Fix acc mismatch for GCU (Enflame) (#330)

<!--
 Copyright 2026 FlagOS Contributors

 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
 You may obtain a copy of the License at

     http://www.apache.org/licenses/LICENSE-2.0

 Unless required by applicable law or agreed to in writing, software
 distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License.
 -->

### PR Category
<!-- One of [Core | Vendor | OP | Tools | Others] -->

### PR Type
<!-- One of [User Experience | New Features | Bug Fixes | Improvements |
Performance | Breaking Change | Deprecations | Test Case | Docs |
Others] -->

### Description
<!-- Describe what this PR does and why. -->

### Related Issues
<!-- Link any related issues: Fixes #issue, Closes #issue, or Related to
#issue -->

### Changes
<!-- List the key changes made in this PR. -->
-

### Testing
<!-- How has this change been tested? Include test commands, hardware
used, etc. -->
-
note: need flaggems commit id :c080be72c7a81772a61b3bad301cb4ec458ee6db

### Checklist
- [ ] I have run the existing tests and they pass
- [ ] I have added tests for my changes (if applicable)
- [ ] I have updated the documentation (if applicable)

---------

Co-authored-by: cyber-pioneer <116002591+cyber-pioneer@users.noreply.github.com>

Author: liuyancong-enflame-tech

vllm_fl/dispatch/backends/vendor/gcu/gcu.py

@@ -73,15 +73,11 @@ def rotary_embedding(
         )
 
     def attention_backend(self, use_mla: bool = False, use_sparse: bool = False) -> str:
-        from vllm.v1.attention.backends.registry import AttentionBackendEnum
-
         if use_mla:
             if use_sparse:
                 raise NotImplementedError("GCU does not support sparse attention yet")
             raise NotImplementedError("GCU does not support MLA yet")
-
-        import flash_attn.vllm_flash_attn
-
-        sys.modules["vllm.vllm_flash_attn"] = flash_attn.vllm_flash_attn
-
-        return AttentionBackendEnum.FLASH_ATTN.get_path()
+        # GCU uses a standalone flash_attn backend (AttentionGCUBackend) that calls
+        # Enflame's native flash_attn_varlen_func directly, without depending on
+        # vllm upstream FlashAttentionBackend / FlashAttentionImpl.
+        return "vllm_fl.dispatch.backends.vendor.gcu.impl.attention.AttentionGCUBackend"