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

Drop 0.11.0 support (#4377)

Browse Source 
There is a lot hack code for v0.11.0, which makes the code hard to
upgrade to newer vLLM version. Since v0.11.0 will release soon. Let's
drop v0.11.0 support first. Then we'll upgrade to v0.11.2 soon.


- vLLM version: v0.11.0
- vLLM main:
2918c1b49c

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>
This commit is contained in:
wangxiyuan
2025-11-24 17:08:20 +08:00
committed by GitHub
parent 41ddb06554
commit a1f142b7ad
80 changed files with 467 additions and 1755 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
vllm_ascend/ops/register_custom_ops.py
View File

@@ -7,17 +7,12 @@ from vllm.distributed import (get_dp_group, get_ep_group,
tensor_model_parallel_all_reduce,
tensor_model_parallel_reduce_scatter)
from vllm.forward_context import get_forward_context
from vllm.utils.torch_utils import direct_register_custom_op
import vllm_ascend.envs as envs_ascend
from vllm_ascend.ascend_forward_context import MoECommType
from vllm_ascend.ops.weight_prefetch import maybe_npu_prefetch
from vllm_ascend.utils import (npu_stream_switch, prefetch_stream,
vllm_version_is)
if vllm_version_is("0.11.0"):
from vllm.utils import direct_register_custom_op
else:
from vllm.utils.torch_utils import direct_register_custom_op
from vllm_ascend.utils import npu_stream_switch, prefetch_stream
def _maybe_all_gather_and_maybe_unpad_impl(