INNER CODE UNIT · Python
get_libfabric_version
awslabs/awsome-distributed-ai · validation/efa-versions.py:21
def get_libfabric_version(container=[]):
try:
version = subprocess.check_output(container + ['fi_info', '--version'])
version = version.decode('utf-8')
version = re.search(r'libfabric: (\d+\.\d+\.\d+)', version).group(1)
return version
except Exception as e:
print(f'Error: {e}')
return None
def get_nccl_version(container=[]):
try:
version = subprocess.check_output(container + ['locate', 'nccl'])
version = version.decode('utf-8')
print(version)
version = re.search(r'/usr/local/cuda-12.8/lib/libnccl.so.(\d+\.\d+\.\d+)', version).group(1)
return version
except Exception as e: