fix message

modelscope/pipelines/nlp/llm_pipeline.py

@@ -370,6 +370,7 @@ class LLMPipeline(Pipeline, PipelineStreamingOutputMixin):
 
     def preprocess(self, inputs: Union[str, Dict], **kwargs):
         is_messages = kwargs.pop('is_messages')
+        print(kwargs)
         if is_messages:
             tokens = self.format_messages(inputs, self.tokenizer, **kwargs)
         else:
@@ -440,6 +441,7 @@ class LLMPipeline(Pipeline, PipelineStreamingOutputMixin):
         # for compatibility, also support input list, but we shall wrap it into Dict
         if isinstance(messages, list):
             messages = {'messages': messages}
+            kwargs['is_message'] = True
         for role, content in LLMPipeline._message_iter(messages):
             tokens = LLMPipeline._concat_with_special_tokens(
                 tokens, role, content, tokenizer)