Add --messages-until-role to truncate messages for regeneration

generate-responses.py

@@ -184,6 +184,7 @@ def main(
     attn_implementation: str = "paged|sdpa",
     subset: Optional[str] = None,
     split: str = "train",
+    messages_until_role: Optional[str] = None,
     skip_long_prompts: bool = True,
     max_samples: Optional[int] = None,
     hf_token: Optional[str] = None,
@@ -289,6 +290,14 @@ def main(
     for i, example in enumerate(dataset):
         if use_messages:
             messages = example[messages_column]
+            # Optionally truncate messages up to last occurrence of a given role
+            if messages_until_role:
+                truncated = []
+                for msg in messages:
+                    truncated.append(msg)
+                    if msg["role"] == messages_until_role:
+                        last_truncated = list(truncated)
+                messages = last_truncated
             prompt = tokenizer.apply_chat_template(
                 messages, tokenize=False, add_generation_prompt=True
             )
@@ -451,6 +460,12 @@ Examples:
             type=str,
             help="Column containing plain text prompts (alternative to --messages-column)",
         )
+        parser.add_argument(
+            "--messages-until-role",
+            type=str,
+            help="Truncate messages up to the last message with this role. "
+            "Use 'user' to strip existing assistant responses and regenerate them.",
+        )
         parser.add_argument(
             "--output-column",
             type=str,
@@ -539,6 +554,7 @@ Examples:
             model_id=args.model_id,
             messages_column=args.messages_column,
             prompt_column=args.prompt_column,
+            messages_until_role=args.messages_until_role,
             output_column=args.output_column,
             temperature=args.temperature,
             top_p=args.top_p,