chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

unified management of environment variables for vlm cuda ipc transport (#14501)

Browse Source
This commit is contained in:
Yuhao Yang
2025-12-18 12:28:06 +08:00
committed by GitHub
parent 9970ee34e8
commit 3d42b7e7b0
3 changed files with 11 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
python/sglang/srt/utils/cuda_ipc_transport_utils.py
View File

@@ -8,21 +8,15 @@ from typing import Tuple
import numpy as np
import torch
from sglang.srt.environ import envs
from sglang.srt.server_args import get_global_server_args
from sglang.srt.utils import get_float_env_var, get_int_env_var
logger = logging.getLogger(__name__)
MM_FEATURE_CACHE_SIZE = (
4 * 1024 * 1024 * 1024
if not get_int_env_var("SGLANG_MM_FEATURE_CACHE_MB")
else get_int_env_var("SGLANG_MM_FEATURE_CACHE_MB") * 1024 * 1024
)
MM_FEATURE_CACHE_SIZE = envs.SGLANG_MM_FEATURE_CACHE_MB.get() * 1024 * 1024
MM_ITEM_MEMORY_POOL_RECYCLE_INTERVAL = (
0.05
if not get_float_env_var("SGLANG_MM_ITEM_MEM_POOL_RECYCLE_INTERVAL_SEC")
else get_float_env_var("SGLANG_MM_ITEM_MEM_POOL_RECYCLE_INTERVAL_SEC")
envs.SGLANG_MM_ITEM_MEM_POOL_RECYCLE_INTERVAL_SEC.get()
)
SHM_LOCK_FILE = "/tmp/shm_wr_lock.lock"