[Main2Main] Upgrade vllm commit to 0113 (#5839)

### What this PR does / why we need it?
Upgrade vllm commit to 0113 (11b6af5280d6d6dfb8953af16e67b25f819b3be9)

- Modify import paths due to the refactors
https://github.com/vllm-project/vllm/pull/31916
https://github.com/vllm-project/vllm/pull/32054

- Fix `TypeError: NPUOffloadingSpec.__init__() takes 2 positional
arguments but 3 were given` due to
https://github.com/vllm-project/vllm/pull/24498

- Skip the async-scheduling tests in
`tests/e2e/multicard/4-cards/long_sequence/test_mtp.py`, which are never
verified
https://github.com/vllm-project/vllm/pull/31998

- Skip some pooling tests, which are caused by
https://github.com/vllm-project/vllm/pull/32148
where vllm is also failed
https://buildkite.com/vllm/ci/builds/46705/steps/canvas?jid=019bb329-3834-4685-862b-1613b8e0f5d4

We will reopen those tests when main2main reachs
https://github.com/vllm-project/vllm/pull/32243

- Skip some cases in
`tests/e2e/multicard/4-cards/long_sequence/test_mtp.py`, which are
broken by
https://github.com/vllm-project/vllm/pull/32118

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

### How was this patch tested?

- vLLM version: v0.13.0
- vLLM main:
2f4e6548ef

Signed-off-by: wjunLu <wjunlu217@gmail.com>
Signed-off-by: hfadzxy <starmoon_zhang@163.com>
Co-authored-by: hfadzxy <starmoon_zhang@163.com>
This commit is contained in:
wjunLu
2026-01-15 09:48:53 +08:00
committed by GitHub
parent e67608041d
commit c11a05c4e1
29 changed files with 229 additions and 54 deletions

View File

@@ -10,7 +10,6 @@ from vllm.distributed import (get_dcp_group,
get_pcp_group)
from vllm.forward_context import ForwardContext, get_forward_context
from vllm.utils.math_utils import cdiv
from vllm.v1.attention.backends.utils import AttentionCGSupport
from vllm.v1.kv_cache_interface import AttentionSpec, MLAAttentionSpec
# isort: off
@@ -28,7 +27,12 @@ from vllm_ascend.attention.context_parallel.common_cp import (
from vllm_ascend.compilation.acl_graph import (get_draft_graph_params,
get_graph_params,
update_graph_params_workspaces)
from vllm_ascend.utils import weak_ref_tensors
from vllm_ascend.utils import weak_ref_tensors, vllm_version_is
if vllm_version_is('0.13.0'):
from vllm.v1.attention.backends.utils import AttentionCGSupport
else:
from vllm.v1.attention.backend import AttentionCGSupport
MAX_O_PROJ_PREFETCH_SIZE = 16 * 1024 * 1024