add ignore chat dataset flag

turbo_alignment/dataset/chat/chat.py

@@ -258,7 +258,14 @@ def _encode(
         inference: bool,
         random_cut: bool,
     ) -> list[dict[str, Any] | None]:
-        conversations = [Conversation(system_prompt=self.source.system_prompt, messages=r.messages) for r in records]
+        conversations = [
+            Conversation(
+                system_prompt=self.source.system_prompt,
+                messages=r.messages,
+                ignore_system_prompt=self.settings.ignore_system_prompt,
+            )
+            for r in records
+        ]
 
         logger.info(f'Tokenizing dataset {self.source.name}')
         tokenized_replicas = self.__tokenize([m.content for c in conversations for m in c.messages])

turbo_alignment/dataset/chat/conversation.py

@@ -9,15 +9,20 @@ def __init__(
         self,
         system_prompt: str | None,
         messages: list[ChatMessage],
+        ignore_system_prompt: bool,
     ):
         self._messages: list[ChatMessage] = []
 
-        if system_prompt:
-            self._messages += [ChatMessage(role=ChatMessageRole.SYSTEM, content=system_prompt, disable_loss=True)]
+        if system_prompt and ignore_system_prompt:
+            raise ValueError('You can not ignore system_prompt and use system prompt in config simultaneoulsy')
 
+        if system_prompt or ignore_system_prompt:
             if messages[0].role == ChatMessageRole.SYSTEM:
                 messages = messages[1:]
 
+            if system_prompt:
+                self._messages += [ChatMessage(role=ChatMessageRole.SYSTEM, content=system_prompt, disable_loss=True)]
+
         self._messages += messages
 
     @property

turbo_alignment/settings/datasets/chat.py

@@ -25,6 +25,7 @@ class ChatDatasetSettings(BaseDatasetSettings):
     keep_end: bool | None = None
     max_tokens_count: int | None
     prompt_template: ChatPromptTemplate
+    ignore_system_prompt: bool = False
 
 
 class ChatMultiDatasetSettings(ChatDatasetSettings, MultiDatasetSettings):