[Refactor] Import global var form vllm instead of overwirte it (#5469)

### What this PR does / why we need it?
Import global var form vllm instead of overwirte it, so that we could
use the correct global variant value

- vLLM version: v0.13.0
- vLLM main:
5326c89803
---------
Signed-off-by: MengqingCao <cmq0113@163.com>
This commit is contained in:
Mengqing Cao
2026-01-07 18:41:45 +08:00
committed by GitHub
parent 380f089fbf
commit 3f4f2b4ae6
10 changed files with 7 additions and 157 deletions

View File

@@ -13,8 +13,7 @@ import torch
import torch.nn.functional as F
import triton
import triton.language as tl
PAD_SLOT_ID = -1
from vllm.attention.backends.utils import PAD_SLOT_ID
def causal_conv1d_ref(