Update app.py

app.py

@@ -141,10 +141,11 @@ def encode_queries(dataset_name, postfix):
         batch_dict = {k: v.to(model.device) for k, v in batch_dict.items()}
 
         with torch.cuda.amp.autocast():
-            outputs = model(**batch_dict)
-            embeds = pool(outputs.last_hidden_state, batch_dict['attention_mask'])
-            embeds = F.normalize(embeds, p=2, dim=-1)
-            encoded_embeds.append(embeds.cpu().numpy())
+            with torch.no_grad():
+                outputs = model(**batch_dict)
+                embeds = pool(outputs.last_hidden_state, batch_dict['attention_mask'])
+                embeds = F.normalize(embeds, p=2, dim=-1)
+                encoded_embeds.append(embeds.cpu().numpy())
 
     return np.concatenate(encoded_embeds, axis=0)