diff --git a/parlai/agents/seq2seq/seq2seq.py b/parlai/agents/seq2seq/seq2seq.py
index 3e050e450fc..42a0d20c2a6 100644
--- a/parlai/agents/seq2seq/seq2seq.py
+++ b/parlai/agents/seq2seq/seq2seq.py
@@ -176,23 +176,23 @@ def build_model(self, states=None):
             print('skipping preinitialization of embeddings for bpe')
         elif not states and opt['embedding_type'] != 'random':
             # `not states`: only set up embeddings if not loading model
-            self._copy_embeddings(self.model.decoder.lt.weight, opt['embedding_type'])
+            self._copy_embeddings(model.decoder.lt.weight, opt['embedding_type'])
             if opt['lookuptable'] in ['unique', 'dec_out']:
                 # also set encoder lt, since it's not shared
                 self._copy_embeddings(
-                    self.model.encoder.lt.weight, opt['embedding_type'], log=False
+                    model.encoder.lt.weight, opt['embedding_type'], log=False
                 )
 
         if states:
             # set loaded states if applicable
-            self.model.load_state_dict(states['model'])
+            model.load_state_dict(states['model'])
 
         if opt['embedding_type'].endswith('fixed'):
             print('Seq2seq: fixing embedding weights.')
-            self.model.decoder.lt.weight.requires_grad = False
-            self.model.encoder.lt.weight.requires_grad = False
+            model.decoder.lt.weight.requires_grad = False
+            model.encoder.lt.weight.requires_grad = False
             if opt['lookuptable'] in ['dec_out', 'all']:
-                self.model.decoder.e2s.weight.requires_grad = False
+                model.output.weight.requires_grad = False
 
         return model