apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Optimize] Improve perf for fd response token with internal adapter (#4991)

Browse Source 
* [Optimize] Improve perf for fd response token with internal adapter

* fix
This commit is contained in:
chenjian
2025-11-13 16:19:48 +08:00
committed by GitHub
parent c599268f57
commit db5d421aa3
9 changed files with 171 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
fastdeploy/splitwise/splitwise_connector.py
View File

@@ -23,7 +23,7 @@ from typing import Dict
import zmq
from fastdeploy import envs
from fastdeploy.engine.request import CompletionOutput, Request, RequestOutput
from fastdeploy.engine.request import Request, RequestOutput
from fastdeploy.inter_communicator import EngineWorkerQueue
from fastdeploy.metrics.metrics import main_process_metrics
from fastdeploy.utils import get_logger
@@ -505,19 +505,5 @@ class SplitwiseConnector:
"""
tasks = []
for task in payload:
tasks.append(
RequestOutput(
request_id=task["request_id"],
outputs=CompletionOutput(
index=task["outputs"]["index"],
send_idx=0,
token_ids=task["outputs"]["token_ids"],
draft_token_ids=task["outputs"]["draft_token_ids"],
),
finished=True,
num_cached_tokens=task["num_cached_tokens"],
error_code=task["error_code"],
error_msg=task["error_msg"],
)
)
tasks.append(RequestOutput.from_dict(task))
self.engine_worker_queue.put_disaggregated_tasks(("decode", tasks))