FastMAE: Graph Neural Network-Based Fast Magnetocrystalline Anisotropy Energy Predictor

import torch
import torch.nn as nn
from torch_geometric.nn import GATConv, global_mean_pool

class FastMAE(nn.Module):
    def __init__(self, num_features=92, hidden_dim=128, num_heads=4):
        super().__init__()
        self.conv1 = GATConv(num_features, hidden_dim, heads=num_heads)
        self.conv2 = GATConv(hidden_dim * num_heads, hidden_dim, heads=num_heads)
        self.conv3 = GATConv(hidden_dim * num_heads, hidden_dim, heads=1)
        self.fc1 = nn.Linear(hidden_dim, 64)
        self.fc2 = nn.Linear(64, 1)
        self.dropout = nn.Dropout(0.2)
        
    def forward(self, data):
        x, edge_index, batch = data.x, data.edge_index, data.batch
        x = torch.relu(self.conv1(x, edge_index))
        x = self.dropout(x)
        x = torch.relu(self.conv2(x, edge_index))
        x = self.dropout(x)
        x = torch.relu(self.conv3(x, edge_index))
        x = global_mean_pool(x, batch)
        x = torch.relu(self.fc1(x))
        x = self.fc2(x)
        return x