Update inference.py, for more general use.

inference.py

@@ -61,8 +61,8 @@ class DualBertModel: # use this class to encode your document and queries.
         with torch.no_grad():
             to_return = []
             for c in self.chunks(corpus, batch_size):
-                to_encode = [d['text'] for d in c]
-                ids, atm = self.process(to_encode, False)
+                # to_encode = [d['text'] for d in c]
+                ids, atm = self.process(c, False)
                 out = self.p_model(ids, atm)
                 to_return.append(out.cpu())