[WIP] Fix node_embeddings grad

* Previously, no grad propagated to the node_embeddings
* Model still doesn't seem to train properly, loss is too noisy.

experiments/reproduction.py

@@ -13,7 +13,7 @@ def train():
     device = "gpu"
     num_workers = 0  # NOTE Set to 0 for single thread debugging!
 
-    model = GraphWavenet(adaptive_embedding_dim=64, n_nodes=207, k_diffusion_hops=1)
+    model = GraphWavenet(adaptive_embedding_dim=None, n_nodes=207, k_diffusion_hops=3)
     plmodule = GWnetForecasting(args, model, missing_value=0.0)
 
     tb_logger = pl_loggers.TensorBoardLogger(save_dir="logs/")
@@ -28,7 +28,7 @@ def train():
     scaler = TrafficStandardScaler.from_dataset(dataset, n_samples=30000)
     plmodule.scaler = scaler
     ## TODO Parametrise.
-    loader = DataLoader(dataset, batch_size=32, num_workers=num_workers)
+    loader = DataLoader(dataset, batch_size=8, num_workers=num_workers)
 
     trainer.fit(model=plmodule, train_dataloaders=loader)
 

src/gwnet/model/gwnet.py

@@ -3,6 +3,7 @@
 
 import torch
 import torch.nn.functional as F
+from torch.nn.parameter import Parameter
 from torch_geometric.data import Data
 from torch_geometric.utils import to_dense_adj
 
@@ -75,7 +76,7 @@ def __init__(
         """
         super().__init__()
 
-        self.id_linear = torch.nn.Linear(in_channels, out_channels, bias=bias)
+        # self.id_linear = torch.nn.Linear(in_channels, out_channels, bias=bias)
 
         # Create one GCN per hop and diffusion direction.
         gcn_dict = {}
@@ -97,6 +98,11 @@ def __init__(
                     in_channels, out_channels, bias=False, node_dim=0
                 )
 
+        if bias:
+            self.bias = Parameter(torch.randn(out_channels))
+        else:
+            self.register_parameter("bias", None)
+
         self.gcns = torch.nn.ModuleDict(gcn_dict)
 
     def forward(self, x: Data, cached_params: dict[str, torch.Tensor]) -> Data:
@@ -115,6 +121,11 @@ def forward(self, x: Data, cached_params: dict[str, torch.Tensor]) -> Data:
                 x.x.transpose(1, 2), adj_index, cached_params["adj_weights"][adj_name]
             ).transpose(1, 2)
 
+        x.x = out_sum
+
+        if self.bias is not None:
+            x.x += self.bias.view(1, -1, 1)
+
         return x
 
 
@@ -161,7 +172,6 @@ def forward(self, x: Data, cached_adj: dict[str, torch.Tensor]) -> Data:
         # TCN works on the features alone, and handles the (N,C,L) shape
         # internally.
         tcn_out = self.tcn(x)
-
         return self.gcn(tcn_out, cached_adj)
 
 
@@ -208,9 +218,6 @@ def __init__(
             self.register_parameter("node_embeddings", None)
             adp = False
 
-        # TODO Why is node_embeddings not registering?
-        # import pdb; pdb.set_trace()
-
         # This model accepts the entire road network, hence can cache
         # the diffusion adjacency matrices, doesn't have to take their
         # powers on every forward pass.
@@ -383,7 +390,8 @@ def forward(self, batch: Data) -> torch.Tensor:
                 )
             )
 
-        self._update_adp_adj(batch.batch.max() + 1, self.global_elements["k_hops"])
+        if self.node_embeddings is not None:
+            self._update_adp_adj(batch.batch.max() + 1, self.global_elements["k_hops"])
 
         # x_dict = batch.x_dict
         # edge_index_dict = batch.edge_index_dict

src/gwnet/model/layer/tempo_gcn.py

@@ -41,6 +41,10 @@ def __init__(  # type: ignore[no-untyped-def]
 
         self.linear = torch.nn.Linear(in_channels, out_channels, bias=bias)
 
+    def message(self, x_j: torch.Tensor, edge_weight: torch.Tensor) -> torch.Tensor:
+        # Message is the original input times the edge weight.
+        return x_j * edge_weight.view(-1, 1, 1)
+
     def forward(
         self,
         x: torch.Tensor,