Unify think_end_id to model_config as single source of truth (#22148)

python/sglang/srt/constrained/base_grammar_backend.py

@@ -204,6 +204,7 @@ def create_grammar_backend(
     tokenizer,
     vocab_size: int,
     eos_token_ids: Optional[set] = None,
+    think_end_id: Optional[int] = None,
 ) -> Optional[BaseGrammarBackend]:
     name = server_args.grammar_backend
 
@@ -258,13 +259,11 @@ def create_grammar_backend(
     else:
         raise ValueError(f"Invalid grammar backend: {name}")
 
-    if server_args.reasoning_parser and hasattr(tokenizer, "think_end_id"):
+    if server_args.reasoning_parser and think_end_id is not None:
         from sglang.srt.constrained.reasoner_grammar_backend import (
             ReasonerGrammarBackend,
         )
 
-        grammar_backend = ReasonerGrammarBackend(
+        grammar_backend = ReasonerGrammarBackend(grammar_backend, think_end_id)
-            grammar_backend, tokenizer.think_end_id
-        )
 
     return grammar_backend

python/sglang/srt/constrained/grammar_manager.py

@@ -32,6 +32,7 @@ class GrammarManager:
                 scheduler.tokenizer,
                 scheduler.model_config.vocab_size,
                 scheduler.model_config.hf_eos_token_id,
+                think_end_id=scheduler.model_config.think_end_id,
             )
         else:
             self.grammar_backend = None

python/sglang/srt/managers/scheduler.py

@@ -549,13 +549,9 @@ class Scheduler(
             reasoning_parser = ReasoningParser(
                 model_type=self.server_args.reasoning_parser, stream_reasoning=False
             )
-            self.tokenizer.think_end_id = self.tokenizer.encode(
+            self.model_config.think_end_id = self.tokenizer.encode(
                 reasoning_parser.detector.think_end_token, add_special_tokens=False
             )[0]
-            self._think_end_id = self.tokenizer.think_end_id
-            self.model_config.think_end_id = self._think_end_id
-        else:
-            self._think_end_id = None
 
     def init_mamba_backend(self) -> None:
         initialize_mamba_selective_state_update_backend(self.server_args)

python/sglang/srt/managers/scheduler_output_processor_mixin.py

@@ -558,8 +558,9 @@ class SchedulerOutputProcessorMixin:
     def _maybe_update_reasoning_tokens(
         self: Scheduler, req: Req, next_token_id: Union[int, List[int]]
     ):
-        if req.require_reasoning and self._think_end_id is not None:
+        think_end_id = self.model_config.think_end_id
-            req.update_reasoning_tokens(next_token_id, self._think_end_id)
+        if req.require_reasoning and think_end_id is not None:
+            req.update_reasoning_tokens(next_token_id, think_end_id)
 
     def _mamba_prefix_cache_update(
         self: Scheduler,