How to correct a validation loss in a regression problem?

I've developed a graph neural network using PyTorch Geometric.

My model looks like:

class GCN(torch.nn.Module):

  def __init__(self):
    super().__init__()
    self.conv1 = GCNConv(3, 32)
    self.conv2 = GCNConv(32, 64)
    self.conv3 = GCNConv(64, 32)
    self.fc1 = Linear(32, 10)

  def forward(self, data):
    x, edge_index, edge_attr = data.x, data.edge_index, data.edge_attr
    x = F.elu(self.conv1(x, edge_index))
    x = F.elu(self.conv2(x, edge_index))
    x = F.elu(self.conv3(x, edge_index))
    x = F.elu(self.fc1(x))
    return x

I've generated a train and validation dataset, with the following learning curves:

I do not know what kind of actions I may take to correct the validation loss. Now, I understand that the model overfits the data, but in a very weird way.

What's happening? Any suggestion?