[Minor] more code cleanup (#4077)

python/sglang/srt/managers/detokenizer_manager.py

@@ -212,6 +212,7 @@ class DetokenizerManager:
             rids=recv_obj.rids,
             finished_reasons=recv_obj.finished_reasons,
             output_strs=output_strs,
+            output_ids=None,
             prompt_tokens=recv_obj.prompt_tokens,
             completion_tokens=recv_obj.completion_tokens,
             cached_tokens=recv_obj.cached_tokens,

python/sglang/srt/managers/io_struct.py

@@ -414,6 +414,12 @@ class BatchTokenIDOut:
 class BatchMultimodalDecodeReq:
     # The request id
     rids: List[str]
+    finished_reasons: List[BaseFinishReason]
+
+    # Token counts
+    prompt_tokens: List[int]
+    completion_tokens: List[int]
+    cached_tokens: List[int]
 
 
 @dataclass
@@ -424,6 +430,8 @@ class BatchStrOut:
     finished_reasons: List[dict]
     # The output decoded strings
     output_strs: List[str]
+    # The token ids
+    output_ids: Optional[List[int]]
 
     # Token counts
     prompt_tokens: List[int]
@@ -453,6 +461,15 @@ class BatchStrOut:
 class BatchMultimodalOut:
     # The request id
     rids: List[str]
+    # The finish reason
+    finished_reasons: List[dict]
+    # The outputs
+    outputs: List[List[Dict]]
+
+    # Token counts
+    prompt_tokens: List[int]
+    completion_tokens: List[int]
+    cached_tokens: List[int]
 
 
 @dataclass

python/sglang/srt/managers/tokenizer_manager.py

@@ -1141,7 +1141,7 @@ async def print_exception_wrapper(func):
 
 
 class SignalHandler:
-    def __init__(self, tokenizer_manager):
+    def __init__(self, tokenizer_manager: TokenizerManager):
         self.tokenizer_manager = tokenizer_manager
 
     def signal_handler(self, signum=None, frame=None):