Shuangli Li

你好 预处理代码可以参考此处：https://github.com/agave233/GeomGCL/blob/master/preprocess.py

In fact, this assert-statement is used to ensure we have removed HOH in the mol2 file, which is not necessary for the data in your case. You can just delete...

Hi, you can obtain the preprocess file here: https://github.com/agave233/GeomGCL/blob/master/preprocess.py

> Unfortunately I get the sense the authors aren't interested in having anyone actually use this tool. I've reached out here and directly to the corresponding scientist on the paper...

为了方便实现论文中message passing第一阶段的node2edge过程，我们首先基于pgl框架的特点来构建了node2node graph，然后通过node2node graph实现了node2edge的消息传递，可以参考如下代码(layers.py): ```python class EdgeAggConv(nn.Layer): def __init__(self, node_in_dim, link_in_dim, hidden_dim, activation=F.relu): super(EdgeAggConv, self).__init__() in_dim = node_in_dim * 2 + link_in_dim self.fc_agg = DenseLayer(in_dim, hidden_dim, activation=activation, bias=True) def agg_func(self,...

Hi, you can check the csv dataset of tox21 and sider, then find the column name for SMILES an labels. You just need to fill these meta information into the...

感谢指出的bug，这里应该是for mol in tqdm(mols)