[Core] Add fault tolerance for RayTokenizerGroupPool (#5748)
This commit is contained in:
@@ -310,6 +310,8 @@ class _AsyncLLMEngine(LLMEngine):
|
||||
)
|
||||
|
||||
async def check_health_async(self) -> None:
|
||||
if self.tokenizer:
|
||||
self.tokenizer.check_health()
|
||||
self.model_executor.check_health()
|
||||
|
||||
|
||||
|
||||
@@ -1013,6 +1013,8 @@ class LLMEngine:
|
||||
return self.model_executor.pin_lora(lora_id)
|
||||
|
||||
def check_health(self) -> None:
|
||||
if self.tokenizer:
|
||||
self.tokenizer.check_health()
|
||||
self.model_executor.check_health()
|
||||
|
||||
def is_tracing_enabled(self) -> bool:
|
||||
|
||||
Reference in New Issue
Block a user