()
| 79 | |
| 80 | |
| 81 | def main(): |
| 82 | device = torch.device('cuda') |
| 83 | path = osp.join(osp.expanduser('~'), 'datasets') |
| 84 | dataset = Planetoid(path, name='Cora', transform=T.NormalizeFeatures()) |
| 85 | data = dataset[0].to(device) |
| 86 | |
| 87 | aug1 = A.Identity() |
| 88 | aug2 = A.PPRDiffusion(alpha=0.2) |
| 89 | gconv1 = GConv(input_dim=dataset.num_features, hidden_dim=512, num_layers=2).to(device) |
| 90 | gconv2 = GConv(input_dim=dataset.num_features, hidden_dim=512, num_layers=2).to(device) |
| 91 | encoder_model = Encoder(encoder1=gconv1, encoder2=gconv2, augmentor=(aug1, aug2), hidden_dim=512).to(device) |
| 92 | contrast_model = DualBranchContrast(loss=L.JSD(), mode='G2L').to(device) |
| 93 | |
| 94 | optimizer = Adam(encoder_model.parameters(), lr=0.001) |
| 95 | |
| 96 | with tqdm(total=200, desc='(T)') as pbar: |
| 97 | for epoch in range(1, 201): |
| 98 | loss = train(encoder_model, contrast_model, data, optimizer) |
| 99 | pbar.set_postfix({'loss': loss}) |
| 100 | pbar.update() |
| 101 | |
| 102 | test_result = test(encoder_model, data) |
| 103 | print(f'(E): Best test F1Mi={test_result["micro_f1"]:.4f}, F1Ma={test_result["macro_f1"]:.4f}') |
| 104 | |
| 105 | |
| 106 | if __name__ == '__main__': |
no test coverage detected