Implementing inference functions

refactor/models/pl_gru.py

@@ -110,9 +110,13 @@ class RecurrentModel(pl.LightningModule):
     def encode(self, lX, l_pad, batch_size=128):
         """
         Returns encoded data (i.e, RNN hidden state at second feed-forward layer - linear1). Dimensionality is 512.
+        # TODO: does not run on gpu..
         :param lX:
+        :param l_pad:
+        :param batch_size:
         :return:
         """
+        with torch.no_grad():
             l_embed = {lang: [] for lang in lX.keys()}
             for lang in sorted(lX.keys()):
                 for i in range(0, len(lX[lang]), batch_size):
@@ -135,7 +139,7 @@ class RecurrentModel(pl.LightningModule):
                     output = self.dropout(F.relu(self.linear1(output)))
                     l_embed[lang].append(output)
             for k, v in l_embed.items():
-            l_embed[k] = torch.cat(v, dim=0)
+                l_embed[k] = torch.cat(v, dim=0).cpu().numpy()
             return l_embed
 
     def training_step(self, train_batch, batch_idx):

refactor/view_generators.py

@@ -229,7 +229,7 @@ class RecurrentGen(ViewGen):
         l_pad = self.multilingualIndex.l_pad()
         data = self.multilingualIndex.l_devel_index()
         # trainer = Trainer(gpus=self.gpus)
-        # self.model.eval()
+        self.model.eval()
         time_init = time()
         l_embeds = self.model.encode(data, l_pad, batch_size=256)
         transform_time = round(time() - time_init, 3)