diff --git a/vllm/attention/backends/utils.py b/vllm/attention/backends/utils.py index 3c5028a66d585..e8a34434122c4 100644 --- a/vllm/attention/backends/utils.py +++ b/vllm/attention/backends/utils.py @@ -612,5 +612,5 @@ def flash_attn_version(): return fa_version VLLM_FLASH_ATTN_VERSION = flash_attn_version() -except ImportError: +except (ImportError, AssertionError): VLLM_FLASH_ATTN_VERSION = None