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

[Model][3/N] Refactor sfa into mla and remove deepseek_v3_2.py (#3769)

Browse Source 
This is the follow-up PR to PR #3189, which continues to refactor sfa
into mla and finally remove deepseek_v3_2.py. This is the last PR of
deepseek modeling refactoring. After this, all deepseek-related model
codes are removed from vllm_ascend.

FurtherMore, after this PR deepseek v3.2 can run chunk-prefill with
correct accuracy.

- vLLM version: v0.11.0rc3
- vLLM main:
83f478bb19

---------

Signed-off-by: whx-sjtu <2952154980@qq.com>
This commit is contained in:
whx
2025-10-30 17:06:38 +08:00
committed by GitHub
parent eff3e5fc6f
commit f6149f3894
10 changed files with 751 additions and 1935 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm_ascend/utils.py
View File

@@ -577,7 +577,6 @@ def register_ascend_customop(vllm_config: Optional[VllmConfig] = None):
from vllm.model_executor.custom_op import CustomOp
from vllm_ascend.models.layers.mla import AscendMultiHeadLatentAttention
from vllm_ascend.models.layers.sfa import AscendSparseFlashAttention
from vllm_ascend.ops.activation import AscendQuickGELU, AscendSiluAndMul
from vllm_ascend.ops.fused_moe.fused_moe import (AscendFusedMoE,
AscendSharedFusedMoE)
@@ -625,10 +624,7 @@ def register_ascend_customop(vllm_config: Optional[VllmConfig] = None):
mla_to_register = "MultiHeadLatentAttention" if vllm_version_is(
"0.11.0") else "MultiHeadLatentAttentionWrapper"
if vllm_config and vllm_config.model_config and vllm_config.model_config.use_mla:
AscendMLAAttentionWarrper = AscendSparseFlashAttention if hasattr(
vllm_config.model_config.hf_config,
"index_topk") else AscendMultiHeadLatentAttention
REGISTERED_ASCEND_OPS[mla_to_register] = AscendMLAAttentionWarrper
REGISTERED_ASCEND_OPS[mla_to_register] = AscendMultiHeadLatentAttention
for name, op_cls in REGISTERED_ASCEND_OPS.items():
CustomOp.register_oot(_decorated_op_cls=op_cls, name=name)