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

Remove VLLM_USE_V1 (#4086)

Browse Source 
Drop VLLM_USE_V1 usage.  This env has been removed from vLLM already.

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

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-11-11 15:43:39 +08:00
committed by GitHub
parent d5567680a2
commit f811a24bf0
21 changed files with 7 additions and 63 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
vllm_ascend/torchair/models/qwen2.py
View File

@@ -21,7 +21,6 @@ from typing import Any, List, Optional, Union
import torch
import torch.nn.functional as F
import vllm
import vllm.envs as envs
from torch import nn
from transformers import Qwen2Config
from vllm.attention import AttentionMetadata, AttentionType
@@ -112,12 +111,9 @@ class CustomQwen2Attention(Qwen2Attention):
is_prefill=False,
is_qwen_torchair=True)
forward_kwargs = {}
if envs.VLLM_USE_V1:
output_shape = q.shape
output = torch.empty(output_shape,
dtype=q.dtype,
device=q.device)
forward_kwargs['output'] = output
output_shape = q.shape
output = torch.empty(output_shape, dtype=q.dtype, device=q.device)
forward_kwargs['output'] = output
attn_output = self.attn.impl.forward(self.attn,
q,

10
vllm_ascend/torchair/models/qwen3_moe.py
View File

@@ -19,7 +19,6 @@
from typing import Any, List, Optional, Union
import torch
import vllm.envs as envs
from torch import nn
from transformers import PretrainedConfig
from vllm.attention import Attention, AttentionMetadata
@@ -244,12 +243,9 @@ class CustomQwen3MoeAttention(Qwen3MoeAttention):
is_prefill=False,
is_qwen_torchair=True)
forward_kwargs = {}
if envs.VLLM_USE_V1:
output_shape = q.shape
output = torch.empty(output_shape,
dtype=q.dtype,
device=q.device)
forward_kwargs['output'] = output
output_shape = q.shape
output = torch.empty(output_shape, dtype=q.dtype, device=q.device)
forward_kwargs['output'] = output
attn_output = self.attn.impl.forward(self.attn,
q,