apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-04 16:22:57 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Launch expert_service before kv_cache initialization in worker_process (#3045)

Browse Source 
* launch expert_service before kv_cache initialization

* add two signal make sure model loading and expert_service lauching finished

* fix the EP bug

* fix ep

* update launching way

* fix ep

* update

* roback ep

* pre-commit all files

---------

Co-authored-by: RAM <gstian5555@outlook.com>
Co-authored-by: Divano <dddivano@outlook.com>
This commit is contained in:
Zero Rains
2025-08-11 19:38:46 +08:00
committed by GitHub
parent c27a3dc43b
commit b23af29d0b
6 changed files with 175 additions and 100 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
fastdeploy/output/token_processor.py
View File

@@ -296,12 +296,14 @@ class TokenProcessor:
else:
batch = self.output_tokens[1, 0]
tokens = tokens[2 : batch + 2]
batch_result = list()
if envs.ENABLE_V1_KVCACHE_SCHEDULER:
need_to_be_reschedule_req_ids = list(self.resource_manager.to_be_rescheduled_request_id_set)
for request_id in need_to_be_reschedule_req_ids:
if self.resource_manager.requests[request_id].idx >= (batch - 1): # No more token generated for preempted request
if self.resource_manager.requests[request_id].idx >= (
batch - 1
): # No more token generated for preempted request
self.resource_manager.reschedule_preempt_task(request_id)
for i in range(batch):
if self.resource_manager.stop_flags[i]: