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

[3/N][Refactor] Move torchair_attention to torchair dir (#2017)

Browse Source 
### What this PR does / why we need it?

1. Move `torchair_attention` to `torchair` dir.
2. Make `AscendAttentionTorchairBackend` extend `AscendAttentionBackend`
to reduce duplicate methods.
3. Make `AscendTorchairMetadata` extend `AscendMetadata` to reduce
duplicate properties.

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?


- vLLM version: v0.10.0
- vLLM main:
0933f9d518

---------

Signed-off-by: shen-shanshan <467638484@qq.com>
This commit is contained in:
Shanshan Shen
2025-08-19 10:25:22 +08:00
committed by GitHub
parent 2a763b8326
commit 83e0f41408
5 changed files with 23 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
vllm_ascend/worker/model_runner_v1.py
View File

@@ -78,7 +78,6 @@ from vllm_ascend.ascend_forward_context import set_ascend_forward_context
from vllm_ascend.attention.attention_mask import AttentionMaskBuilder
from vllm_ascend.attention.attention_v1 import (AscendAttentionState,
AscendMetadata)
from vllm_ascend.attention.attention_v1_torchair import AscendTorchairMetadata
from vllm_ascend.attention.mla_v1 import AscendMLAMetadata
from vllm_ascend.distributed.moe_comm_method import (AllGatherCommImpl,
DummyCommImpl,
@@ -86,6 +85,7 @@ from vllm_ascend.distributed.moe_comm_method import (AllGatherCommImpl,
from vllm_ascend.multistream.ms_split import compute_split_seq_index
from vllm_ascend.platform import NPUPlatform
from vllm_ascend.sample.rejection_sampler import AscendRejectionSampler
from vllm_ascend.torchair.torchair_attention import AscendTorchairMetadata
from vllm_ascend.utils import (ACL_FORMAT_FRACTAL_ND, ACL_FORMAT_FRACTAL_NZ,
ProfileExecuteDuration, is_310p,
maybe_converting_weight_acl_format)