address PR comments

besskge/scoring.py

@@ -965,8 +965,8 @@ def __init__(
         sharding: Sharding,
         n_relation_type: int,
         embedding_size: int,
-        embedding_width: int,
         embedding_height: int,
+        embedding_width: int,
         entity_initializer: Union[torch.Tensor, List[Callable[..., torch.Tensor]]] = [
             init_xavier_norm,
             torch.nn.init.zeros_,
@@ -977,8 +977,8 @@ def __init__(
         inverse_relations: bool = True,
         input_channels: int = 1,
         output_channels: int = 32,
-        kernel_width: int = 3,
         kernel_height: int = 3,
+        kernel_width: int = 3,
         input_dropout: float = 0.2,
         feature_map_dropout: float = 0.2,
         hidden_dropout: float = 0.3,
@@ -995,12 +995,12 @@ def __init__(
             Number of relation types in the knowledge graph.
         :param embedding_size:
             Size of entity and relation embeddings.
-        :param embedding_width:
-            Width of the 2D-reshaping of the concatenation of
-            head and relation embeddings.
         :param embedding_height:
             Height of the 2D-reshaping of the concatenation of
             head and relation embeddings.
+        :param embedding_width:
+            Width of the 2D-reshaping of the concatenation of
+            head and relation embeddings.
         :param entity_initializer:
             Initialization functions or table for entity embeddings.
             If not passing a table, two functions are needed: the initializer
@@ -1013,10 +1013,10 @@ def __init__(
             Number of input channels of the Conv2D operator. Default: 1.
         :param output_channels:
             Number of output channels of the Conv2D operator. Default: 32.
-        :param kernel_width:
-            Width of the Conv2D kernel. Default: 3.
         :param kernel_height:
             Height of the Conv2D kernel. Default: 3.
+        :param kernel_width:
+            Width of the Conv2D kernel. Default: 3.
         :param input_dropout:
             Rate of Dropout applied before the convolution. Default: 0.2.
         :param feature_map_dropout:
@@ -1056,23 +1056,23 @@ def __init__(
         self.embedding_size = embedding_size
 
         self.inp_channels = input_channels
-        self.emb_w = embedding_width
         self.emb_h = embedding_height
+        self.emb_w = embedding_width
         conv_layers = [
             torch.nn.Dropout(input_dropout),
             torch.nn.Conv2d(
                 in_channels=self.inp_channels,
                 out_channels=output_channels,
-                kernel_size=(kernel_width, kernel_height),
+                kernel_size=(kernel_height, kernel_width),
             ),
             torch.nn.ReLU(),
             torch.nn.Dropout2d(feature_map_dropout),
         ]
         fc_layers = [
             torch.nn.Linear(
                 output_channels
-                * (2 * self.emb_w - kernel_width + 1)
-                * (self.emb_h - kernel_height + 1),
+                * (2 * self.emb_h - kernel_height + 1)
+                * (self.emb_w - kernel_width + 1),
                 embedding_size,
             ),
             torch.nn.Dropout(hidden_dropout),
@@ -1100,8 +1100,8 @@ def score_triple(
         tail_emb, tail_bias = torch.split(tail_emb, self.embedding_size, dim=-1)
         hr_cat = torch.cat(
             [
-                head_emb.view(-1, self.inp_channels, self.emb_w, self.emb_h),
-                relation_emb.view(-1, self.inp_channels, self.emb_w, self.emb_h),
+                head_emb.view(-1, self.inp_channels, self.emb_h, self.emb_w),
+                relation_emb.view(-1, self.inp_channels, self.emb_h, self.emb_w),
             ],
             dim=-2,
         )
@@ -1115,30 +1115,7 @@ def score_heads(
         relation_id: torch.Tensor,
         tail_emb: torch.Tensor,
     ) -> torch.Tensor:
-        relation_emb = torch.index_select(
-            self.relation_embedding, index=relation_id, dim=0
-        )
-        # Discard bias for heads
-        head_emb = head_emb[..., :-1]
-        tail_emb, tail_bias = torch.split(tail_emb, self.embedding_size, dim=-1)
-        if self.negative_sample_sharing:
-            head_emb = head_emb.view(1, -1, self.embedding_size)
-        broadcast_dim = (relation_emb.shape[0], head_emb.shape[1], self.embedding_size)
-        hr_cat = torch.cat(
-            [
-                head_emb.expand(*broadcast_dim).view(
-                    -1, self.inp_channels, self.emb_w, self.emb_h
-                ),
-                relation_emb.unsqueeze(1)
-                .expand(*broadcast_dim)
-                .view(-1, self.inp_channels, self.emb_w, self.emb_h),
-            ],
-            dim=-2,
-        )
-        hr_conv = self.fc_layers(self.conv_layers(hr_cat).flatten(start_dim=1)).view(
-            *broadcast_dim
-        )
-        return self.reduce_embedding(hr_conv * tail_emb.unsqueeze(1)) + tail_bias
+        raise NotImplementedError("ConvE should not be used with head corruption")
 
     # docstr-coverage: inherited
     def score_tails(
@@ -1159,8 +1136,8 @@ def score_tails(
             tail_bias = tail_bias.squeeze(-1)
         hr_cat = torch.cat(
             [
-                head_emb.view(-1, self.inp_channels, self.emb_w, self.emb_h),
-                relation_emb.view(-1, self.inp_channels, self.emb_w, self.emb_h),
+                head_emb.view(-1, self.inp_channels, self.emb_h, self.emb_w),
+                relation_emb.view(-1, self.inp_channels, self.emb_h, self.emb_w),
             ],
             dim=-2,
         )