EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Refactor]7/N Extract common code to common_cp (#5490)

Browse Source 
RFC: https://github.com/vllm-project/vllm-ascend/issues/4629
Reason：
Eliminate duplicate code for two file(mla_cp.py attention_cp.py) to
common_cp.py.

vLLM version: 0.13.0rc3
vLLM main:
ad32e3e19c

vLLM version: release/v0.13.0
vLLM main:
5fbfa8d9ef

- vLLM version: v0.13.0
- vLLM main:
5326c89803

---------

Signed-off-by: wujinyuan1 <wjy9595@qq.com>
Signed-off-by: wujinyuan1 <wujinyuan1@huawei.com>
Co-authored-by: wujinyuan1 <wjy9595@qq.com>
This commit is contained in:
wujinyuan1
2026-01-05 17:41:12 +08:00
committed by GitHub
parent 755caeb06e
commit 4a3663327b
10 changed files with 252 additions and 301 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
vllm_ascend/attention/attention_v1.py
View File

@@ -34,10 +34,10 @@ from vllm.v1.attention.backends.utils import (AttentionCGSupport,
from vllm.v1.core.sched.output import SchedulerOutput
from vllm.v1.kv_cache_interface import AttentionSpec
from vllm_ascend.attention.context_parallel.common_cp import (
AscendMetadataForDecode, AscendMetadataForPrefill)
from vllm_ascend.attention.utils import (AscendCommonAttentionMetadata,
AscendMetadataForDecode,
AscendMetadataForPrefill, enable_cp,
split_decodes_and_prefills,
enable_cp, split_decodes_and_prefills,
using_paged_attention)
from vllm_ascend.compilation.acl_graph import (
get_draft_graph_params, get_graph_params,
@@ -63,7 +63,7 @@ class AscendAttentionBackend(AttentionBackend):
@staticmethod
def get_impl_cls() -> Type["AscendAttentionBackendImpl"]:
if enable_cp():
from vllm_ascend.attention.attention_cp import \
from vllm_ascend.attention.context_parallel.attention_cp import \
AscendAttentionCPImpl
return AscendAttentionCPImpl
return AscendAttentionBackendImpl
@@ -71,7 +71,7 @@ class AscendAttentionBackend(AttentionBackend):
@staticmethod
def get_builder_cls() -> type["AscendAttentionMetadataBuilder"]:
if enable_cp():
from vllm_ascend.attention.attention_cp import \
from vllm_ascend.attention.context_parallel.attention_cp import \
AscendAttentionCPMetadataBuilder
return AscendAttentionCPMetadataBuilder
return AscendAttentionMetadataBuilder