apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 08:37:06 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature] Support block scheduler v1 for FD (#2928)

Browse Source 
* Support FD block scheduler v1

* Support FD block scheduler v1

* Support FD block scheduler v1

* Fix according to copilot review

* Fix according to review

* Remove is_dummy

* Fix bug when real_bsz=1

* Fix infer first token cost time

---------

Co-authored-by: Jiang-Jia-Jun <163579578+Jiang-Jia-Jun@users.noreply.github.com>
This commit is contained in:
chenjian
2025-07-23 20:31:31 +08:00
committed by GitHub
parent ca0f71bd39
commit 85a78d695d
16 changed files with 898 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
fastdeploy/worker/gpu_worker.py
View File

@@ -22,6 +22,7 @@ import paddle
import pynvml
from paddle import nn
from fastdeploy import envs
from fastdeploy.config import FDConfig
from fastdeploy.engine.request import Request
from fastdeploy.platforms import current_platform
@@ -183,7 +184,10 @@ class GpuWorker(WorkerBase):
TODO(gongshaotian):The scheduler should schedule the handling of prefill,
and workers and modelrunners should not perceive it.
"""
self.model_runner.insert_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.insert_prefill_inputs(req_dicts=req_dicts)
def graph_optimize_and_warm_up_model(self) -> None:
"""