Add files via upload

char-seq.ipynb

@@ -0,0 +1,231 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### character-level LSTM in Pytorch"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 该笔记记录的模型如下所示\n",
+    "<img src=\"assets/charseq.jpeg\" width=\"500\">"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 导入相关的包"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np \n",
+    "import torch \n",
+    "from torch import nn\n",
+    "import torch.nn.functional as F"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "####  读取数据"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "with open('data/anna.txt', 'r') as f:\n",
+    "    text = f.read()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'Chapter 1\\n\\n\\nHappy families are all alike; every unhappy family is unhappy in its own\\nway.\\n\\nEverythin'"
+      ]
+     },
+     "execution_count": 3,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "text[:100]"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 对其进行标记（tokenization）\n",
+    "即是为其数据建造一个词典，让每个字符均有对应的标签"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "chars = tuple(set(text))\n",
+    "int2char = dict(enumerate(chars))\n",
+    "char2int = {ch:li for li,ch in int2char.items()}\n",
+    "\n",
+    "encoded = np.array([char2int[ch] for ch in text])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "[14 45 48 54 76 72 74 46 75 22 22 22 82 48 54 54 18 46 80 48 12 49 27 49\n",
+      " 72 62 46 48 74 72 46 48 27 27 46 48 27 49 28 72  4 46 72  2 72 74 18 46\n",
+      " 70 39 45 48 54 54 18 46 80 48 12 49 27 18 46 49 62 46 70 39 45 48 54 54\n",
+      " 18 46 49 39 46 49 76 62 46  9 51 39 22 51 48 18  0 22 22 58  2 72 74 18\n",
+      " 76 45 49 39]\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(encoded[:100])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 将其转换为one-hot编码\n",
+    "构造一个函数，使其具有使得原本的文本转化为one-hot功能"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def one_hot_encoded(arr,n_labels):\n",
+    "    \n",
+    "    one_hot = np.zeros((np.multiply(*arr.shape),n_labels),dtype = np.float32)\n",
+    "    \n",
+    "    one_hot [np.arange(one_hot.shape[0]),arr.flatten()] = 1\n",
+    "    \n",
+    "    ont_hot = one_hot.reshape((*arr.shape,n_labels))\n",
+    "    \n",
+    "    return one_hot"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "[[0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]\n",
+      " [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]\n",
+      " [0. 0. 0. 0. 0. 1. 0. 0. 0. 0.]\n",
+      " [0. 0. 0. 0. 0. 0. 0. 1. 0. 0.]\n",
+      " [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]]\n"
+     ]
+    }
+   ],
+   "source": [
+    "test_seq = np.array([[2,4,5,7,2]])\n",
+    "one_hot = one_hot_encoded(test_seq , 10)\n",
+    "print(one_hot)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### 建造一个mini-batch\n",
+    "此处对应的是英文文本的batch，且训练数据是一段英文文本，故其batch的组成是其不同的英文序列\n",
+    "且为了保留batch的完整，需要丢弃掉最后一部分空余的数据"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def get_batches(arr,batch_size,seq_length):\n",
+    "    \n",
+    "    batch_size_total = batch_size * seq_length\n",
+    "    \n",
+    "    n_batches = len(arr) //batch_size_total\n",
+    "    \n",
+    "    arr = arr[:n_batches * batch_size_total]\n",
+    "    \n",
+    "    arr = arr.reshape((batch_size , -1))\n",
+    "    \n",
+    "    for i in range(0,arr.shape[1],seq_length):\n",
+    "    "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "np.multiply?"
+   ]
+  }
+ ],
+ "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": "3.6.6"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}