Bugs for training

[mgao97]

Hi, I have encountered a problem when I run the code following the readme.

The problem shows _"Traceback (most recent call last):
File "/users/Min/HyperGCL/src/train_auto.py", line 903, in
out = model.forward_cl(data_aug1)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/users/Min/HyperGCL/src/models.py", line 718, in forward_cl
x = F.relu(self.V2EConvs[i](x, edge_index, norm, self.aggr, aug_weight))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/users/Min/miniconda/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/users/Min/miniconda/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1527, in call_impl
return forward_call(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/users/Min/HyperGCL/src/layers.py", line 640, in forward
x = self.propagate(edge_index, x=x, norm=norm, aggr=aggr, aug_weight=aug_weight)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/users/Min/miniconda/lib/python3.11/site-packages/torch_geometric/nn/conv/message_passing.py", line 480, in propagate
out = self.aggregate(out, **aggr_kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/users/Min/HyperGCL/src/layers.py", line 665, in aggregate
raise ValueError("aggr was not passed!")
ValueError: aggr was not passed!"

Your response is very helpful to me!

[YESLUP]

嗨，我在按照自述文件运行代码时遇到了问题。

问题显示 _“回溯（最近一次调用最后一次）：文件”/users/Min/HyperGCL/src/train_auto.py“，第 903 行，输入输出 = model.forward_cl（data_aug1） ^^^^^^^^^^^^^^^^^^^^^^^^^^^ 文件”/users/Min/HyperGCL/src/models.py“，第 718 行，在forward_cl x = F.relu（self.V2EConvs[i]（x， edge_index， norm， self.aggr， aug_weight）） ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 文件“/users/Min/miniconda/lib/python3.11/site-packages/torch/nn/modules/module.py”，第 1518 行，在 _wrapped_call_impl 返回 self.call_impl（*args， **kwargs） ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 文件“/users/Min/miniconda/lib/python3.11/site-packages/torch/nn/modules/module.py”，第 1527 行，在_call_impl返回 forward_call（*args， **kwargs） ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 文件 “/users/Min/HyperGCL/src/layers.py”，第 640 行，向前 x = self.propagate（edge_index， x=x， norm=norm， aggr=aggr， aug_weight=aug_weight） ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 文件 “/users/Min/miniconda/lib/python3.11/site-packages/torch_geometric/nn/conv/message_passing.py”，行480，in propagate out = self.aggregate（out， **aggr_kwargs） ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File “/users/Min/HyperGCL/src/layers.py”，第 665 行，在聚合中引发 ValueError（“aggr was not passed！”） ValueError：未传递 aggr！

你的回答对我非常有帮助！

I also had the same issue, how did you end up solving this problem？

[yyou1996]

@weitianxin Hi Tianxin, please help on this.

[weitianxin]

Thank you for your interest in our work, and I apologize for the late reply. I also tested the code from our GitHub repository and it ran smoothly in my environment. It seems the issue you're experiencing might stem from a version mismatch of torch geometric. Our code was developed in an environment with python==3.6.13, torch==1.10.2, and torch-geometric==2.0.3. I will also include a file with the details of our environment setup (requirements.txt) and a log from a successful run using the cora dataset (example_log.txt) for your reference (the running speed is currently slow due to high usage). I suggest setting up a new virtual environment and installing these specific versions to replicate our experiments. Hope you all the best.