{ "cells": [ { "cell_type": "code", "execution_count": 61, "metadata": {}, "outputs": [], "source": [ "import pandas as pd\n", "import numpy as np\n", "import torch\n", "from nltk.tokenize import word_tokenize\n", "import gensim.downloader as api" ] }, { "cell_type": "code", "execution_count": 62, "metadata": {}, "outputs": [], "source": [ "# Wczytanie X i Y do Train oraz X do Dev i Test\n", "X_train = pd.read_table('train/in.tsv', sep='\\t', error_bad_lines=False, quoting=3, header=None, names=['content', 'id'], usecols=['content'])\n", "y_train = pd.read_table('train/expected.tsv', sep='\\t', error_bad_lines=False, quoting=3, header=None, names=['label'])\n", "X_dev = pd.read_table('dev-0/in.tsv', sep='\\t', error_bad_lines=False, header=None, quoting=3, names=['content', 'id'], usecols=['content'])\n", "X_test = pd.read_table('test-A/in.tsv', sep='\\t', error_bad_lines=False, header=None, quoting=3, names=['content', 'id'], usecols=['content'])" ] }, { "cell_type": "code", "execution_count": 63, "metadata": {}, "outputs": [], "source": [ "# lowercase-ing zbiorów\n", "# https://www.datacamp.com/community/tutorials/case-conversion-python\n", "X_train = X_train.content.str.lower()\n", "X_dev = X_dev.content.str.lower()\n", "X_test = X_test.content.str.lower()\n", "\n", "y_train = y_train['label'] #Df do Series?" ] }, { "cell_type": "code", "execution_count": 64, "metadata": {}, "outputs": [], "source": [ "# tokenizacja zbiorów\n", "#https://www.nltk.org/_modules/nltk/tokenize.html\n", "X_train = [word_tokenize(doc) for doc in X_train]\n", "X_dev = [word_tokenize(doc) for doc in X_dev]\n", "X_test = [word_tokenize(doc) for doc in X_test]" ] }, { "cell_type": "code", "execution_count": 67, "metadata": {}, "outputs": [], "source": [ "# word2vec zgodnie z poradą Pana Jakuba\n", "# https://radimrehurek.com/gensim/auto_examples/howtos/run_downloader_api.html\n", "# https://www.kaggle.com/kstathou/word-embeddings-logistic-regression\n", "w2v = api.load('word2vec-google-news-300')\n", "\n", "def document_vector(doc):\n", " \"\"\"Create document vectors by averaging word vectors. Remove out-of-vocabulary words.\"\"\"\n", " return np.mean([w2v[w] for w in doc if w in w2v] or [np.zeros(300)], axis=0)\n", "\n", "X_train = [document_vector(doc) for doc in X_train]\n", "X_dev = [document_vector(doc) for doc in X_dev]\n", "X_test = [document_vector(doc) for doc in X_test]" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "#Sieć neuronowa z ćwiczeń 8\n", "#https://git.wmi.amu.edu.pl/filipg/aitech-eks-pub/src/branch/master/cw/08_regresja_logistyczna.ipynb\n", "class NeuralNetwork(torch.nn.Module): \n", " def __init__(self, hidden_size):\n", " super(NeuralNetwork, self).__init__()\n", " self.l1 = torch.nn.Linear(300, hidden_size) #Korzystamy z word2vec-google-news-300 który ma zawsze na wejściu wymiar 300\n", " self.l2 = torch.nn.Linear(hidden_size, 1)\n", "\n", " def forward(self, x):\n", " x = self.l1(x)\n", " x = torch.relu(x)\n", " x = self.l2(x)\n", " x = torch.sigmoid(x)\n", " return x" ] }, { "cell_type": "code", "execution_count": 45, "metadata": {}, "outputs": [], "source": [ "model = NeuralNetwork(600)\n", "\n", "criterion = torch.nn.BCELoss()\n", "optimizer = torch.optim.SGD(model.parameters(), lr = 0.1)\n", "\n", "batch_size = 15" ] }, { "cell_type": "code", "execution_count": 46, "metadata": {}, "outputs": [], "source": [ "# Trening modelu z ćwiczeń 8\n", "#https://git.wmi.amu.edu.pl/filipg/aitech-eks-pub/src/branch/master/cw/08_regresja_logistyczna.ipynb\n", "for epoch in range(5):\n", " model.train()\n", " for i in range(0, y_train.shape[0], batch_size):\n", " X = X_train[i:i+batch_size]\n", " X = torch.tensor(X)\n", " y = y_train[i:i+batch_size]\n", " y = torch.tensor(y.astype(np.float32).to_numpy()).reshape(-1,1)\n", "\n", " outputs = model(X.float())\n", " loss = criterion(outputs, y)\n", "\n", " optimizer.zero_grad()\n", " loss.backward()\n", " optimizer.step()" ] }, { "cell_type": "code", "execution_count": 59, "metadata": {}, "outputs": [], "source": [ "y_dev = []\n", "y_test = []\n", "\n", "#model.eval() will notify all your layers that you are in eval mode\n", "model.eval()\n", "\n", "#torch.no_grad() impacts the autograd engine and deactivate it. It will reduce memory usage and speed up\n", "with torch.no_grad():\n", " for i in range(0, len(X_dev), batch_size):\n", " X = X_dev[i:i+batch_size]\n", " X = torch.tensor(X)\n", " \n", " outputs = model(X.float())\n", " \n", " y = (outputs > 0.5)\n", " y_dev.extend(y)\n", "\n", " for i in range(0, len(X_test), batch_size):\n", " X = X_test[i:i+batch_size]\n", " X = torch.tensor(X)\n", "\n", " outputs = model(X.float())\n", "\n", " y = (outputs > 0.5)\n", " y_test.extend(y)" ] }, { "cell_type": "code", "execution_count": 60, "metadata": {}, "outputs": [], "source": [ "y_dev = np.asarray(y_dev, dtype=np.int32)\n", "y_test = np.asarray(y_test, dtype=np.int32)\n", "\n", "y_dev_df = pd.DataFrame({'label':y_dev})\n", "y_test_df = pd.DataFrame({'label':y_test})\n", "\n", "y_dev_df.to_csv(r'dev-0/out.tsv', sep='\\t', index=False, header=False)\n", "y_test_df.to_csv(r'test-A/out.tsv', sep='\\t', index=False, header=False)" ] }, { "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": "3.8.5" } }, "nbformat": 4, "nbformat_minor": 4 }