6.3 LSTM的记忆能力实验

长短期记忆网络（Long Short-Term Memory Network，LSTM）是一种可以有效缓解长程依赖问题的循环神经网络．LSTM 的特点是引入了一个新的内部状态（Internal State）c∈R^D 和门控机制（Gating Mechanism）．不同时刻的内部状态以近似线性的方式进行传递，从而缓解梯度消失或梯度爆炸问题．同时门控机制进行信息筛选，可以有效地增加记忆能力．
使用LSTM模型重新进行数字求和实验，验证LSTM模型的长程依赖能力。

6.3.1 模型构建

使用第6.1.2.4节中定义Model_RNN4SeqClass模型，并构建 LSTM 算子．

只需要实例化 LSTM ，并传入Model_RNN4SeqClass模型，就可以用 LSTM 进行数字求和实验。

6.3.1.1 LSTM层

自定义LSTM算子
nn.LSTM
将自定义LSTM与pytorch内置的LSTM进行对比
LSTM层的代码与SRN层结构相似，只是在SRN层的基础上增加了内部状态、输入门、遗忘门和输出门的定义和计算。这里LSTM层的输出也依然为序列的最后一个位置的隐状态向量。代码实现如下：

class LSTM(nn.Module):def __init__(self, input_size, hidden_size, Wi_attr=None, Wf_attr=None, Wo_attr=None, Wc_attr=None,Ui_attr=None, Uf_attr=None, Uo_attr=None, Uc_attr=None, bi_attr=None, bf_attr=None,bo_attr=None, bc_attr=None):super(LSTM, self).__init__()self.input_size = input_sizeself.hidden_size = hidden_sizeW_i = torch.randn([input_size, hidden_size])W_f = torch.randn([input_size, hidden_size])W_o = torch.randn([input_size, hidden_size])W_c = torch.randn([input_size, hidden_size])U_i = torch.randn([hidden_size, hidden_size])U_f = torch.randn([hidden_size, hidden_size])U_o = torch.randn([hidden_size, hidden_size])U_c = torch.randn([hidden_size, hidden_size])b_i = torch.randn([1, hidden_size])b_f = torch.randn([1, hidden_size])b_o = torch.randn([1, hidden_size])b_c = torch.randn([1, hidden_size])self.W_i = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(W_i, dtype=torch.float32), gain=1.0))# 初始化模型参数self.W_f = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(W_f, dtype=torch.float32), gain=1.0))self.W_o = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(W_o, dtype=torch.float32), gain=1.0))self.W_c = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(W_c, dtype=torch.float32), gain=1.0))self.U_i = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(U_i, dtype=torch.float32), gain=1.0))self.U_f = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(U_f, dtype=torch.float32), gain=1.0))self.U_o = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(U_o, dtype=torch.float32), gain=1.0))self.U_c = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(U_c, dtype=torch.float32), gain=1.0))self.b_i = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(b_i, dtype=torch.float32), gain=1.0))self.b_f = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(b_f, dtype=torch.float32), gain=1.0))self.b_o = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(b_o, dtype=torch.float32), gain=1.0))self.b_c = torch.nn.Parameter(torch.nn.init.xavier_uniform_(torch.as_tensor(b_c, dtype=torch.float32), gain=1.0))# 初始化状态向量和隐状态向量def init_state(self, batch_size):hidden_state = torch.zeros([batch_size, self.hidden_size])cell_state = torch.zeros([batch_size, self.hidden_size])return hidden_state, cell_state# 定义前向计算def forward(self, inputs, states=None):# inputs: 输入数据，其shape为batch_size x seq_len x input_sizebatch_size, seq_len, input_size = inputs.shape# 初始化起始的单元状态和隐状态向量，其shape为batch_size x hidden_sizeif states is None:states = self.init_state(batch_size)hidden_state, cell_state = states# 执行LSTM计算，包括：输入门、遗忘门和输出门、候选内部状态、内部状态和隐状态向量for step in range(seq_len):# 获取当前时刻的输入数据step_input: 其shape为batch_size x input_sizestep_input = inputs[:, step, :]# 计算输入门, 遗忘门和输出门, 其shape为：batch_size x hidden_sizeI_gate = F.sigmoid(torch.matmul(step_input, self.W_i) + torch.matmul(hidden_state, self.U_i) + self.b_i)F_gate = F.sigmoid(torch.matmul(step_input, self.W_f) + torch.matmul(hidden_state, self.U_f) + self.b_f)O_gate = F.sigmoid(torch.matmul(step_input, self.W_o) + torch.matmul(hidden_state, self.U_o) + self.b_o)# 计算候选状态向量, 其shape为：batch_size x hidden_sizeC_tilde = F.tanh(torch.matmul(step_input, self.W_c) + torch.matmul(hidden_state, self.U_c) + self.b_c)# 计算单元状态向量, 其shape为：batch_size x hidden_sizecell_state = F_gate * cell_state + I_gate * C_tilde# 计算隐状态向量，其shape为：batch_size x hidden_sizehidden_state = O_gate * F.tanh(cell_state)return hidden_state
Wi_attr = torch.nn.Parameter(torch.tensor([[0.1, 0.2], [0.1, 0.2]]))
Wf_attr = torch.nn.Parameter(torch.tensor([[0.1, 0.2], [0.1, 0.2]]))
Wo_attr = torch.nn.Parameter(torch.tensor([[0.1, 0.2], [0.1, 0.2]]))
Wc_attr = torch.nn.Parameter(torch.tensor([[0.1, 0.2], [0.1, 0.2]]))
Ui_attr = torch.nn.Parameter(torch.tensor([[0.0, 0.1], [0.1, 0.0]]))
Uf_attr = torch.nn.Parameter(torch.tensor([[0.0, 0.1], [0.1, 0.0]]))
Uo_attr = torch.nn.Parameter(torch.tensor([[0.0, 0.1], [0.1, 0.0]]))
Uc_attr = torch.nn.Parameter(torch.tensor([[0.0, 0.1], [0.1, 0.0]]))
bi_attr = torch.nn.Parameter(torch.tensor([[0.1, 0.1]]))
bf_attr = torch.nn.Parameter(torch.tensor([[0.1, 0.1]]))
bo_attr = torch.nn.Parameter(torch.tensor([[0.1, 0.1]]))
bc_attr = torch.nn.Parameter(torch.tensor([[0.1, 0.1]]))lstm = LSTM(2, 2, Wi_attr=Wi_attr, Wf_attr=Wf_attr, Wo_attr=Wo_attr, Wc_attr=Wc_attr,Ui_attr=Ui_attr, Uf_attr=Uf_attr, Uo_attr=Uo_attr, Uc_attr=Uc_attr,bi_attr=bi_attr, bf_attr=bf_attr, bo_attr=bo_attr, bc_attr=bc_attr)inputs = torch.tensor([[[1, 0]]], dtype=torch.float32)
hidden_state = lstm(inputs)
print(hidden_state)

运行结果：

将自己实现的SRN和Torch框架内置的SRN返回的结果进行打印展示：

# 这里创建一个随机数组作为测试数据，数据shape为batch_size x seq_len x input_size
batch_size, seq_len, input_size = 8, 20, 32
inputs = torch.randn([batch_size, seq_len, input_size])# 设置模型的hidden_size
hidden_size = 32
torch_lstm = nn.LSTM(input_size, hidden_size)
self_lstm = LSTM(input_size, hidden_size)self_hidden_state = self_lstm(inputs)
torch_outputs, (torch_hidden_state, torch_cell_state) = torch_lstm(inputs)print("self_lstm hidden_state: ", self_hidden_state.shape)
print("torch_lstm outpus:", torch_outputs.shape)
print("torch_lstm hidden_state:", torch_hidden_state.shape)
print("torch_lstm cell_state:", torch_cell_state.shape)

运行结果：

定义输入数据inputs，然后将该数据分别传入Torch内置的LSTM与自己实现的LSTM模型中，最后通过对比两者的隐状态输出向量：

import torch
torch.manual_seed(0)# 这里创建一个随机数组作为测试数据，数据shape为batch_size x seq_len x input_size
batch_size, seq_len, input_size, hidden_size = 2, 5, 10, 10
inputs = torch.randn([batch_size, seq_len, input_size])# 设置模型的hidden_size
# bih_attr = torch.nn.Parameter(torch.tensor(torch.zeros([4*hidden_size, ])))
torch_lstm = nn.LSTM(input_size, hidden_size, bias=True)# 获取torch_lstm中的参数，并设置相应的paramAttr,用于初始化lstm
print(torch_lstm.weight_ih_l0.T.shape)
chunked_W = torch.split(torch_lstm.weight_ih_l0.T, 4, dim=-1)
chunked_U = torch.split(torch_lstm.weight_hh_l0.T, 4, dim=-1)
chunked_b = torch.split(torch_lstm.bias_hh_l0.T, 4, dim=-1)Wi_attr = torch.nn.Parameter(torch.tensor(chunked_W[0].clone().detach().requires_grad_(True)))
Wf_attr = torch.nn.Parameter(torch.tensor(chunked_W[1].clone().detach().requires_grad_(True)))
Wc_attr = torch.nn.Parameter(torch.tensor(chunked_W[2].clone().detach().requires_grad_(True)))
Wo_attr = torch.nn.Parameter(torch.tensor(chunked_W[3].clone().detach().requires_grad_(True)))
Ui_attr = torch.nn.Parameter(torch.tensor(chunked_U[0].clone().detach().requires_grad_(True)))
Uf_attr = torch.nn.Parameter(torch.tensor(chunked_U[1].clone().detach().requires_grad_(True)))
Uc_attr = torch.nn.Parameter(torch.tensor(chunked_U[2].clone().detach().requires_grad_(True)))
Uo_attr = torch.nn.Parameter(torch.tensor(chunked_U[3].clone().detach().requires_grad_(True)))
bi_attr = torch.nn.Parameter(torch.tensor(chunked_b[0].clone().detach().requires_grad_(True)))
bf_attr = torch.nn.Parameter(torch.tensor(chunked_b[1].clone().detach().requires_grad_(True)))
bc_attr = torch.nn.Parameter(torch.tensor(chunked_b[2].clone().detach().requires_grad_(True)))
bo_attr = torch.nn.Parameter(torch.tensor(chunked_b[3].clone().detach().requires_grad_(True)))
self_lstm = LSTM(input_size, hidden_size, Wi_attr=Wi_attr, Wf_attr=Wf_attr, Wo_attr=Wo_attr, Wc_attr=Wc_attr,Ui_attr=Ui_attr, Uf_attr=Uf_attr, Uo_attr=Uo_attr, Uc_attr=Uc_attr,bi_attr=bi_attr, bf_attr=bf_attr, bo_attr=bo_attr, bc_attr=bc_attr)# 进行前向计算，获取隐状态向量，并打印展示
self_hidden_state = self_lstm(inputs)
torch_outputs, (torch_hidden_state, _) = torch_lstm(inputs)
print("torch SRN:\n", torch_hidden_state.detach().numpy().squeeze(0))
print("self SRN:\n", self_hidden_state.detach().numpy())

运行结果：

可以看到，两者的输出基本是一致的。另外，还可以进行对比两者在运算速度方面的差异。代码实现如下：

import time# 这里创建一个随机数组作为测试数据，数据shape为batch_size x seq_len x input_size
batch_size, seq_len, input_size = 8, 20, 32
inputs = torch.randn([batch_size, seq_len, input_size])# 设置模型的hidden_size
hidden_size = 32
self_lstm = LSTM(input_size, hidden_size)
torch_lstm = nn.LSTM(input_size, hidden_size)# 计算自己实现的SRN运算速度
model_time = 0
for i in range(100):strat_time = time.time()hidden_state = self_lstm(inputs)# 预热10次运算，不计入最终速度统计if i < 10:continueend_time = time.time()model_time += (end_time - strat_time)
avg_model_time = model_time / 90
print('self_lstm speed:', avg_model_time, 's')# 计算torch内置的SRN运算速度
model_time = 0
for i in range(100):strat_time = time.time()outputs, (hidden_state, cell_state) = torch_lstm(inputs)# 预热10次运算，不计入最终速度统计if i < 10:continueend_time = time.time()model_time += (end_time - strat_time)
avg_model_time = model_time / 90
print('torch_lstm speed:', avg_model_time, 's')

运行结果：

可以看到，由于PyTorch底层采用了C++实现并进行优化，Paddle框架内置的LSTM运行效率远远高于自己实现的LSTM。

6.3.1.2 模型汇总

# 基于RNN实现数字预测的模型
class Model_RNN4SeqClass(nn.Module):def __init__(self, model, num_digits, input_size, hidden_size, num_classes):super(Model_RNN4SeqClass, self).__init__()# 传入实例化的RNN层，例如SRNself.rnn_model = model# 词典大小self.num_digits = num_digits# 嵌入向量的维度self.input_size = input_size# 定义Embedding层self.embedding = Embedding(num_digits, input_size)# 定义线性层self.linear = nn.Linear(hidden_size, num_classes)def forward(self, inputs):# 将数字序列映射为相应向量inputs_emb = self.embedding(inputs)# 调用RNN模型hidden_state = self.rnn_model(inputs_emb)# 使用最后一个时刻的状态进行数字预测logits = self.linear(hidden_state)return logits

6.3.2 模型训练

6.3.2.1 训练指定长度的数字预测模型

本节将基于RunnerV3类进行训练，首先定义模型训练的超参数，并保证和简单循环网络的超参数一致. 然后定义一个train函数，其可以通过指定长度的数据集，并进行训练. 在train函数中，首先加载长度为length的数据，然后实例化各项组件并创建对应的Runner，然后训练该Runner。同时在本节将使用4.5.4节定义的准确度（Accuracy）作为评估指标，代码实现如下：

import os
import random
import torch
import numpy as np# 训练轮次
num_epochs = 500
# 学习率
lr = 0.001
# 输入数字的类别数
num_digits = 10
# 将数字映射为向量的维度
input_size = 32
# 隐状态向量的维度
hidden_size = 32
# 预测数字的类别数
num_classes = 19
# 批大小
batch_size = 8
# 模型保存目录
save_dir = "./checkpoints"# 可以设置不同的length进行不同长度数据的预测实验
def train(length):print(f"\n====> Training LSTM with data of length {length}.")np.random.seed(0)random.seed(0)torch.manual_seed(0)# 加载长度为length的数据data_path = f"D:/datasets/{length}"train_examples, dev_examples, test_examples = load_data(data_path)train_set, dev_set, test_set = DigitSumDataset(train_examples), DigitSumDataset(dev_examples), DigitSumDataset(test_examples)train_loader = DataLoader(train_set, batch_size=batch_size)dev_loader = DataLoader(dev_set, batch_size=batch_size)test_loader = DataLoader(test_set, batch_size=batch_size)# 实例化模型base_model = LSTM(input_size, hidden_size)model = Model_RNN4SeqClass(base_model, num_digits, input_size, hidden_size, num_classes)# 指定优化器optimizer = torch.optim.Adam(model.parameters(),lr)# 定义评价指标metric = Accuracy()# 定义损失函数loss_fn = torch.nn.CrossEntropyLoss()# 基于以上组件，实例化Runnerrunner = RunnerV3(model, optimizer, loss_fn, metric)# 进行模型训练model_save_path = os.path.join(save_dir, f"best_lstm_model_{length}.pdparams")runner.train(train_loader, dev_loader, num_epochs=num_epochs, eval_steps=100, log_steps=100, save_path=model_save_path)return runner

RunnerV3:

class RunnerV3(object):def __init__(self, model, optimizer, loss_fn, metric, **kwargs):self.model = modelself.optimizer = optimizerself.loss_fn = loss_fnself.metric = metric  # 只用于计算评价指标# 记录训练过程中的评价指标变化情况self.dev_scores = []# 记录训练过程中的损失函数变化情况self.train_epoch_losses = []  # 一个epoch记录一次lossself.train_step_losses = []  # 一个step记录一次lossself.dev_losses = []# 记录全局最优指标self.best_score = 0def train(self, train_loader, dev_loader=None, **kwargs):# 将模型切换为训练模式self.model.train()# 传入训练轮数，如果没有传入值则默认为0num_epochs = kwargs.get("num_epochs", 0)# 传入log打印频率，如果没有传入值则默认为100log_steps = kwargs.get("log_steps", 100)# 评价频率eval_steps = kwargs.get("eval_steps", 0)# 传入模型保存路径，如果没有传入值则默认为"best_model.pdparams"save_path = kwargs.get("save_path", "best_model.pdparams")custom_print_log = kwargs.get("custom_print_log", None)# 训练总的步数num_training_steps = num_epochs * len(train_loader)if eval_steps:if self.metric is None:raise RuntimeError('Error: Metric can not be None!')if dev_loader is None:raise RuntimeError('Error: dev_loader can not be None!')# 运行的step数目global_step = 0# 进行num_epochs轮训练for epoch in range(num_epochs):# 用于统计训练集的损失total_loss = 0for step, data in enumerate(train_loader):X, y = data# 获取模型预测logits = self.model(X)y = torch.tensor(y, dtype=torch.int64)loss = self.loss_fn(logits, y)  # 默认求meantotal_loss += loss# 训练过程中，每个step的loss进行保存self.train_step_losses.append((global_step, loss.item()))if log_steps and global_step % log_steps == 0:print(f"[Train] epoch: {epoch}/{num_epochs}, step: {global_step}/{num_training_steps}, loss: {loss.item():.5f}")# 梯度反向传播，计算每个参数的梯度值loss.backward()torch.nn.utils.clip_grad_norm_(parameters=self.model.parameters(), max_norm=5, norm_type=1)if custom_print_log:custom_print_log(self)# 小批量梯度下降进行参数更新self.optimizer.step()# 梯度归零self.optimizer.zero_grad()# 判断是否需要评价if eval_steps > 0 and global_step > 0 and \(global_step % eval_steps == 0 or global_step == (num_training_steps - 1)):dev_score, dev_loss = self.evaluate(dev_loader, global_step=global_step)print(f"[Evaluate]  dev score: {dev_score:.5f}, dev loss: {dev_loss:.5f}")# 将模型切换为训练模式self.model.train()# 如果当前指标为最优指标，保存该模型if dev_score > self.best_score:self.save_model(save_path)print(f"[Evaluate] best accuracy performence has been updated: {self.best_score:.5f} --> {dev_score:.5f}")self.best_score = dev_scoreglobal_step += 1# 当前epoch 训练loss累计值trn_loss = (total_loss / len(train_loader)).item()# epoch粒度的训练loss保存self.train_epoch_losses.append(trn_loss)print("[Train] Training done!")# 模型评估阶段，使用'paddle.no_grad()'控制不计算和存储梯度@torch.no_grad()def evaluate(self, dev_loader, **kwargs):assert self.metric is not None# 将模型设置为评估模式self.model.eval()global_step = kwargs.get("global_step", -1)# 用于统计训练集的损失total_loss = 0# 重置评价self.metric.reset()# 遍历验证集每个批次for batch_id, data in enumerate(dev_loader):X, y = data# 计算模型输出logits = self.model(X)y = y.clone().detach()# 计算损失函数loss = self.loss_fn(logits, y).item()# 累积损失total_loss += loss# 累积评价self.metric.update(logits, y)dev_loss = (total_loss / len(dev_loader))dev_score = self.metric.accumulate()# 记录验证集lossif global_step != -1:self.dev_losses.append((global_step, dev_loss))self.dev_scores.append(dev_score)return dev_score, dev_loss# 模型评估阶段，使用'paddle.no_grad()'控制不计算和存储梯度@torch.no_grad()def predict(self, x, **kwargs):# 将模型设置为评估模式self.model.eval()# 运行模型前向计算，得到预测值logits = self.model(x)return logitsdef save_model(self, save_path):torch.save(self.model.state_dict(), save_path)def load_model(self, model_path):model_state_dict = torch.load(model_path)self.model.state_dict(model_state_dict)

Accuracy:

class Accuracy():def __init__(self, is_logist=True):"""输入：- is_logist: outputs是logist还是激活后的值"""# 用于统计正确的样本个数self.num_correct = 0# 用于统计样本的总数self.num_count = 0self.is_logist = is_logistdef update(self, outputs, labels):"""输入：- outputs: 预测值, shape=[N,class_num]- labels: 标签值, shape=[N,1]"""# 判断是二分类任务还是多分类任务，shape[1]=1时为二分类任务，shape[1]>1时为多分类任务if outputs.shape[1] == 1:  # 二分类outputs = torch.squeeze(outputs, dim=-1)if self.is_logist:# logist判断是否大于0preds = torch.tensor((outputs >= 0), dtype=torch.float32)else:# 如果不是logist，判断每个概率值是否大于0.5，当大于0.5时，类别为1，否则类别为0preds = torch.tensor((outputs >= 0.5), dtype=torch.float32)else:# 多分类时，使用'paddle.argmax'计算最大元素索引作为类别preds = torch.argmax(outputs, dim=1)preds = preds.clone().detach()# 获取本批数据中预测正确的样本个数labels = torch.squeeze(labels, dim=-1)batch_correct = torch.sum((preds == labels).clone().detach()).numpy()batch_count = len(labels)# 更新num_correct 和 num_countself.num_correct += batch_correctself.num_count += batch_countdef accumulate(self):# 使用累计的数据，计算总的指标if self.num_count == 0:return 0return self.num_correct / self.num_countdef reset(self):# 重置正确的数目和总数self.num_correct = 0self.num_count = 0def name(self):return "Accuracy"

6.3.2.2 多组训练

接下来，分别进行数据长度为10, 15, 20, 25, 30, 35的数字预测模型训练实验，训练后的runner保存至runners字典中。

# LSTM训练
lstm_runners = {}
lengths = [10, 15, 20, 25, 30, 35]
for length in lengths:runner = train(length)lstm_runners[length] = runner

运行结果：

6.3.2.3 损失曲线展示

# # 画出训练过程中的损失图
for length in lengths:runner = lstm_runners[length]fig_name = f"D:/datasets/images/6.11_{length}.pdf"plot_training_loss(runner, fig_name, sample_step=100)

L=10：

L=15：

L=20：

L=25：

L=30：

L=35：

【思考题1】LSTM与SRN实验结果对比，谈谈看法。（选做）

LSTM模型在不同长度数据集上进行训练后的损失变化，同SRN模型一样，随着序列长度的增加，训练集上的损失逐渐不稳定，验证集上的损失整体趋向于变大，这说明当序列长度增加时，保持长期依赖的能力同样在逐渐变弱。LSTM模型在序列长度增加时，收敛情况比SRN模型更好，确率也要优于SRN。

6.3.3 模型评价

6.3.3.1 在测试集上进行模型评价

#lstm
lstm_dev_scores = []
lstm_test_scores = []
for length in lengths:print(f"Evaluate LSTM with data length {length}.")runner = lstm_runners[length]# 加载训练过程中效果最好的模型model_path = os.path.join(save_dir, f"best_lstm_model_{length}.pdparams")runner.load_model(model_path)# 加载长度为length的数据data_path = f"D:/datasets/{length}"train_examples, dev_examples, test_examples = load_data(data_path)test_set = DigitSumDataset(test_examples)test_loader = DataLoader(test_set, batch_size=batch_size)# 使用测试集评价模型，获取测试集上的预测准确率score, _ = runner.evaluate(test_loader)lstm_test_scores.append(score)lstm_dev_scores.append(max(runner.dev_scores))for length, dev_score, test_score in zip(lengths, lstm_dev_scores, lstm_test_scores):print(f"[LSTM] length:{length}, dev_score: {dev_score}, test_score: {test_score: .5f}")#训练SRN模型
srn_runners = {}
lengths = [10, 15, 20, 25, 30, 35]
for length in lengths:runner = train(length)srn_runners[length] = runner
srn_dev_scores = []
srn_test_scores = []
for length in lengths:print(f"Evaluate SRN with data length {length}.")runner = srn_runners[length]# 加载训练过程中效果最好的模型model_path = os.path.join(save_dir, f"best_srn_model_{length}.pdparams")runner.load_model(model_path)# 加载长度为length的数据data_path = f"D:/datasets/{length}"train_examples, dev_examples, test_examples = load_data(data_path)test_set = DigitSumDataset(test_examples)test_loader = torch.utils.data.DataLoader(test_set, batch_size=batch_size)# 使用测试集评价模型，获取测试集上的预测准确率score, _ = runner.evaluate(test_loader)srn_test_scores.append(score)srn_dev_scores.append(max(runner.dev_scores))for length, dev_score, test_score in zip(lengths, srn_dev_scores, srn_test_scores):print(f"[SRN] length:{length}, dev_score: {dev_score}, test_score: {test_score: .5f}")

运行结果：

6.3.3.2 模型在不同长度的数据集上的准确率变化图

#绘制全部图
import matplotlib.pyplot as pltplt.plot(lengths, lstm_dev_scores, '-o', color='#e8609b',  label="LSTM Dev Accuracy")
plt.plot(lengths, lstm_test_scores,'-o', color='#000000', label="LSTM Test Accuracy")#绘制坐标轴和图例
plt.ylabel("accuracy", fontsize='large')
plt.xlabel("sequence length", fontsize='large')
plt.legend(loc='lower left', fontsize='x-large')fig_name = "D:/datasets/images/6.12.pdf"
plt.savefig(fig_name)
plt.show()

【思考题2】LSTM与SRN在不同长度数据集上的准确度对比，谈谈看法。（选做）

随着数据集长度的增加，LSTM模型和SRN模型的准确率降低，但是LSTM模型的准确率显著高于SRN模型，说明LSTM模型保持长期依赖的能力要优于SRN模型。

6.3.3.3 LSTM模型门状态和单元状态的变化

import torch.nn.functional as F# 实例化模型
model = LSTM(input_size, hidden_size)
model = Model_RNN4SeqClass(model, num_digits, input_size, hidden_size, num_classes)
# 指定优化器
lr = 0.001
optimizer = torch.optim.Adam(model.parameters(),lr)
# 定义评价指标
metric = Accuracy()
# 定义损失函数
loss_fn = torch.nn.CrossEntropyLoss()
# 基于以上组件，重新实例化Runner
runner = RunnerV3(model, optimizer, loss_fn, metric)length = 10
# 加载训练过程中效果最好的模型
model_path = os.path.join(save_dir, f"best_lstm_model_{length}.pdparams")
runner.load_model(model_path)import seaborn as snsdef plot_tensor(inputs, tensor,  save_path, vmin=0, vmax=1):import matplotlib.pyplot as plttensor = np.stack(tensor, axis=0)tensor = np.squeeze(tensor, 1).Tplt.figure(figsize=(16,6))# vmin, vmax定义了色彩图的上下界ax = sns.heatmap(tensor, vmin=vmin, vmax=vmax)ax.set_xticklabels(inputs)ax.figure.savefig(save_path)# 定义模型输入
inputs = [6, 7, 0, 0, 1, 0, 0, 0, 0, 0]
X = torch.tensor(inputs.copy())
X = X.unsqueeze(0)
# 进行模型预测，并获取相应的预测结果
logits = runner.predict(X)
predict_label = torch.argmax(logits, dim=-1)
print(f"predict result: {predict_label.numpy()[0]}")
# 输入门
Is= runner.model.rnn_model.Is
plot_tensor(inputs, Is, save_path="D:/datasets/images/6.13_I.pdf")
# 遗忘门
Fs = runner.model.rnn_model.Fs
plot_tensor(inputs, Fs, save_path="D:/datasets/images/6.13_F.pdf")
# 输出门
Os = runner.model.rnn_model.Os
plot_tensor(inputs, Os, save_path="D:/datasets/images/6.13_O.pdf")
# 单元状态
Cs = runner.model.rnn_model.Cs
plot_tensor(inputs, Cs, save_path="D:/datasets/images/6.13_C.pdf", vmin=-5, vmax=5)

输出门：

遗忘门：

输入门：

单元状态：

【思考题3】分析LSTM中单元状态和门数值的变化图，并用自己的话解释该图。

横坐标为输入数字，纵坐标为相应门或单元状态向量的维度，颜色的深浅代表数值的大小。
当输入门遇到不同位置的数字0时，保持了相对一致的数值大小，表明对于0元素保持相同的门控过滤机制，避免输入信息的变化给当前模型带来困扰；
遗忘门数值在一些维度上变小，表明对某些信息进行了遗忘；
输出门和单元状态在某些维度上数值变小，在某些维度上数值变大，表明输出门在根据信息的重要性选择信息进行输出，同时单元状态也在保持着对文本预测重要的一些信息.

全面总结RNN（必做）

ref：

NNDL 实验6（上） - HBU_DAVID - 博客园 (cnblogs.com)

NNDL 实验6（下） - HBU_DAVID - 博客园
(cnblogs.com)