biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[core] Add tags parameter to wake_up() (#15500)

Browse Source 
Signed-off-by: Eric <erictang000@gmail.com>
This commit is contained in:
Eric Tang
2025-04-02 01:59:27 -07:00
committed by GitHub
parent 90969fb39a
commit ddb94c2605
18 changed files with 144 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/v1/engine/async_llm.py
View File

@@ -424,8 +424,8 @@ class AsyncLLM(EngineClient):
async def sleep(self, level: int = 1) -> None:
await self.engine_core.sleep_async(level)
async def wake_up(self) -> None:
await self.engine_core.wake_up_async()
async def wake_up(self, tags: Optional[list[str]] = None) -> None:
await self.engine_core.wake_up_async(tags)
async def is_sleeping(self) -> bool:
return await self.engine_core.is_sleeping_async()

4
vllm/v1/engine/core.py
View File

@@ -264,8 +264,8 @@ class EngineCore:
def sleep(self, level: int = 1):
self.model_executor.sleep(level)
def wake_up(self):
self.model_executor.wake_up()
def wake_up(self, tags: Optional[list[str]] = None):
self.model_executor.wake_up(tags)
def is_sleeping(self) -> bool:
return self.model_executor.is_sleeping

16
vllm/v1/engine/core_client.py
View File

@@ -92,7 +92,7 @@ class EngineCoreClient(ABC):
def sleep(self, level: int = 1) -> None:
raise NotImplementedError
def wake_up(self) -> None:
def wake_up(self, tags: Optional[list[str]] = None) -> None:
raise NotImplementedError
def is_sleeping(self) -> bool:
@@ -141,7 +141,7 @@ class EngineCoreClient(ABC):
async def sleep_async(self, level: int = 1) -> None:
raise NotImplementedError
async def wake_up_async(self) -> None:
async def wake_up_async(self, tags: Optional[list[str]] = None) -> None:
raise NotImplementedError
async def is_sleeping_async(self) -> bool:
@@ -206,8 +206,8 @@ class InprocClient(EngineCoreClient):
def sleep(self, level: int = 1) -> None:
self.engine_core.sleep(level)
def wake_up(self) -> None:
self.engine_core.wake_up()
def wake_up(self, tags: Optional[list[str]] = None) -> None:
self.engine_core.wake_up(tags)
def is_sleeping(self) -> bool:
return self.engine_core.is_sleeping()
@@ -520,8 +520,8 @@ class SyncMPClient(MPClient):
def sleep(self, level: int = 1) -> None:
self.call_utility("sleep", level)
def wake_up(self) -> None:
self.call_utility("wake_up")
def wake_up(self, tags: Optional[list[str]] = None) -> None:
self.call_utility("wake_up", tags)
def is_sleeping(self) -> bool:
return self.call_utility("is_sleeping")
@@ -647,8 +647,8 @@ class AsyncMPClient(MPClient):
async def sleep_async(self, level: int = 1) -> None:
await self.call_utility_async("sleep", level)
async def wake_up_async(self) -> None:
await self.call_utility_async("wake_up")
async def wake_up_async(self, tags: Optional[list[str]] = None) -> None:
await self.call_utility_async("wake_up", tags)
async def is_sleeping_async(self) -> bool:
return await self.call_utility_async("is_sleeping")

4
vllm/v1/engine/llm_engine.py
View File

@@ -245,8 +245,8 @@ class LLMEngine:
def sleep(self, level: int = 1):
self.engine_core.sleep(level)
def wake_up(self):
self.engine_core.wake_up()
def wake_up(self, tags: Optional[list[str]] = None):
self.engine_core.wake_up(tags)
def is_sleeping(self) -> bool:
return self.engine_core.is_sleeping()