Merge branch 'add_steerlm_links' of github.com:NVIDIA/NeMo-Aligner in…

…to steerlm_update
NVIDIA · Dec 2, 2023 · 7640597 · 7640597
2 parents 564ae7e + 68d75ff
commit 7640597
Showing 1 changed file with 9 additions and 1 deletion.
diff --git a/examples/nlp/data/steerlm/preprocess_openassistant_data.py b/examples/nlp/data/steerlm/preprocess_openassistant_data.py
@@ -86,8 +86,16 @@ def get_data_records(objs, mask_role, type):
             if len(conversations) <= 1:
                 # remove single turn conversations
                 continue
+
+            # mask out labels from user turns
+            updated_conversation = []
+            for turn in conversations:
+                if turn["from"] == "User":
+                    turn["label"] = None
+                updated_conversation.append(turn)
+
             conversation_obj = {
-                "conversations": conversations,
+                "conversations": updated_conversation,
                 "system": system_prompt,
                 "mask": mask_role,
                 "type": type,