bd571cf6d6
### What this PR does / why we need it?
break:
- https://github.com/vllm-project/vllm/pull/34102
Disable_full param replaced with valid_modes/invalid_modes API
- https://github.com/vllm-project/vllm/pull/35503
Now must return float compilation_time
- https://github.com/vllm-project/vllm/pull/35564
New sequence_lengths param added
- https://github.com/vllm-project/vllm/pull/33807
A check was performed (if runner_backend != "auto")
- https://github.com/vllm-project/vllm/pull/34861
`BaseDeviceCommunicator` now accesses PyTorch's internal `pg_map` to
check process group state
- https://github.com/vllm-project/vllm/pull/35274
**Important change:**
- https://github.com/vllm-project/vllm/pull/28672
`matcher_utils` directly accesses `torch.ops._C.*` during the import
phase. In the Ascend environment, some unregistered ops trigger
`AttributeError`, causing e2e initialization failure.
https://github.com/vllm-project/vllm-ascend/actions/runs/22607260487/job/65502047131#step:10:2323
https://github.com/vllm-project/vllm/blob/main/vllm/compilation/passes/fusion/matcher_utils.py#L29
This PR adds temporary compatibility placeholders (rms_norm,
fused_add_rms_norm, rotate_embedding, static/dynamic fp8 quant,
silu_and_mul) to
`vllm_ascend/patch/platform/patch_fusion_matcher_compat_ops.py` to
ensure no crashes during the import phase. Upstream repairs will be
considered later.
### Does this PR introduce _any_ user-facing change?
### How was this patch tested?
- vLLM version: v0.16.0
- vLLM main:
15d76f74e2
---------
Signed-off-by: MrZ20 <2609716663@qq.com>
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
Co-authored-by: Meihan-chen <jcccx.cmh@gmail.com>
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: gcanlin <canlinguosdu@gmail.com>
47 lines
1.6 KiB
Docker
47 lines
1.6 KiB
Docker
#
|
|
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
# This file is a part of the vllm-ascend project.
|
|
#
|
|
|
|
FROM ascendai/python:3.11-ubuntu22.04
|
|
|
|
ARG TARGETARCH
|
|
|
|
RUN apt-get update -y && \
|
|
apt-get install -y curl git gcc g++ cmake libnuma-dev jq && \
|
|
rm -rf /var/cache/apt/* && \
|
|
rm -rf /var/lib/apt/lists/*
|
|
|
|
|
|
ARG VLLM_REPO=https://github.com/vllm-project/vllm.git
|
|
# For lint purpose, actually we need make a main2main matching.
|
|
ARG VLLM_COMMIT=4034c3d32e30d01639459edd3ab486f56993876d
|
|
RUN git clone $VLLM_REPO /vllm-workspace/vllm && \
|
|
cd /vllm-workspace/vllm && \
|
|
git checkout $VLLM_COMMIT
|
|
|
|
# # Install vLLM common dependencies
|
|
RUN python3 -m pip install -r /vllm-workspace/vllm/requirements/common.txt --extra-index https://download.pytorch.org/whl/cpu/ && \
|
|
python3 -m pip uninstall -y triton && \
|
|
python3 -m pip cache purge
|
|
|
|
COPY . /vllm-workspace/vllm-ascend/
|
|
|
|
RUN pip install -r /vllm-workspace/vllm-ascend/requirements-dev.txt --extra-index-url https://download.pytorch.org/whl/cpu && \
|
|
pip cache purge && \
|
|
rm -fr /vllm-workspace/
|
|
|
|
CMD ["/bin/bash"]
|