apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[fix] fix ep group all-reduce (#4140)

Browse Source 
* [fix] fix ep group all-reduce

* [fix] fix clear/update lock not working when workers > 1

* [chore] add preemption triggered info log

* [fix] fix code style

* fix model_weights_signal (#4092)

* fix model_weights_signal

---------

Co-authored-by: Yuanle Liu <yuanlehome@163.com>
This commit is contained in:
李泳桦
2025-09-18 10:34:49 +08:00
committed by GitHub
parent cffde70949
commit 0fa28b1068
6 changed files with 41 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
fastdeploy/entrypoints/engine_client.py
View File

@@ -16,12 +16,12 @@
import inspect
import os
import threading
import time
import traceback
import uuid
import numpy as np
from filelock import FileLock
from fastdeploy import envs
from fastdeploy.config import ModelConfig
@@ -132,7 +132,7 @@ class EngineClient:
pid, max_connections=int(os.getenv("FD_DEALER_CONNECTIONS", 50))
)
self.connection_initialized = False
self.clear_update_lock = threading.Lock()
self.clear_update_lock = FileLock(f"/tmp/fd_weight_clear_update_lock__pid{pid}_port{port}.lock")
def create_zmq_client(self, model, mode):
"""
@@ -351,7 +351,9 @@ class EngineClient:
if self.model_weights_status_signal.value[0] == ModelWeightsStatus.NORMAL:
return True, ""
if self.model_weights_status_signal.value[0] == ModelWeightsStatus.UPDATING:
return False, "updating model weight already"
return False, "worker is updating model weight already"
if self.model_weights_status_signal.value[0] == ModelWeightsStatus.CLEARING:
return False, "worker is clearing model weight, cannot update now"
self.model_weights_status_signal.value[0] = ModelWeightsStatus.UPDATING
if self.enable_prefix_caching or self.enable_splitwise:
@@ -395,7 +397,9 @@ class EngineClient:
if self.model_weights_status_signal.value[0] == ModelWeightsStatus.CLEARED:
return True, ""
if self.model_weights_status_signal.value[0] == ModelWeightsStatus.CLEARING:
return False, "clearing model weight already"
return False, "worker is clearing model weight already"
if self.model_weights_status_signal.value[0] == ModelWeightsStatus.UPDATING:
return False, "worker is updating model weight, cannot clear now"
self.model_weights_status_signal.value[0] = ModelWeightsStatus.CLEARING
if self.enable_prefix_caching or self.enable_splitwise: