Compatibility wrapper for HPC API changes.
Users of vLLM should always import only these wrappers.
Functions:
-
has_hpc – Return True if hpc package is available.
has_hpc() cached
Return True if hpc package is available.
Source code in vllm/utils/hpc.py
| @functools.cache
def has_hpc() -> bool:
"""Return `True` if hpc package is available."""
# Use find_spec to check if the module exists without importing it
# This avoids potential CUDA initialization side effects
if importlib.util.find_spec("hpc") is None:
logger.warning_once(
"HPC attention requires the hpc module to be installed. "
"Please install it from https://github.com/Tencent/hpc-ops"
)
return False
return True
|