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

[Feature] Support XiaoMi MIMO Flash V2 (#62)

Browse Source 
* [Feature] Support MIMO Flash V2
This commit is contained in:
Xinyu Dong
2025-12-31 10:16:33 +08:00
committed by GitHub
parent 341dc7f296
commit b3c30a3cb9
12 changed files with 1530 additions and 690 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm_kunlun/ops/paged_attn.py
View File

@@ -8,14 +8,8 @@ from typing import List, Optional, Tuple
from vllm.platforms import current_platform
if current_platform.is_kunlun():
from vllm_kunlun.ops._kunlun_ops import KunlunOps as ops
else:
from vllm import _custom_ops as ops
from vllm.triton_utils.importing import HAS_TRITON
from vllm_kunlun.ops._kunlun_ops import KunlunOps as ops
if HAS_TRITON:
from vllm.attention.ops.prefix_prefill import context_attention_fwd
# Should be the same as PARTITION_SIZE in `paged_attention_v2_launcher`.
_PARTITION_SIZE = 512