「代码词典 · 一词一义」（四）epochs和batch_size

``` import torch import numpy as np from torch import nn, optim from torch.nn.utils.rnn import pad_sequence # 如果没有安装Jellyfish库，可以用其他方法处理序列 # 定义词典：单词到索引映射，假设有5个不同的单词 word_to_idx = { 'cat': 0, 'dog': 1, 'apple': 2, 'banana': 3, 'orange': 4 } # 加载训练数据和测试数据，这里简单模拟，假设序列长度为5 training_data = [ ['cat', 'dog', 'apple'], ['banana', 'orange'], ['cat', 'dog', 'apple', 'banana'] ] test_data = [ ['apple', 'banana', 'orange'], ['dog', 'cat'] ] # 定义模型参数 batch_size = 2 seq_len = 5 hidden_size = 128 embedding_size = 100 class SimpleLSTM(nn.Module): def __init__(self, vocab_size, hidden_size, embedding_size): super().__init__() self.embedding = nn.Embedding(vocab_size, embedding_size) self.lstm = nn.LSTM(hidden_size, hidden_size, num_layers=1) self.fc = nn.Linear(hidden_size, hidden_size) def forward(self, x): # 假设x是 tensor形式，形状为 (batch_size, seq_len) x_embed = self.embedding(x) # 形状为 (batch_size, seq_len, embedding_size) lstm_out, (hn, cn) = self.lstm(x_embed) output = self.fc(hn.mean(dim=1)) # 取最后一个时间步的输出 return output # 初始化模型 model = SimpleLSTM(len(word_to_idx), hidden_size, embedding_size) # 定义优化器 optimizer = optim.Adam(model.parameters(), lr=0.001) # 定义损失函数（交叉熵损失） criterion = nn.CrossEntropyLoss() # 训练循环 num_epochs = 10 for epoch in range(num_epochs): # 将数据转换为 tensors 并进行处理 train_sequences = [torch.tensor(s, dtype=torch.long) for s in training_data] train_labels = torch.tensor([0]*len(training_data), dtype=torch.long) sequences = pad_sequence(train_sequences, padding_value=9999) labels = torch.zeros(len(sequences)//seq_len * batch_size) # 前向传播 outputs = model(sequences[:, :seq_len]) loss = criterion(outputs, labels) # 反向传播和优化 optimizer.zero_grad() loss.backward() optimizer.step() print(f'Epoch {epoch+1}, Loss: {loss.item():.4f}') # 测试模型 test_sequences = [torch.tensor(s, dtype=torch.long) for s in test_data] test_labels = torch.tensor([0]*len(test_data), dtype=torch.long) test_sequences_padded = pad_sequence(test_sequences, padding_value=9999) with torch.no_grad(): outputs = model(test_sequences_padded[:, :seq_len]) _, predicted = torch.max(outputs.data, 1) print("Test Accuracy:", np.mean(predicted == test_labels))```代码检查

epoch与数据分割