在使用GCN网络以及数据集为MUTAG，IMDB-B和PTC-MR进行图分类任务时产生的问题

[Raymond]

🐛 Bug

To Reproduce

Steps to reproduce the behavior:

在我使用GCN网络以及数据集为MUTAG，IMDB-B和PTC-MR进行图分类任务时，都会产生以下问题
experiment(dataset="mutag", model="gcn", dw="graph_classification_dw", mw="graph_classification_mw",batch_size=16)

MUTAG---ValueError: Expected input batch_size (308) to match target batch_size (16).
IMDB-B---ValueError: Expected input batch_size (289) to match target batch_size (16).
PTC-MR---ValueError: Expected input batch_size (227) to match target batch_size (16).

在我更改batch_size超参数时前面的数字会变得更大，且后面的数字会不一样
experiment(dataset="mutag", model="gcn", dw="graph_classification_dw", mw="graph_classification_mw",batch_size=308)
ValueError: Expected input batch_size (1668) to match target batch_size (94).

Expected behavior

正常运行

Environment

• CogDL version:0.5.3
• OS (e.g., Linux):Windows 10
• Python version:3.8.13
• PyTorch version:1.12.1
• CUDA/cuDNN version:11.6
• Software:Pycharm

[Raymond]

其中报错为：

Namespace(activation='relu', actnn=False, batch_size=308, checkpoint_path='./checkpoints/model.pt', cpu=False, cpu_inference=False, dataset=['mutag'], degree_node_features=False, devices=[0], distributed=False, do_test=True, do_valid=True, dropout=0.5, dw='graph_classification_dw', epochs=500, eval_step=1, fp16=False, hidden_size=64, load_emb_path=None, local_rank=0, log_path='.', logger=None, lr=0.01, master_addr='localhost', master_port=13425, max_epoch=None, model=['gcn'], mw='graph_classification_mw', n_trials=3, n_warmup_steps=0, norm=None, nstage=1, num_classes=None, num_features=None, num_layers=2, patience=100, progress_bar='epoch', project='cogdl-exp', residual=False, resume_training=False, return_model=False, rp_ratio=1, save_emb_path=None, seed=[1], split=[0], test_ratio=0.3, train_ratio=0.5, unsup=False, use_best_config=False, weight_decay=0)

|-------------------------------------------------------------------------------------------|

*** Running (mutag, gcn, graph_classification_dw, graph_classification_mw)
Model Parameters: 642
0%
warnings.warn(f'Error checking compiler version for {compiler}: {error}')
��Ϣ: ���ṩ��ģʽ�޷��ҵ��ļ���
C:\Users\Administrator\Anaconda\envs\pytorch_gpu\lib\site-packages\torch\utils\cpp_extension.py:346: UserWarning: Error checking compiler version for cl: [WinError 2] 系统找不到指定的文件。
warnings.warn(f'Error checking compiler version for {compiler}: {error}')
��Ϣ: ���ṩ��ģʽ�޷��ҵ��ļ���
C:\Users\Administrator\Anaconda\envs\pytorch_gpu\lib\site-packages\torch\utils\cpp_extension.py:346: UserWarning: Error checking compiler version for cl: [WinError 2] 系统找不到指定的文件。
warnings.warn(f'Error checking compiler version for {compiler}: {error}')
��Ϣ: ���ṩ��ģʽ�޷��ҵ��ļ���
C:\Users\Administrator\Anaconda\envs\pytorch_gpu\lib\site-packages\torch\utils\cpp_extension.py:346: UserWarning: Error checking compiler version for cl: [WinError 2] 系统找不到指定的文件。
warnings.warn(f'Error checking compiler version for {compiler}: {error}')
��Ϣ: ���ṩ��ģʽ�޷��ҵ��ļ���
��Ϣ: ���ṩ��ģʽ�޷��ҵ��ļ���
��Ϣ: ���ṩ��ģʽ�޷��ҵ��ļ���
0%
Traceback (most recent call last):
File "D:/cogdl-master/myGNN.py", line 5, in

experiment(dataset="mutag", model="gcn", dw="graph_classification_dw", mw="graph_classification_mw",batch_size=308)

File "D:\cogdl-master\cogdl\experiments.py", line 360, in experiment
return raw_experiment(args)
File "D:\cogdl-master\cogdl\experiments.py", line 264, in raw_experiment
results = [train(args) for args in variant_args_generator(args, variants)]
File "D:\cogdl-master\cogdl\experiments.py", line 264, in
results = [train(args) for args in variant_args_generator(args, variants)]
File "D:\cogdl-master\cogdl\experiments.py", line 218, in train
result = trainer.run(model_wrapper, dataset_wrapper)
File "D:\cogdl-master\cogdl\trainer\trainer.py", line 204, in run
self.train(self.devices[0], model_w, dataset_w)
File "D:\cogdl-master\cogdl\trainer\trainer.py", line 361, in train
training_loss = self.train_step(model_w, train_loader, optimizers, lr_schedulers, rank, scaler)
File "D:\cogdl-master\cogdl\trainer\trainer.py", line 515, in train_step
loss = model_w.on_train_step(batch)
File "D:\cogdl-master\cogdl\wrappers\model_wrapper\base_model_wrapper.py", line 73, in on_train_step
return self.train_step(*args, **kwargs)
File "D:\cogdl-master\cogdl\wrappers\model_wrapper\graph_classification\graph_classification_mw.py", line 15, in train_step
loss = self.default_loss_fn(pred, y)
File "D:\cogdl-master\cogdl\utils\evaluator.py", line 121, in call
return torch.nn.functional.nll_loss(y_pred, y_true)
File "C:\Users\Administrator\Anaconda\envs\pytorch_gpu\lib\site-packages\torch\nn\functional.py", line 2689, in nll_loss
return torch._C._nn.nll_loss_nd(input, target, weight, _Reduction.get_enum(reduction), ignore_index)
ValueError: Expected input batch_size (1668) to match target batch_size (94).

Process finished with exit code 1

[cenyk1230]

Hi @zhanghaoxuan184 ,

感谢对cogdl的关注。您能否先测试一下gin模型能否使用？gcn模型目前主要用于节点分类任务

[Raymond]

gin模型可以使用，非常感谢您！请问cogdl当中的哪些模型可以进行图分类任务

[cenyk1230]

GNN模型应该是这些：["gin", "patchy_san", "diffpool", "infograph", "dgcnn", "sortpool"]
embedding模型应该是这些：["dgk","graph2vec"]

[Raymond]

好的，非常感谢！！