diff --git a/spacy/pipeline.pyx b/spacy/pipeline.pyx
index 8c3ecacd6..41aefdd59 100644
--- a/spacy/pipeline.pyx
+++ b/spacy/pipeline.pyx
@@ -160,6 +160,7 @@ class TokenVectorEncoder(BaseThincComponent):
         self.vocab = vocab
         self.doc2feats = doc2feats()
         self.model = model
+        self.cfg = dict(cfg)
 
     def __call__(self, doc):
         """Add context-sensitive vectors to a `Doc`, e.g. from a CNN or LSTM
@@ -239,9 +240,10 @@ class TokenVectorEncoder(BaseThincComponent):
 
 class NeuralTagger(BaseThincComponent):
     name = 'tagger'
-    def __init__(self, vocab, model=True):
+    def __init__(self, vocab, model=True, **cfg):
         self.vocab = vocab
         self.model = model
+        self.cfg = dict(cfg)
 
     def __call__(self, doc):
         tags = self.predict([doc.tensor])
@@ -416,10 +418,18 @@ class NeuralTagger(BaseThincComponent):
 
 class NeuralLabeller(NeuralTagger):
     name = 'nn_labeller'
-    def __init__(self, vocab, model=True):
+    def __init__(self, vocab, model=True, **cfg):
         self.vocab = vocab
         self.model = model
-        self.labels = {}
+        self.cfg = dict(cfg)
+
+    @property
+    def labels(self):
+        return self.cfg.get('labels', {})
+
+    @labels.setter
+    def labels(self, value):
+        self.cfg['labels'] = value
 
     def set_annotations(self, docs, dep_ids):
         pass
@@ -478,9 +488,10 @@ class SimilarityHook(BaseThincComponent):
     Where W is a vector of dimension weights, initialized to 1.
     """
     name = 'similarity'
-    def __init__(self, vocab, model=True):
+    def __init__(self, vocab, model=True, **cfg):
         self.vocab = vocab
         self.model = model
+        self.cfg = dict(cfg)
 
     @classmethod
     def Model(cls, length):
@@ -527,7 +538,7 @@ class TextCategorizer(BaseThincComponent):
     def __init__(self, vocab, model=True, **cfg):
         self.vocab = vocab
         self.model = model
-        self.cfg = cfg
+        self.cfg = dict(cfg)
 
     @property
     def labels(self):