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

Use SGLANG_CACHE_DIR env for gpu_p2p_access_cache path (#25686)

Browse Source 
Co-authored-by: Ian O'Connell <ianoc@meta.com>
Co-authored-by: ianoc <ianoc@fb.com>
This commit is contained in:
Lianmin Zheng
2026-05-18 16:00:54 -07:00
committed by GitHub
co-authored by Ian O'Connell ianoc
parent 745abd6cc0
commit 314dedf7c6
1 changed files with 3 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/distributed/device_communicators/custom_all_reduce_utils.py
+3 -2
View File
@@ -21,6 +21,7 @@ from typing_extensions import ParamSpec
from sglang.srt.distributed.device_communicators.cuda_wrapper import CudaRTLibrary from sglang.srt.distributed.device_communicators.cuda_wrapper import CudaRTLibrary
from sglang.srt.distributed.parallel_state import in_the_same_node_as from sglang.srt.distributed.parallel_state import in_the_same_node_as
from sglang.srt.environ import envs as sglang_envs
from sglang.srt.utils import is_cuda, is_hip, is_musa from sglang.srt.utils import is_cuda, is_hip, is_musa
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
@@ -261,8 +262,8 @@ def gpu_p2p_access_check(src: int, tgt: int) -> bool:
cuda_visible_devices = ",".join(str(i) for i in range(num_dev)) cuda_visible_devices = ",".join(str(i) for i in range(num_dev))
# VLLM_CACHE_ROOT -> SGLANG_CACHE_ROOT # VLLM_CACHE_ROOT -> SGLANG_CACHE_ROOT
# "~/.cache/vllm" -> "~/.cache/sglang" # "~/.cache/vllm" -> envs.SGLANG_CACHE_DIR
SGLANG_CACHE_ROOT = os.path.expanduser("~/.cache/sglang") SGLANG_CACHE_ROOT = os.path.expanduser(sglang_envs.SGLANG_CACHE_DIR.get())
path = os.path.join( path = os.path.join(
SGLANG_CACHE_ROOT, f"gpu_p2p_access_cache_for_{cuda_visible_devices}.json" SGLANG_CACHE_ROOT, f"gpu_p2p_access_cache_for_{cuda_visible_devices}.json"
) )