diff --git a/spacy/cli/train_from_config.py b/spacy/cli/train_from_config.py
index 95fdb10d5..7dc6143f2 100644
--- a/spacy/cli/train_from_config.py
+++ b/spacy/cli/train_from_config.py
@@ -231,8 +231,8 @@ def train(
         # check whether the setting 'exclusive_classes' corresponds to the provided training data
         if textcat_multilabel:
             multilabel_found = False
-            for ex in corpus.train_examples:
-                cats = ex.doc_annotation.cats
+            for eg in corpus.train_annotations:
+                cats = eg.reference.cats
                 textcat_labels.update(cats.keys())
                 if list(cats.values()).count(1.0) != 1:
                     multilabel_found = True
@@ -244,8 +244,8 @@ def train(
                     "mutually exclusive classes more accurately."
                 )
         else:
-            for ex in corpus.train_examples:
-                cats = ex.doc_annotation.cats
+            for eg in corpus.train_annotations:
+                cats = eg.reference.cats
                 textcat_labels.update(cats.keys())
                 if list(cats.values()).count(1.0) != 1:
                     msg.fail(
@@ -346,10 +346,8 @@ def train(
                 progress = tqdm.tqdm(total=training["eval_frequency"], leave=False)
             # Clean up the objects to faciliate garbage collection.
             for eg in batch:
-                eg.doc = None
-                eg.goldparse = None
-                eg.doc_annotation = None
-                eg.token_annotation = None
+                eg.reference = None
+                eg.predicted = None
     except Exception as e:
         msg.warn(
             f"Aborting and saving the final best model. "
diff --git a/spacy/pipeline/simple_ner.py b/spacy/pipeline/simple_ner.py
index 692c74a38..3ef6a48ce 100644
--- a/spacy/pipeline/simple_ner.py
+++ b/spacy/pipeline/simple_ner.py
@@ -143,8 +143,7 @@ def _has_ner(eg):
 def _get_labels(examples):
     labels = set()
     for eg in examples:
-        for ner_tag in eg.token_annotation.entities:
+        for ner_tag in eg.get_aligned("ENT_TYPE", as_string=True):
             if ner_tag != 'O' and ner_tag != '-':
-                _, label = ner_tag.split('-', 1)
-                labels.add(label)
+                labels.add(ner_tag)
     return list(sorted(labels))