diff --git a/spacy/syntax/nn_parser.pyx b/spacy/syntax/nn_parser.pyx
index 653235aff..fa91c697e 100644
--- a/spacy/syntax/nn_parser.pyx
+++ b/spacy/syntax/nn_parser.pyx
@@ -827,6 +827,7 @@ cdef class Parser:
         for action, labels in actions.items():
             for label in labels:
                 self.moves.add_action(action, label)
+        cfg.setdefault('token_vector_width', 128)
         if self.model is True:
             cfg['pretrained_dims'] = self.vocab.vectors_length
             self.model, cfg = self.Model(self.moves.n_moves, **cfg)
@@ -836,11 +837,12 @@ cdef class Parser:
                     self.model[1].ops.allocate((5, cfg['token_vector_width'])))
             self.init_multitask_objectives(gold_tuples, pipeline, sgd=sgd, **cfg)
             link_vectors_to_models(self.vocab)
-            self.cfg.update(cfg)
-        elif sgd is None:
-            sgd = self.create_optimizer()
-        self.model[1].begin_training(
-            self.model[1].ops.allocate((5, cfg['token_vector_width'])))
+        else:
+            if sgd is None:
+                sgd = self.create_optimizer()
+            self.model[1].begin_training(
+                self.model[1].ops.allocate((5, cfg['token_vector_width'])))
+        self.cfg.update(cfg)
         return sgd
 
     def add_multitask_objective(self, target):
diff --git a/spacy/tests/regression/test_issue1919.py b/spacy/tests/regression/test_issue1919.py
new file mode 100644
index 000000000..ffb592b1e
--- /dev/null
+++ b/spacy/tests/regression/test_issue1919.py
@@ -0,0 +1,10 @@
+'''Test that nlp.begin_training() doesn't require missing cfg properties.'''
+from __future__ import unicode_literals
+import pytest
+from ... import load as load_spacy
+
+@pytest.mark.models('en')
+def test_issue1919():
+    nlp = load_spacy('en')
+    opt = nlp.begin_training()
+