Closed osllogon closed 1 year ago
I have been trying to train a temporal GNN with the following structure on WindmillOutputMediumDatasetLoader dataset:
`class RecurrentGCN(torch.nn.Module):
def __init__(self, node_features: int) -> None: # call super class constructor super().__init__() # define layers self.recurrent: torch.nn.Module = GConvLSTM(node_features, 32, 2) self.relu: torch.nn.Module = torch.nn.ReLU() self.linear: torch.nn.Module = torch.nn.Linear(32, 1) def forward(self, x: torch.Tensor, edge_index: torch.Tensor, edge_weight: torch.Tensor) -> torch.Tensor: # get outputs outputs = self.recurrent(x, edge_index, edge_weight)[0] outputs = self.relu(outputs) outputs = self.linear(outputs) return outputs`
But I find impossible to optimize. I have tried many different learning rates and optimizers. Can anyone help me with this or give me any tips? I have tried the same with other datasets from the library and I found no problem
Is there any solution or update for this?
I have been trying to train a temporal GNN with the following structure on WindmillOutputMediumDatasetLoader dataset:
`class RecurrentGCN(torch.nn.Module):
But I find impossible to optimize. I have tried many different learning rates and optimizers. Can anyone help me with this or give me any tips? I have tried the same with other datasets from the library and I found no problem