Pytorch+PyG实现GIN过程示例详解

import torch.nn.functional as F
from torch_geometric.datasets import Planetoid
from torch_geometric.utils import from_networkx, to_networkx
# 加载Cora数据集
dataset = Planetoid(root='./cora', name='Cora')
data = dataset[0]
# 将nx.Graph形式的图转换成PyG需要的格式
graph = to_networkx(data)
data = from_networkx(graph)
# 获取节点数量和特征向量维度
num_nodes = data.num_nodes
num_features = dataset.num_features
num_classes = dataset.num_classes
# 建立需要训练的节点分割数据集
data.train_mask = torch.zeros(num_nodes, dtype=torch.bool)
data.val_mask = torch.zeros(num_nodes, dtype=torch.bool)
data.test_mask = torch.zeros(num_nodes, dtype=torch.bool)
data.train_mask[:num_nodes - 1000] = True
data.test_mask[-1000:] = True
data.val_mask[num_nodes - 2000: num_nodes - 1000] = True

from torch_geometric.nn import global_mean_pool
class GIN(torch.nn.Module):
    def __init__(self, hidden_dim, num_layers):
        super(GIN, self).__init__()
        self.conv1 = GINConv(mlp=nn.Sequential(nn.Linear(num_features, hidden_dim),
                                                nn.ReLU(),
                                                nn.Linear(hidden_dim, hidden_dim)))
        self.convs = nn.ModuleList()
        for _ in range(num_layers - 1):
            self.convs.append(GINConv(mlp=nn.Sequential(nn.Linear(hidden_dim, hidden_dim),
                                                        nn.ReLU(),
                                                        nn.Linear(hidden_dim, hidden_dim))))
        self.classify = nn.Sequential(nn.Linear(hidden_dim, num_classes))
    def forward(self, data):
        x, edge_index, batch = data.x, data.edge_index, data.batch
        x = F.relu(self.conv1(x, edge_index))
        for conv in self.convs:
            x = F.relu(conv(x, edge_index))
        out = global_mean_pool(x, batch)
        return self.classify(out)

from torch_geometric.nn import GINConv, global_add_pool
# 初始化GIN并指定参数
num_layers = 5
hidden_dim = 1024
model = GIN(hidden_dim=hidden_dim, num_layers=num_layers).to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=0.001, weight_decay=1e-06)
loss_func = nn.CrossEntropyLoss()
# 开始训练
for epoch in range(100):
    model.train()
    optimizer.zero_grad()
    pred = model(train_data)
    loss = loss_func(pred[train_mask], train_labels)
    loss.backward()
    optimizer.step()
    # 在各个测试阶段检测一下准确率
    with torch.no_grad():
        model.eval()
        pred = model(test_data)
        test_loss = loss_func(pred[test_mask], test_labels).item()
        pred = pred.argmax(dim=-1, keepdim=True)
        correct = float(pred[test_mask].eq(test_labels.view(-1, 1)[test_mask]).sum().item())
        acc = correct / test_mask.sum().item()
        if epoch % 10 == 0:
            print("Epoch {:03d}, Train Loss {:.4f}, Test Loss {:.4f}, Test Acc {:.4f}".format(
                epoch, loss.item(), test_loss, acc))