refactor: minor refactors regarding multimodal processing (#6187)

2025-05-18 13:53:20 +08:00
parent b3f3d610fd
commit 01dd39bac1
15 changed files with 140 additions and 98 deletions
--- a/python/sglang/srt/managers/schedule_batch.py
+++ b/python/sglang/srt/managers/schedule_batch.py
@@ -324,8 +324,9 @@ class MultimodalInputs:
    video_token_id: Optional[int] = None

    # audio
-    audio_start_id: Optional[torch.Tensor] = None
-    audio_end_id: Optional[torch.Tensor] = None
+    audio_token_id: Optional[int] = None
+    audio_start_id: Optional[int] = None
+    audio_end_id: Optional[int] = None

    @staticmethod
    def from_dict(obj: dict):
@@ -349,6 +350,7 @@ class MultimodalInputs:
            "slice_end_id",
            "audio_start_id",
            "audio_end_id",
+            "audio_token_id",
        ]
        for arg in optional_args:
            if arg in obj: