biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[CI/Build] vLLM cache directory for images (#6444)

Browse Source
This commit is contained in:
Cyrus Leung
2024-07-16 14:12:25 +08:00
committed by GitHub
parent 37d776606f
commit d97011512e
13 changed files with 123 additions and 140 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/distributed/device_communicators/custom_all_reduce_utils.py
View File

@@ -189,10 +189,10 @@ def gpu_p2p_access_check(src: int, tgt: int) -> bool:
cuda_visible_devices = envs.CUDA_VISIBLE_DEVICES
if cuda_visible_devices is None:
cuda_visible_devices = ",".join(str(i) for i in range(num_dev))
VLLM_CONFIG_ROOT = envs.VLLM_CONFIG_ROOT
path = os.path.expanduser(
f"{VLLM_CONFIG_ROOT}/vllm/gpu_p2p_access_cache_for_{cuda_visible_devices}.json"
)
path = os.path.join(
envs.VLLM_CACHE_ROOT,
f"gpu_p2p_access_cache_for_{cuda_visible_devices}.json")
os.makedirs(os.path.dirname(path), exist_ok=True)
from vllm.distributed.parallel_state import get_world_group
if ((not is_distributed or get_world_group().local_rank == 0)