From b2988e1b9fa46887909318515873f2448798855d Mon Sep 17 00:00:00 2001 From: Chendi Xue Date: Fri, 3 Oct 2025 15:49:32 +0000 Subject: [PATCH] remove _backend Signed-off-by: Chendi Xue --- vllm_gaudi/platform.py | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/vllm_gaudi/platform.py b/vllm_gaudi/platform.py index dcf76e6d..da17a537 100644 --- a/vllm_gaudi/platform.py +++ b/vllm_gaudi/platform.py @@ -8,10 +8,11 @@ from vllm import envs -from vllm.platforms import Platform, PlatformEnum, _Backend +from vllm.platforms import Platform, PlatformEnum from vllm_gaudi.extension.runtime import get_config if TYPE_CHECKING: + from vllm.attention.backends.registry import _Backend from vllm.config import ModelConfig, VllmConfig else: ModelConfig = None @@ -39,7 +40,7 @@ class HpuPlatform(Platform): additional_env_vars = [k for k, v in os.environ.items() if retain_envs(k)] @classmethod - def get_attn_backend_cls(cls, selected_backend: _Backend, head_size: int, dtype: torch.dtype, + def get_attn_backend_cls(cls, selected_backend: "_Backend", head_size: int, dtype: torch.dtype, kv_cache_dtype: Optional[str], block_size: int, use_v1: bool, use_mla: bool, has_sink: bool, use_sparse: bool) -> str: assert use_v1, 'Only V1 is supported!'