@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 743a0b7402
Fork 0

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 17, 04:15 UTC

[Bugfix] use SoftLockFile instead of LockFile (#3578)

kota-iizuka committed Mar 24, 2024 at 03:43 UTC 743a0b74021b466088924d1a1228031bdedba896
1 file changed +1 -1
vllm/model_executor/weight_utils.py
+1 -1
@@ -34,7 +34,7 @@ def get_lock(model_name_or_path: str, cache_dir: Optional[str] = None):
34 lock_dir = cache_dir if cache_dir is not None else _vllm_filelocks_path
35 os.makedirs(os.path.dirname(lock_dir), exist_ok=True)
36 lock_file_name = model_name_or_path.replace("/", "-") + ".lock"
37 - lock = filelock.FileLock(os.path.join(lock_dir, lock_file_name))
37 + lock = filelock.SoftFileLock(os.path.join(lock_dir, lock_file_name))
38 return lock
39
40