update utils

Configuration/config.ini

@@ -26,7 +26,7 @@ dropout=0.1
 sinkhorn_iters=3
 
 [MODEL_TRAINING]
-batch_size=16
+batch_size=32
 epoch=30
 seed_val=42
 learning_rate=2e-4

Linker/Linker.py

@@ -257,7 +257,8 @@ class Linker(Module):
                 logits, sentences_embedding = self.Supertagger.forward(batch_sentences_tokens, batch_sentences_mask)
 
                 # Run the kinker on the categories predictions
-                logits_predictions = self(batch_atoms, batch_pos_idx, batch_neg_idx, sentences_embedding, batch_sentences_mask)
+                logits_predictions = self(batch_atoms, batch_pos_idx, batch_neg_idx, sentences_embedding,
+                                          batch_sentences_mask)
 
                 linker_loss = self.cross_entropy_loss(logits_predictions, batch_true_links)
                 # Perform a backward pass to calculate the gradients.
@@ -366,9 +367,10 @@ class Linker(Module):
         :param positional_ids:
             A List of batch_size elements, each being a List of num_atoms LongTensors.
             Each LongTensor in positional_ids[b][a] indexes the location of atoms of type a in sentence b.
-        :param device:
+        :param atom_type:
         :return:
         """
 
-        return [[bsd_tensor.select(0, index=i).index_select(0, index=atom.to(self.device)) for atom in sentence]
-                for i, sentence in enumerate(positional_ids[atom_map_redux[atom_type]])]
\ No newline at end of file
+        return torch.stack([torch.stack([bsd_tensor.select(0, index=i).select(0,index=int(atom))
+                 if atom != -1 else torch.zeros(self.dim_embedding_atoms) for atom in sentence])
+                for i, sentence in enumerate(positional_ids[:, atom_map_redux[atom_type], :])])