[Frontend] Refactor ServingTokenization entrypoint. (#46022)

Signed-off-by: wang.yuqi <yuqi.wang@daocloud.io>
This commit is contained in:
wang.yuqi
2026-06-22 06:27:58 +00:00
committed by GitHub
parent 31124749d1
commit 1eb2cc961e
15 changed files with 322 additions and 302 deletions
@@ -42,6 +42,10 @@ class OpenAIModelRegistry:
) -> None:
self.model_config = model_config
self.base_model_paths = base_model_paths
self.lora_requests: dict[str, LoRARequest] = {}
def model_name(self, lora_request: LoRARequest | None = None) -> str:
return self.base_model_paths[0].name
def is_base_model(self, model_name: str) -> bool:
return any(model.name == model_name for model in self.base_model_paths)
@@ -72,6 +76,9 @@ class OpenAIModelRegistry:
]
)
async def resolve_lora(self, lora_name: str):
raise RuntimeError("The OpenAIModelRegistry has no LoRA support.")
class OpenAIServingModels:
"""Shared instance to hold data about the loaded base model(s) and adapters.