mirror of
https://github.com/PaddlePaddle/FastDeploy.git
synced 2025-10-05 08:37:06 +08:00
[XPU] Support kvblock centralized management (#3017)
This commit is contained in:
@@ -20,6 +20,7 @@ from typing import List, Optional
|
||||
import paddle
|
||||
from paddle import nn
|
||||
|
||||
from fastdeploy import envs
|
||||
from fastdeploy.config import FDConfig
|
||||
from fastdeploy.engine.request import Request
|
||||
from fastdeploy.utils import get_logger
|
||||
@@ -154,7 +155,10 @@ class XpuWorker(WorkerBase):
|
||||
TODO(gongshaotian):The scheduler should schedule the handling of prefill,
|
||||
and workers and modelrunners should not perceive it.
|
||||
"""
|
||||
self.model_runner.process_prefill_inputs(req_dicts=req_dicts)
|
||||
if envs.ENABLE_V1_KVCACHE_SCHEDULER:
|
||||
self.model_runner.insert_tasks_v1(req_dicts=req_dicts)
|
||||
else:
|
||||
self.model_runner.process_prefill_inputs(req_dicts=req_dicts)
|
||||
|
||||
def check_health(self) -> bool:
|
||||
""" """
|
||||
|
Reference in New Issue
Block a user