[NPU] adapt dsv3.2 nsa prefill context parallel (#14541)

2025-12-11 20:12:59 +08:00
parent fca8e88f28
commit 388018a5bd
8 changed files with 282 additions and 135 deletions
@@ -36,6 +36,7 @@ class TestAscendTp4Bf16(CustomTestCase):
            32,
            "--attention-backend",
            "ascend",
+            "--disable-radix-cache",
            "--cuda-graph-max-bs",
            32,
            "--tp-size",