n", + "y_test_tensor = y_test_tensor.to(device)" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "用训练好的模型进行预测" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ "y_pred = model(x_test_tensor).squeeze()" ] }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "计算 loss function (MSE)" + ] + }, { "cell_type": "code", "execution_count": 55, @@ -546,7 +600,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.8.5" + "version": "3.7.9" } }, "nbformat": 4, diff --git a/dnn-fdtm_ree.ipynb b/dnn-fdtm_ree.ipynb index b5ad5b5..bf3d501 100644 --- a/dnn-fdtm_ree.ipynb +++ b/dnn-fdtm_ree.ipynb @@ -410,7 +410,9 @@ { "cell_type": "code", "execution_count": 61, - "metadata": {}, + "metadata": { + "collapsed": true + }, "outputs": [ { "ename": "ValueError", @@ -1025,7 +1027,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.8.3" + "version": "3.7.9" } }, "nbformat": 4,