GRU(门控循环单元)的原理与代码实现

1.GRU的原理

1.1重置门和更新门

1.2候选隐藏状态

1.3隐状态

2. GRU的代码实现

#导包
import torch
from torch import nn
import dltools#加载数据
batch_size, num_steps = 32, 35
train_iter, vocab = dltools.load_data_time_machine(batch_size, num_steps)#封装函数：实现初始化模型参数
def get_params(vocab_size, num_hiddens, device):num_inputs = num_outputs = vocab_sizedef normal(shape):return torch.randn(size=shape, device=device) * 0.01def three():return (normal((num_inputs, num_hiddens)),normal((num_hiddens, num_hiddens)),torch.zeros(num_hiddens, device=device))# 更新门参数W_xz, W_hz, b_z = three()# 重置门W_xr, W_hr, b_r = three()# 候选隐藏状态参数W_xh, W_hh, b_h = three()# 输出层参数W_hq = normal((num_hiddens, num_outputs))b_q = torch.zeros(num_outputs, device=device)params = [W_xz, W_hz, b_z, W_xr, W_hr, b_r, W_xh, W_hh, b_h, W_hq, b_q]for param in params:param.requires_grad_(True)return params#定义函数：初始化隐藏状态
def init_gru_state(batch_size, num_hiddens, device):return (torch.zeros((batch_size, num_hiddens), device=device))#定义函数：构建GRU网络结构
def gru(inputs, state, params):[W_xz, W_hz, b_z, W_xr, W_hr, b_r, W_xh, W_hh, b_h, W_hq, b_q] = paramsH, = stateoutputs = []for X in inputs:Z = torch.sigmoid((X @ W_xz) + (H @ W_hz) + b_z)R = torch.sigmoid((X @ W_xr) + (H @ W_hr) + b_r)H_tilda = torch.tanh((X @ W_xh) + ((R * H) @ W_hh) + b_h)H = Z * H + (1 - Z) * H_tildaY = H @ W_hq + b_qoutputs.append(Y)return torch.cat(outputs, dim=0), (H, )#训练和预测
vocab_size, num_hiddens, device = len(vocab), 256, dltools.try_gpu()
num_epochs, lr = 500, 5
model = dltools.RNNModelScratch(len(vocab), num_hiddens, device, get_params, init_gru_state, gru)
dltools.train_ch8(model, train_iter, vocab, lr, num_epochs, device)

3.pytorch 简洁实现版_GRU调包实现

num_inputs = vocab_size
#创建网络层
gru_layer = nn.GRU(num_inputs, num_hiddens)
#建模
model = dltools.RNNModel(gru_layer, len(vocab))
#将模型转到device上
model = model.to(device)
#模型训练
dltools.train_ch8(model, train_iter, vocab, lr, num_epochs, device)