reward can use same batch check
This commit is contained in:
@@ -21,9 +21,6 @@ class BTChatTemplateStrategy(ChatTemplateStrategy):
|
||||
Bradley-Terry reward model pairwise chat template prompt strategy.
|
||||
"""
|
||||
|
||||
def is_prompt_batched(self, prompt: dict[str, Any]) -> bool:
|
||||
return all(isinstance(v, list) for v in prompt.values())
|
||||
|
||||
def _tokenize_single_prompt(self, prompt):
|
||||
"""
|
||||
|
||||
|
||||
Reference in New Issue
Block a user