transformer分类预测

点击代码获取

部分代码：

import torch
import torch.nn as nn
import torch.optim as optim
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelBinarizer
from sklearn.metrics import accuracy_score
import numpy as np
from util import *


# 定义 Transformer 模型
class TransformerModel(nn.Module):
    def __init__(self, input_dim, output_dim, hidden_dim, num_heads, num_layers):
        super(TransformerModel, self).__init__()

        self.transformer_encoder = nn.TransformerEncoder(
            nn.TransformerEncoderLayer(input_dim, num_heads, hidden_dim),num_layers
        )
        self.fc = nn.Linear(input_dim, output_dim)

    def forward(self, x):
        x = x.unsqueeze(1) #维度扩展 (batch_size, 1, input_dim)
        x = self.transformer_encoder(x)
        x = x.squeeze(1)   #维度压缩  (batch_size, hidden_dim)
        x = self.fc(x)
        return x

#定义训练函数
def train_model():
    # 训练模型
    train_losses = []
    train_accs = []

    for epoch in range(num_epochs):
        model.train()
        optimizer.zero_grad()    #梯度清零
        outputs = model(X_train) # 前向传播
        loss = criterion(outputs, torch.argmax(y_train, dim=1))
        loss.backward() # 反向传播和优化
        optimizer.step()

        # 计算训练准确率
        _, predicted = torch.max(outputs, 1)
        train_acc = accuracy_score(torch.argmax(y_train, dim=1).numpy(), predicted.numpy())

        # 记录损失和准确率
        train_losses.append(loss.item())
        train_accs.append(train_acc)

        if (epoch + 1) % 10 == 0:
            print(f'Epoch {epoch + 1}/{num_epochs}, Loss: {loss.item():.4f}, Train Acc: {train_acc:.4f}')

    return train_losses,train_accs

部分数据：

训练损失

预测对比：

混淆矩阵