Files
enginex-ascend-910-vllm/vllm_ascend/distributed/kv_transfer/utils/mooncake_transfer_engine.py
Sun Ruoxi 7f8a1b1f7a init v0.23.0
Signed-off-by: Sun Ruoxi <sunruoxi@4paradigm.com>
2026-08-27 15:11:51 +08:00

44 lines
2.0 KiB
Python

import threading
class GlobalTE:
def __init__(self):
self.transfer_engine = None
self.is_register_buffer: bool = False
self.transfer_engine_lock = threading.Lock()
self.register_buffer_lock = threading.Lock()
def get_transfer_engine(self, hostname: str, device_name: str | None):
if self.transfer_engine is None:
with self.transfer_engine_lock:
# Double-Checked Locking
if self.transfer_engine is None:
try:
from mooncake.engine import TransferEngine # type: ignore
except ImportError as e:
raise ImportError(
"Please install mooncake by following the instructions at "
"https://github.com/kvcache-ai/Mooncake/blob/main/doc/en/build.md " # noqa: E501
"to run vLLM with MooncakeConnector."
) from e
self.transfer_engine = TransferEngine()
device_name = device_name if device_name is not None else ""
ret_value = self.transfer_engine.initialize(hostname, "P2PHANDSHAKE", "ascend", device_name)
if ret_value != 0:
raise RuntimeError(f"TransferEngine initialization failed with ret_value: {ret_value}")
return self.transfer_engine
def register_buffer(self, ptrs: list[int], sizes: list[int]):
with self.register_buffer_lock:
assert self.transfer_engine is not None, "Transfer engine must be initialized"
if self.is_register_buffer:
return
for ptr, size in zip(ptrs, sizes):
ret_value = self.transfer_engine.register_memory(ptr, size)
if ret_value != 0:
raise RuntimeError("Mooncake memory registration failed.")
self.is_register_buffer = True
global_te = GlobalTE()