Revert "[feature] Rework Ascend NPU graph support" (#9385)

This commit is contained in:
Even Zhou
2025-08-20 15:35:10 +08:00
committed by GitHub
parent 1ec9769753
commit de2dd73831
18 changed files with 81 additions and 546 deletions

View File

@@ -376,7 +376,7 @@ class MHATokenToKVPool(KVCache):
v_scale: Optional[float] = None,
layer_id_override: Optional[int] = None,
):
from sglang.srt.model_executor.graph_runner import get_is_capture_mode
from sglang.srt.model_executor.cuda_graph_runner import get_is_capture_mode
if layer_id_override is not None:
layer_id = layer_id_override