[NPU] bugfix with Kimi-k2 and bge-reranker-v2 model (#17478)

Co-authored-by: amote-i <49533125+amote-i@users.noreply.github.com> Co-authored-by: cy <chenyang08056032@163.com>
2026-01-22 22:02:05 +08:00
parent a4dc432587
commit 5d299c25c0
5 changed files with 130 additions and 11 deletions
@@ -115,9 +115,10 @@ class RotaryEmbedding(MultiPlatformOp):
            cache = cache.to(dtype)

        if (
-            (not (_is_cuda or _is_npu) or self.head_size not in [64, 128, 256, 512])
+            (not (_is_cuda) or self.head_size not in [64, 128, 256, 512])
            and not (_is_cpu)
            and not (_is_xpu)
+            and not (_is_npu)
        ):
            if _is_cuda or _is_hip:
                from sgl_kernel import rotary_embedding