Support Multi Process Tokenizer Manager(#6555) (#8964)

Signed-off-by: ybyang <ybyang7@iflytek.com>
Signed-off-by: huanglong <huanglong@linux.alibaba.com>
Co-authored-by: Huang Long <121648372+LLLL114@users.noreply.github.com>
Co-authored-by: huanglong <huanglong@linux.alibaba.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
This commit is contained in:
ybyang
2025-09-01 16:00:13 +08:00
committed by GitHub
parent 4750cddf68
commit 5f77e1292d
11 changed files with 1030 additions and 80 deletions

View File

@@ -983,6 +983,11 @@ class AbortReq:
abort_all: bool = False
# The finished reason data
finished_reason: Optional[Dict[str, Any]] = None
# used in MultiTokenzierManager mode
rids: Optional[Union[List[str], str]] = None
def __post_init__(self):
self.rids = self.rid
@dataclass
@@ -1183,6 +1188,18 @@ class LoRAUpdateResult:
LoadLoRAAdapterReqOutput = UnloadLoRAAdapterReqOutput = LoRAUpdateResult
@dataclass
class MultiTokenizerRegisterReq:
rids: Optional[Union[List[str], str]] = None
ipc_name: Optional[str] = None
@dataclass
class MultiTokenizerWarpper:
worker_id: int
obj: Optional[Any] = None
class BlockReqType(Enum):
BLOCK = 1
UNBLOCK = 2