[Bugfix] Use dedicated MM processor cache in /tokenize to prevent sender-cache pollution (#38545)

Signed-off-by: Sergey Zinchenko <sergey.zinchenko.rnd@gmail.com>
This commit is contained in:
Sergey Zinchenko
2026-04-01 21:14:49 -07:00
committed by GitHub
parent 5f96f9aff1
commit 5a2d420c17
4 changed files with 192 additions and 24 deletions
@@ -86,12 +86,14 @@ class OpenAIServingTokenization(OpenAIServing):
default_template_content_format=self.chat_template_content_format,
default_template_kwargs=self.default_chat_template_kwargs,
tool_dicts=tool_dicts,
skip_mm_cache=True,
)
else:
engine_inputs = await self.openai_serving_render.preprocess_completion(
request,
prompt_input=request.prompt,
prompt_embeds=None,
skip_mm_cache=True,
)
input_ids: list[int] = []