add learnable finite difference and fix __init__.py

ThermalSolver/model/finite_difference.py

@@ -15,21 +15,21 @@ class FiniteDifferenceStep(MessagePassing):
             aggr == "add"
         ), "Per somme pesate, l'aggregazione deve essere 'add'."
         self.root_weight = float(root_weight)
+        self.p = torch.nn.Parameter(torch.tensor(0.5))
 
-    def forward(self, x, edge_index, edge_attr, deg, weight=1.0):
+    def forward(self, x, edge_index, edge_attr, deg):
         """
         TODO: add docstring.
         """
-        out = self.propagate(
-            edge_index, x=x, edge_attr=edge_attr, deg=deg, weight=weight
-        )
+        out = self.propagate(edge_index, x=x, edge_attr=edge_attr, deg=deg)
         return out
 
     def message(self, x_j, edge_attr):
         """
         TODO: add docstring.
         """
-        return edge_attr.view(-1, 1) * x_j
+        p = torch.clamp(self.p, 0.0, 1.0)
+        return p * edge_attr.view(-1, 1) * x_j
 
     def aggregate(self, inputs, index, deg):
         """
@@ -39,12 +39,11 @@ class FiniteDifferenceStep(MessagePassing):
         deg = deg + 1e-7
         return out / deg.view(-1, 1)
 
-    def update(self, aggr_out, x, weight):
+    def update(self, aggr_out, x):
         """
         TODO: add docstring.
         """
-        print(weight)
-        return weight * aggr_out + (1 - weight) * x
+        return self.root_weight * aggr_out + (1 - self.root_weight) * x
 
 
 class GraphFiniteDifference(nn.Module):
@@ -94,13 +93,27 @@ class GraphFiniteDifference(nn.Module):
         c_ij = self._compute_c_ij(c, edge_index)
         edge_attr = edge_attr * c_ij
         deg = self._compute_deg(edge_index, edge_attr, x.size(0))
-        conv_thres = self.threshold * torch.norm(x)
-        weight = 1.0
+
+        # Calcola la soglia staccando x dal grafo
+        conv_thres = self.threshold * torch.norm(x.detach())
+
         for _i in range(self.max_iters):
-            out = self.fd_step(x, edge_index, edge_attr, deg, weight=weight)
-            weight = weight * 0.9999
+            out = self.fd_step(x, edge_index, edge_attr, deg)
             out[boundary_mask] = boundary_values.unsqueeze(-1)
-            if torch.norm(out - x) < conv_thres:
+
+            # Controllo convergenza senza tracciamento gradienti
+            with torch.no_grad():
+                residual_norm = torch.norm(out - x)
+
+            if residual_norm < conv_thres:
                 break
-            x = out
+
+            # --- OTTIMIZZAZIONE CHIAVE ---
+            # Stacca 'out' dal grafo prima della prossima iterazione
+            # per evitare BPTT e risparmiare memoria.
+            x = out.detach()
+
+        # Il 'out' finale restituito mantiene i gradienti
+        # dell'ULTIMA chiamata a fd_step, permettendo al modello
+        # di apprendere correttamente.
         return out, _i + 1