{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "af0090bb",
   "metadata": {},
   "source": [
    "# 从 RNN、LSTM、GRU 到 ELMo：四个可运行实验\n",
    "\n",
    "这份 Notebook 不追求训练出真正可用的语言模型，而是用小数据把四个关键现象直接跑出来：隐藏状态怎样传递、长期梯度为什么消失、门控模型怎样学习记忆任务，以及同一个词怎样得到不同的上下文表示。所有实验都可以在 CPU 上运行。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d2f02822",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:23.492698Z",
     "iopub.status.busy": "2026-08-24T16:46:23.492497Z",
     "iopub.status.idle": "2026-08-24T16:46:24.820218Z",
     "shell.execute_reply": "2026-08-24T16:46:24.819228Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Python: 3.12.13\n",
      "PyTorch: 2.13.0\n",
      "Device: CPU\n"
     ]
    }
   ],
   "source": [
    "from pathlib import Path\n",
    "import random\n",
    "import sys\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import torch\n",
    "from torch import nn\n",
    "import torch.nn.functional as F\n",
    "\n",
    "torch.set_num_threads(2)\n",
    "torch.manual_seed(42)\n",
    "np.random.seed(42)\n",
    "random.seed(42)\n",
    "\n",
    "OUTPUT_DIR = Path('outputs')\n",
    "OUTPUT_DIR.mkdir(exist_ok=True)\n",
    "\n",
    "print('Python:', sys.version.split()[0])\n",
    "print('PyTorch:', torch.__version__)\n",
    "print('Device: CPU')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "10ac95b8",
   "metadata": {},
   "source": [
    "## 实验一：RNN、LSTM、GRU 到底输出了什么？\n",
    "\n",
    "下面把两句话编码为形状为 `[batch, time]` 的整数张量，再经过词嵌入。三种循环层的 `output` 都保存每个时间步的状态；区别是 LSTM 最后还会返回细胞状态 `c_n`。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "498a046e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:24.822803Z",
     "iopub.status.busy": "2026-08-24T16:46:24.822631Z",
     "iopub.status.idle": "2026-08-24T16:46:24.870795Z",
     "shell.execute_reply": "2026-08-24T16:46:24.869998Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "token_ids shape: (2, 5)\n",
      "embedded shape : (2, 5, 5) = [batch, time, embedding_dim]\n",
      "RNN : output=(2, 5, 6), h_n=(1, 2, 6)\n",
      "      最后一个 output 等于 h_n: True\n",
      "LSTM: output=(2, 5, 6), h_n=(1, 2, 6), c_n=(1, 2, 6)\n",
      "      最后一个 output 等于 h_n: True\n",
      "GRU : output=(2, 5, 6), h_n=(1, 2, 6)\n",
      "      最后一个 output 等于 h_n: True\n"
     ]
    }
   ],
   "source": [
    "sentences = [\n",
    "    ['我', '喜欢', '学习', '人工智能', '课程'],\n",
    "    ['我', '不想', '继续', '熬夜', '工作'],\n",
    "]\n",
    "vocab = sorted({word for sentence in sentences for word in sentence})\n",
    "word_to_id = {word: index for index, word in enumerate(vocab)}\n",
    "token_ids = torch.tensor([[word_to_id[word] for word in sentence] for sentence in sentences])\n",
    "\n",
    "embedding = nn.Embedding(num_embeddings=len(vocab), embedding_dim=5)\n",
    "embedded = embedding(token_ids)\n",
    "print('token_ids shape:', tuple(token_ids.shape))\n",
    "print('embedded shape :', tuple(embedded.shape), '= [batch, time, embedding_dim]')\n",
    "\n",
    "layers = {\n",
    "    'RNN': nn.RNN(5, 6, batch_first=True),\n",
    "    'LSTM': nn.LSTM(5, 6, batch_first=True),\n",
    "    'GRU': nn.GRU(5, 6, batch_first=True),\n",
    "}\n",
    "\n",
    "for name, layer in layers.items():\n",
    "    output, state = layer(embedded)\n",
    "    if name == 'LSTM':\n",
    "        h_n, c_n = state\n",
    "        print(f'{name:4s}: output={tuple(output.shape)}, h_n={tuple(h_n.shape)}, c_n={tuple(c_n.shape)}')\n",
    "    else:\n",
    "        h_n = state\n",
    "        print(f'{name:4s}: output={tuple(output.shape)}, h_n={tuple(h_n.shape)}')\n",
    "    print('      最后一个 output 等于 h_n:', torch.allclose(output[:, -1], h_n[-1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f2fdd8ff",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:24.874051Z",
     "iopub.status.busy": "2026-08-24T16:46:24.873652Z",
     "iopub.status.idle": "2026-08-24T16:46:25.155839Z",
     "shell.execute_reply": "2026-08-24T16:46:25.155436Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tokens: 我 -> 喜欢 -> 学习 -> 人工智能 -> 课程\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 画出第一句话经过 RNN 时，6 维隐藏状态如何随词语变化。\n",
    "rnn_output, _ = layers['RNN'](embedded)\n",
    "trajectory = rnn_output[0].detach().numpy()\n",
    "\n",
    "plt.figure(figsize=(8, 4))\n",
    "for dim in range(trajectory.shape[1]):\n",
    "    plt.plot(range(len(sentences[0])), trajectory[:, dim], marker='o', label=f'h[{dim}]')\n",
    "print('tokens:', ' -> '.join(sentences[0]))\n",
    "plt.xticks(range(len(sentences[0])), [f'T{index + 1}' for index in range(len(sentences[0]))])\n",
    "plt.xlabel('Token')\n",
    "plt.ylabel('Hidden-state value')\n",
    "plt.title('RNN hidden state changes after every token')\n",
    "plt.grid(alpha=0.25)\n",
    "plt.legend(ncol=3)\n",
    "plt.tight_layout()\n",
    "plt.savefig(OUTPUT_DIR / 'rnn-hidden-state-trajectory.png', dpi=160)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c557ce6",
   "metadata": {},
   "source": [
    "图中的线不代表某个维度已经学会了固定语义，因为网络还没有训练。它只说明：读入每个词以后，`h_t` 都会发生变化，并继续成为下一个时间步的输入。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "efd18b24",
   "metadata": {},
   "source": [
    "## 实验二：让自动求导告诉我们，远处的梯度还剩多少\n",
    "\n",
    "创建三个尚未训练的网络，让最终分类分数对 40 个位置的输入向量求梯度。越靠前的位置距离输出越远。如果梯度接近 0，模型就很难知道应该怎样修改那个位置对应的计算。\n",
    "\n",
    "为了让例子可复现，三个网络使用固定随机种子；LSTM 的遗忘门偏置初始化为 1，这是一种常见的、鼓励初期保留信息的初始化方式。这个实验展示的是一次初始化下的梯度通道，不是三种模型准确率的最终结论。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "a9c189b6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:25.157667Z",
     "iopub.status.busy": "2026-08-24T16:46:25.157542Z",
     "iopub.status.idle": "2026-08-24T16:46:25.199647Z",
     "shell.execute_reply": "2026-08-24T16:46:25.199023Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "位置（从 1 开始）: [1, 5, 10, 20, 30, 40]\n",
      "RNN : ['1.47e-13', '9.11e-13', '4.78e-11', '4.11e-08', '1.19e-04', '1.06e-01']\n",
      "LSTM: ['1.30e-03', '1.10e-03', '1.53e-03', '2.90e-03', '5.74e-03', '2.71e-02']\n",
      "GRU : ['1.14e-08', '4.66e-08', '3.32e-07', '1.63e-05', '7.09e-04', '7.74e-02']\n"
     ]
    }
   ],
   "source": [
    "class MemoryClassifier(nn.Module):\n",
    "    def __init__(self, cell_type, vocab_size=8, embedding_dim=8, hidden_dim=24):\n",
    "        super().__init__()\n",
    "        self.embedding = nn.Embedding(vocab_size, embedding_dim)\n",
    "        recurrent_class = {'RNN': nn.RNN, 'LSTM': nn.LSTM, 'GRU': nn.GRU}[cell_type]\n",
    "        self.recurrent = recurrent_class(embedding_dim, hidden_dim, batch_first=True)\n",
    "        self.classifier = nn.Linear(hidden_dim, 2)\n",
    "        if cell_type == 'LSTM':\n",
    "            self._initialize_forget_gate_bias()\n",
    "\n",
    "    def _initialize_forget_gate_bias(self):\n",
    "        # PyTorch 的 LSTM bias 顺序为 input、forget、cell、output。\n",
    "        for names in self.recurrent._all_weights:\n",
    "            for name in [item for item in names if 'bias' in item]:\n",
    "                bias = getattr(self.recurrent, name)\n",
    "                quarter = bias.numel() // 4\n",
    "                bias.data[quarter:2 * quarter].fill_(1.0)\n",
    "\n",
    "    def forward_from_embeddings(self, embedded):\n",
    "        output, _ = self.recurrent(embedded)\n",
    "        return self.classifier(output[:, -1])\n",
    "\n",
    "    def forward(self, token_ids):\n",
    "        return self.forward_from_embeddings(self.embedding(token_ids))\n",
    "\n",
    "\n",
    "def gradient_profile(cell_type, sequence_length=40):\n",
    "    torch.manual_seed(42)\n",
    "    model = MemoryClassifier(cell_type)\n",
    "    sequence = torch.full((1, sequence_length), 2, dtype=torch.long)\n",
    "    sequence[:, 0] = 1\n",
    "    embedded = model.embedding(sequence).detach().requires_grad_()\n",
    "    score = model.forward_from_embeddings(embedded)[0, 1]\n",
    "    score.backward()\n",
    "    return embedded.grad.norm(dim=2)[0].detach().numpy()\n",
    "\n",
    "\n",
    "gradient_profiles = {name: gradient_profile(name) for name in ['RNN', 'LSTM', 'GRU']}\n",
    "positions_to_show = [0, 4, 9, 19, 29, 39]\n",
    "print('位置（从 1 开始）:', [position + 1 for position in positions_to_show])\n",
    "for name, values in gradient_profiles.items():\n",
    "    print(f'{name:4s}:', [f'{values[position]:.2e}' for position in positions_to_show])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "300a99ec",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:25.202472Z",
     "iopub.status.busy": "2026-08-24T16:46:25.202296Z",
     "iopub.status.idle": "2026-08-24T16:46:25.593541Z",
     "shell.execute_reply": "2026-08-24T16:46:25.591026Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x450 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 4.5))\n",
    "for name, values in gradient_profiles.items():\n",
    "    plt.semilogy(range(1, len(values) + 1), values, label=name, linewidth=2)\n",
    "plt.xlabel('Input position (1 is farthest from the final output)')\n",
    "plt.ylabel('Gradient norm, log scale')\n",
    "plt.title('How much learning signal reaches each time step?')\n",
    "plt.grid(alpha=0.25)\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "plt.savefig(OUTPUT_DIR / 'gradient-flow.png', dpi=160)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0db1762b",
   "metadata": {},
   "source": [
    "在这次固定初始化中，普通 RNN 的第 1 个位置只收到约 $10^{-13}$ 量级的梯度，而 LSTM 仍处于 $10^{-3}$ 量级。门控结构没有保证梯度永不消失，但它提供了更容易保留学习信号的通道。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fe8e2376",
   "metadata": {},
   "source": [
    "## 实验三：训练一个“记住第一个词”的任务\n",
    "\n",
    "每条序列有 40 个词：第一个词是 `红` 或 `蓝`，后面 39 个都是无关占位词。模型只能读取最后一个时间步的状态，然后回答第一个词是什么。\n",
    "\n",
    "这不是为了证明某个模型永远更好，而是把长期依赖变成一个可以观察的训练过程。固定随机种子后，我们比较达到同样目标时的收敛速度。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "6fee4282",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:25.596846Z",
     "iopub.status.busy": "2026-08-24T16:46:25.596653Z",
     "iopub.status.idle": "2026-08-24T16:46:32.373055Z",
     "shell.execute_reply": "2026-08-24T16:46:32.371631Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RNN  accuracy at epochs 1/5/10/15/20: ['49.8%', '49.8%', '49.8%', '100.0%', '100.0%']\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LSTM accuracy at epochs 1/5/10/15/20: ['49.8%', '100.0%', '100.0%', '100.0%', '100.0%']\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GRU  accuracy at epochs 1/5/10/15/20: ['49.8%', '100.0%', '100.0%', '100.0%', '100.0%']\n"
     ]
    }
   ],
   "source": [
    "def make_memory_data(sample_count, sequence_length=40):\n",
    "    labels = torch.randint(0, 2, (sample_count,))\n",
    "    sequences = torch.full((sample_count, sequence_length), 2, dtype=torch.long)\n",
    "    sequences[:, 0] = labels  # 0 代表“红”，1 代表“蓝”，2 是无关词。\n",
    "    return sequences, labels\n",
    "\n",
    "\n",
    "torch.manual_seed(42)\n",
    "x_train, y_train = make_memory_data(1200)\n",
    "x_test, y_test = make_memory_data(400)\n",
    "\n",
    "\n",
    "def train_memory_model(cell_type, epochs=20):\n",
    "    torch.manual_seed(42)\n",
    "    model = MemoryClassifier(cell_type)\n",
    "    optimizer = torch.optim.Adam(model.parameters(), lr=0.005)\n",
    "    loss_function = nn.CrossEntropyLoss()\n",
    "    accuracy_history = []\n",
    "\n",
    "    for epoch in range(epochs):\n",
    "        model.train()\n",
    "        for indices in torch.randperm(len(x_train)).split(64):\n",
    "            loss = loss_function(model(x_train[indices]), y_train[indices])\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)\n",
    "            optimizer.step()\n",
    "\n",
    "        model.eval()\n",
    "        with torch.no_grad():\n",
    "            prediction = model(x_test).argmax(dim=1)\n",
    "            accuracy = (prediction == y_test).float().mean().item()\n",
    "        accuracy_history.append(accuracy)\n",
    "\n",
    "    return model, accuracy_history\n",
    "\n",
    "\n",
    "trained_models = {}\n",
    "accuracy_histories = {}\n",
    "for cell_type in ['RNN', 'LSTM', 'GRU']:\n",
    "    model, history = train_memory_model(cell_type)\n",
    "    trained_models[cell_type] = model\n",
    "    accuracy_histories[cell_type] = history\n",
    "    checkpoints = [history[index] for index in [0, 4, 9, 14, 19]]\n",
    "    print(f'{cell_type:4s} accuracy at epochs 1/5/10/15/20:', [f'{value:.1%}' for value in checkpoints])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "28fe1a3e",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:32.376945Z",
     "iopub.status.busy": "2026-08-24T16:46:32.376126Z",
     "iopub.status.idle": "2026-08-24T16:46:32.535846Z",
     "shell.execute_reply": "2026-08-24T16:46:32.534668Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x450 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8, 4.5))\n",
    "for name, history in accuracy_histories.items():\n",
    "    plt.plot(range(1, len(history) + 1), history, marker='o', markersize=3, label=name)\n",
    "plt.axhline(0.5, color='gray', linestyle='--', label='Random guess')\n",
    "plt.ylim(0.4, 1.05)\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Test accuracy')\n",
    "plt.title('Remember the first token across 40 time steps')\n",
    "plt.grid(alpha=0.25)\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "plt.savefig(OUTPUT_DIR / 'memory-task-accuracy.png', dpi=160)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f8f54619",
   "metadata": {},
   "source": [
    "在当前随机种子下，LSTM 与 GRU 较早学会任务，普通 RNN 需要更多轮训练。更换初始化、序列长度或加入随机干扰后，曲线会变化；这恰好说明不能凭一次小实验宣布某个模型永远更好。这个实验真正想展示的是：门控模型通常更容易为远处的信息建立稳定通道。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8aa92f85",
   "metadata": {},
   "source": [
    "## 实验四：做一个迷你 ELMo 式动态词向量\n",
    "\n",
    "真正的 ELMo 使用字符级 CNN 和两层双向 LSTM，并在大规模语料上训练。这里把它缩小成共享词嵌入、一个前向 LSTM、一个反向 LSTM和共享输出层，只保留最关键的思想：分别训练两个方向的语言模型，再拼接同一位置的双向状态。\n",
    "\n",
    "语料中既有水果含义的“苹果”，也有科技公司含义的“苹果”。静态嵌入查表时两者必然完全相同；经过上下文编码以后，向量则会发生变化。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "add37868",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:32.537498Z",
     "iopub.status.busy": "2026-08-24T16:46:32.537406Z",
     "iopub.status.idle": "2026-08-24T16:46:32.541343Z",
     "shell.execute_reply": "2026-08-24T16:46:32.540664Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "语料句数: 16 词表大小: 59 张量形状: (16, 8)\n"
     ]
    }
   ],
   "source": [
    "fruit_sentences = [\n",
    "    '我 喜欢 吃 苹果 很 甜', '今天 买 的 苹果 很 甜',\n",
    "    '这个 苹果 吃起来 很 香 啊', '她 削开 苹果 慢慢 吃 了',\n",
    "    '新鲜 苹果 富含 多种 维生素 啊', '桌上 放着 两个 红 苹果 啊',\n",
    "    '孩子 喜欢 苹果 的 味道 啊', '水果店 苹果 今天 正在 打折 啊',\n",
    "]\n",
    "technology_sentences = [\n",
    "    '我 使用 苹果 电脑 工作 啊', '苹果 公司 发布 新 手机 啊',\n",
    "    '这台 苹果 手机 运行 很 流畅', '他 买了 苹果 最新 电脑 啊',\n",
    "    '苹果 系统 今天 推送 新 更新', '开发者 使用 苹果 电脑 写 代码',\n",
    "    '苹果 发布会 展示 全新 芯片 啊', '这家 苹果 门店 正在 营业 啊',\n",
    "]\n",
    "toy_sentences = [sentence.split() for sentence in fruit_sentences + technology_sentences]\n",
    "toy_vocab = ['<pad>', '<bos>', '<eos>'] + sorted({word for sentence in toy_sentences for word in sentence})\n",
    "toy_word_to_id = {word: index for index, word in enumerate(toy_vocab)}\n",
    "max_length = max(map(len, toy_sentences)) + 2\n",
    "\n",
    "\n",
    "def encode_toy_sentence(sentence):\n",
    "    ids = [toy_word_to_id['<bos>']] + [toy_word_to_id[word] for word in sentence] + [toy_word_to_id['<eos>']]\n",
    "    return ids + [toy_word_to_id['<pad>']] * (max_length - len(ids))\n",
    "\n",
    "\n",
    "toy_batch = torch.tensor([encode_toy_sentence(sentence) for sentence in toy_sentences])\n",
    "print('语料句数:', len(toy_sentences), '词表大小:', len(toy_vocab), '张量形状:', tuple(toy_batch.shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "20de5541",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:32.543527Z",
     "iopub.status.busy": "2026-08-24T16:46:32.543345Z",
     "iopub.status.idle": "2026-08-24T16:46:33.141057Z",
     "shell.execute_reply": "2026-08-24T16:46:33.139649Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "双向语言模型损失：8.144 -> 0.797\n"
     ]
    }
   ],
   "source": [
    "class TinyBidirectionalLanguageModel(nn.Module):\n",
    "    def __init__(self, vocab_size, embedding_dim=16, hidden_dim=24):\n",
    "        super().__init__()\n",
    "        self.embedding = nn.Embedding(vocab_size, embedding_dim, padding_idx=0)\n",
    "        self.forward_lstm = nn.LSTM(embedding_dim, hidden_dim, batch_first=True)\n",
    "        self.backward_lstm = nn.LSTM(embedding_dim, hidden_dim, batch_first=True)\n",
    "        self.shared_output = nn.Linear(hidden_dim, vocab_size)\n",
    "\n",
    "    def directional_states(self, token_ids):\n",
    "        embedded = self.embedding(token_ids)\n",
    "        forward_states, _ = self.forward_lstm(embedded)\n",
    "        reversed_states, _ = self.backward_lstm(torch.flip(embedded, dims=[1]))\n",
    "        return forward_states, reversed_states\n",
    "\n",
    "\n",
    "torch.manual_seed(11)\n",
    "toy_bilm = TinyBidirectionalLanguageModel(len(toy_vocab))\n",
    "optimizer = torch.optim.Adam(toy_bilm.parameters(), lr=0.02)\n",
    "loss_history = []\n",
    "\n",
    "for step in range(300):\n",
    "    forward_states, backward_states_reversed = toy_bilm.directional_states(toy_batch)\n",
    "    reversed_tokens = torch.flip(toy_batch, dims=[1])\n",
    "\n",
    "    forward_loss = F.cross_entropy(\n",
    "        toy_bilm.shared_output(forward_states[:, :-1]).reshape(-1, len(toy_vocab)),\n",
    "        toy_batch[:, 1:].reshape(-1),\n",
    "        ignore_index=toy_word_to_id['<pad>'],\n",
    "    )\n",
    "    backward_loss = F.cross_entropy(\n",
    "        toy_bilm.shared_output(backward_states_reversed[:, :-1]).reshape(-1, len(toy_vocab)),\n",
    "        reversed_tokens[:, 1:].reshape(-1),\n",
    "        ignore_index=toy_word_to_id['<pad>'],\n",
    "    )\n",
    "    loss = forward_loss + backward_loss\n",
    "    optimizer.zero_grad()\n",
    "    loss.backward()\n",
    "    optimizer.step()\n",
    "    loss_history.append(loss.item())\n",
    "\n",
    "print(f'双向语言模型损失：{loss_history[0]:.3f} -> {loss_history[-1]:.3f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "a159fd05",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:33.144328Z",
     "iopub.status.busy": "2026-08-24T16:46:33.143950Z",
     "iopub.status.idle": "2026-08-24T16:46:33.162882Z",
     "shell.execute_reply": "2026-08-24T16:46:33.156096Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "静态查表：任意两个“苹果”的余弦相似度 = 1.000\n",
      "上下文化后：Fruit-1 与 Tech-1 的余弦相似度 = 0.470\n",
      "注意：小语料只演示“向量会随上下文改变”，不能把具体数值当成真实语义质量。\n"
     ]
    }
   ],
   "source": [
    "def contextual_apple_vectors(model, batch, sentences):\n",
    "    model.eval()\n",
    "    with torch.no_grad():\n",
    "        forward_states, backward_states_reversed = model.directional_states(batch)\n",
    "        backward_states = torch.flip(backward_states_reversed, dims=[1])\n",
    "        vectors = []\n",
    "        for row, sentence in enumerate(sentences):\n",
    "            apple_position = sentence.index('苹果') + 1  # 加 1 是因为句首有 <bos>。\n",
    "            vectors.append(torch.cat([\n",
    "                forward_states[row, apple_position],\n",
    "                backward_states[row, apple_position],\n",
    "            ]))\n",
    "    return F.normalize(torch.stack(vectors), dim=1)\n",
    "\n",
    "\n",
    "apple_vectors = contextual_apple_vectors(toy_bilm, toy_batch, toy_sentences)\n",
    "selected_indices = [0, 1, 8, 9]\n",
    "selected_vectors = apple_vectors[selected_indices]\n",
    "similarity_matrix = (selected_vectors @ selected_vectors.T).numpy()\n",
    "labels = ['Fruit-1', 'Fruit-2', 'Tech-1', 'Tech-2']\n",
    "\n",
    "apple_id = toy_word_to_id['苹果']\n",
    "static_vector = toy_bilm.embedding.weight[apple_id].detach()\n",
    "static_similarity = F.cosine_similarity(static_vector, static_vector, dim=0).item()\n",
    "print(f'静态查表：任意两个“苹果”的余弦相似度 = {static_similarity:.3f}')\n",
    "print(f'上下文化后：Fruit-1 与 Tech-1 的余弦相似度 = {similarity_matrix[0, 2]:.3f}')\n",
    "print('注意：小语料只演示“向量会随上下文改变”，不能把具体数值当成真实语义质量。')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "0ce049ae",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2026-08-24T16:46:33.165427Z",
     "iopub.status.busy": "2026-08-24T16:46:33.165167Z",
     "iopub.status.idle": "2026-08-24T16:46:33.410358Z",
     "shell.execute_reply": "2026-08-24T16:46:33.409413Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x500 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(6, 5))\n",
    "image = plt.imshow(similarity_matrix, vmin=-1, vmax=1, cmap='coolwarm')\n",
    "plt.colorbar(image, label='Cosine similarity')\n",
    "plt.xticks(range(len(labels)), labels, rotation=25)\n",
    "plt.yticks(range(len(labels)), labels)\n",
    "for row in range(len(labels)):\n",
    "    for column in range(len(labels)):\n",
    "        plt.text(column, row, f'{similarity_matrix[row, column]:.2f}', ha='center', va='center')\n",
    "plt.title('Same token, different contextual vectors')\n",
    "plt.tight_layout()\n",
    "plt.savefig(OUTPUT_DIR / 'elmo-context-similarity.png', dpi=160)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e7a7a12f",
   "metadata": {},
   "source": [
    "## 小结\n",
    "\n",
    "1. `output[:, t]` 就是循环网络读到第 $t$ 个词后的隐藏状态。\n",
    "2. 普通 RNN 的远距离梯度可能快速衰减；LSTM 和 GRU 提供了更稳定的门控通道。\n",
    "3. 小型记忆任务能够直观看到不同结构的训练速度，但一次实验不能替代严谨比较。\n",
    "4. 静态词嵌入只做一次查表；ELMo 式表示则把双向语言模型在当前位置的状态作为词表示，因此同一个词会随上下文变化。"
   ]
  }
 ],
 "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": "3.12.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}
