DrugEncoder.MPG

Click here to view source code.

class MPG(ft_dim: int = 768, MPG_dim: int = 768, freeze: bool = True, conv: bool = True,
num_layer=5, emb_dim=768, heads=12, num_message_passing=3, drop_ratio=0, pt_path=None)

MPG can be used to encode graphs of drugs.

PARAMETERS:

  • ft_dim (int, optional) - Dimension of encoded node features. (default: 768)

  • MPG_dim (int, optional) - Dimension of MPG features. (default: 768)

  • freeze (bool, optional) - Freeze MPG or not. (default: True)

  • conv (bool, optional) - Use torch_geometric.nn.conv.GCNConv as output layer or not. (default: True)

  • num_layer (optional) - Parameter num_layer of MPG. (default: 5)

  • emb_dim (optional) - Parameter emb_dim of MPG. (default: 768)

  • heads (optional) - Parameter heads of MPG. (default: 12)

  • num_message_passing (optional) - Parameter num_message_passing of MPG. (default: 3)

  • drop_ratio (optional) - Parameter drop_ratio of MPG. (default: 0)

  • pt_path (optional) - Path of MolGNet.pt. (default: None)

SHAPES:

  • input: Preprocessed graphs

  • output: (Encoded node features [node_num, ft_dim], Preprocessed graphs)

forward(g)

  • g - The input of MPG.