From 6360eb1deab4e118116f563fc3d39167c3d6e09e Mon Sep 17 00:00:00 2001
From: wangxiyuan <wangxiyuan1007@gmail.com>
Date: Tue, 2 Dec 2025 13:15:47 +0800
Subject: [PATCH] Revert "[Bugfix] Fix Qwen2.5-Omni-7B accuarcy test (#4556)"
 (#4619)

This reverts commit 71e9b379c84d4d68764fcda4ef833e21884f8341. It breaks vllm-ascend/Qwen3-30B-A3B-W8A8 test
---
 vllm_ascend/ops/layernorm.py           | 6 +++---
 vllm_ascend/ops/register_custom_ops.py | 8 ++------
 2 files changed, 5 insertions(+), 9 deletions(-)

diff --git a/vllm_ascend/ops/layernorm.py b/vllm_ascend/ops/layernorm.py
index 8dad11c2..da5051c0 100644
--- a/vllm_ascend/ops/layernorm.py
+++ b/vllm_ascend/ops/layernorm.py
@@ -108,13 +108,13 @@ class AscendRMSNorm(RMSNorm):
         residual: Optional[torch.Tensor] = None,
     ) -> Union[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]]:
         import torch_npu
+
         if residual is not None:
             residual = torch.ops.vllm.maybe_chunk_residual(x, residual)
             assert x.size(0) == residual.size(0)
-            next_need_quant_fusion_linear = getattr(
-                self, 'next_need_quant_fusion_linear', None)
             x, residual = _addrmsnorm_forward_oot(
-                self, x, residual, next_need_quant_fusion_linear, self.bias)
+                self, x, residual, self.next_need_quant_fusion_linear,
+                self.bias)
             return x, residual
         x, residual = torch_npu.npu_rms_norm(x, self.weight,
                                              self.variance_epsilon)
diff --git a/vllm_ascend/ops/register_custom_ops.py b/vllm_ascend/ops/register_custom_ops.py
index 7c7fd6f0..03bea554 100644
--- a/vllm_ascend/ops/register_custom_ops.py
+++ b/vllm_ascend/ops/register_custom_ops.py
@@ -173,9 +173,7 @@ def _maybe_prefetch_mlp_down_proj_impl(x_dependency: torch.Tensor) -> None:
     except AssertionError:
         return
 
-    prefetch_mlp_enabled = getattr(forward_context, 'prefetch_mlp_enabled',
-                                   False)
-    if not prefetch_mlp_enabled:
+    if not forward_context.prefetch_mlp_enabled:
         return
     forward_context.prefetch_mlp_down_proj = True
     model_instance = forward_context.model_instance
@@ -204,9 +202,7 @@ def _maybe_wait_prefetch_done_impl(x: torch.Tensor) -> None:
     except AssertionError:
         return
 
-    prefetch_mlp_enabled = getattr(forward_context, 'prefetch_mlp_enabled',
-                                   False)
-    if not prefetch_mlp_enabled:
+    if not forward_context.prefetch_mlp_enabled:
         return
     if forward_context.prefetch_mlp_gate_up_proj or \
         forward_context.prefetch_mlp_down_proj: