A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 5, 16:00 UTC
[Bugfix] Lazy import diskcache to avoid sqlite3/libstdc++ ImportError at startup (#38649)
Signed-off-by: Jeffrey Wang <jeffreywang@anyscale.com>
Jeffrey Wang committed
Mar 31, 2026 at 22:31 UTC
2d725b89c5257378fc4b64b7f9134855f8c54590
1 file changed
+2
-1
vllm/v1/structured_output/utils.py
+2
-1
@@ -12,7 +12,6 @@ import numpy as np
12
import regex as re
13
import torch
14
from cachetools import LRUCache
15
-from diskcache import Cache
15
16
import vllm.envs as envs
17
from vllm.logger import init_logger
@@ -181,6 +180,8 @@ def get_outlines_cache():
180
181
cache_dir = get_outlines_cache_path()
182
if envs.VLLM_V1_USE_OUTLINES_CACHE:
183
+ from diskcache import Cache
184
+
185
logger.warning(
186
"Enabling outlines cache. This is an unbounded on-disk "
187
"cache. It may consume a lot of disk space and should "