[Bug] Cannot install torch-npu==2.3.1, torch==2.3.1 and torchvision==0.18.1 because these package versions have conflicting dependencies. #2745
Comments
|
@jiabao-wang Hi, are you building docker image on x86_64 platform? Currently, the Dockerfile is only supported on aarch64. For x86_64 platform, the pypi package of dlinfer is not uploaded, as well as the problem you mentioned above. |
|
@jiabao-wang Here is a new Dockerfile for ascend x86_64 platform, it's only tested for building on x86_64 machine, the inference of models is not tested yet since we don't have any x86_64 ascend npu machine. FROM ubuntu:20.04 as base_builder
WORKDIR /tmp
ARG http_proxy
ARG https_proxy
ARG DEBIAN_FRONTEND=noninteractive
RUN sed -i 's@http://.*.ubuntu.com@http://mirrors.tuna.tsinghua.edu.cn@g' /etc/apt/sources.list && \
apt update && \
apt install --no-install-recommends ca-certificates -y && \
apt install --no-install-recommends bc wget -y && \
apt install --no-install-recommends git curl gcc make g++ pkg-config unzip -y && \
apt install --no-install-recommends libsqlite3-dev libblas3 liblapack3 gfortran vim -y && \
apt install --no-install-recommends liblapack-dev libblas-dev libhdf5-dev libffi-dev -y && \
apt install --no-install-recommends libssl-dev zlib1g-dev xz-utils cython3 python3-h5py -y && \
apt install --no-install-recommends libopenblas-dev libgmpxx4ldbl liblzma-dev -y && \
apt install --no-install-recommends libicu66 libxml2 pciutils libgl1-mesa-glx libbz2-dev -y && \
apt install --no-install-recommends libreadline-dev libncurses5 libncurses5-dev libncursesw5 -y && \
sed -i 's@http://mirrors.tuna.tsinghua.edu.cn@https://mirrors.tuna.tsinghua.edu.cn@g' /etc/apt/sources.list && \
apt clean && rm -rf /var/lib/apt/lists/*
ARG PYVERSION=3.10.5
ENV LD_LIBRARY_PATH=/usr/local/python${PYVERSION}/lib: \
PATH=/usr/local/python${PYVERSION}/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
RUN umask 0022 && \
wget https://repo.huaweicloud.com/python/${PYVERSION}/Python-${PYVERSION}.tar.xz && \
tar -xf Python-${PYVERSION}.tar.xz && cd Python-${PYVERSION} && ./configure --prefix=/usr/local/python${PYVERSION} --enable-shared && \
make -j 16 && make install && \
ln -sf /usr/local/python${PYVERSION}/bin/python3 /usr/bin/python3 && \
ln -sf /usr/local/python${PYVERSION}/bin/python3 /usr/bin/python && \
ln -sf /usr/local/python${PYVERSION}/bin/pip3 /usr/bin/pip3 && \
ln -sf /usr/local/python${PYVERSION}/bin/pip3 /usr/bin/pip && \
cd .. && \
rm -rf Python*
RUN --mount=type=cache,target=/root/.cache/pip pip3 config set global.index-url http://mirrors.aliyun.com/pypi/simple && \
pip3 config set global.trusted-host mirrors.aliyun.com && \
pip3 install -U pip && \
pip3 install wheel==0.43.0 scikit-build==0.18.0 numpy==1.24 setuptools==69.5.1 && \
pip3 install decorator sympy cffi && \
pip3 install cmake ninja pyyaml && \
pip3 install pathlib2 protobuf attrs attr scipy && \
pip3 install requests psutil absl-py
ENV LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu/hdf5/serial:$LD_LIBRARY_PATH
FROM ubuntu:20.04 as build_temp
COPY . /tmp
FROM base_builder as cann_builder
ARG ASCEND_BASE=/usr/local/Ascend
ARG TOOLKIT_PATH=$ASCEND_BASE/ascend-toolkit/latest
ENV LD_LIBRARY_PATH=\
$ASCEND_BASE/driver/lib64:\
$ASCEND_BASE/driver/lib64/common:\
$ASCEND_BASE/driver/lib64/driver:\
$ASCEND_BASE/driver/tools/hccn_tool/:\
$TOOLKIT_PATH/opp/built-in/op_impl/ai_core/tbe/op_tiling/lib/linux/x86_64/:\
$LD_LIBRARY_PATH
# run files should be placed at the root dir of repo
ARG CHIP=all
ARG TOOLKIT_PKG=Ascend-cann-toolkit_*.run
ARG KERNELS_PKG=Ascend-cann-kernels-*.run
ARG NNAL_PKG=Ascend-cann-nnal_*.run
RUN --mount=type=cache,target=/tmp,from=build_temp,source=/tmp \
umask 0022 && \
mkdir -p $ASCEND_BASE/driver && \
if [ "$CHIP" != "all" ]; \
then \
CHIPOPTION="--chip=$CHIP"; \
else \
CHIPOPTION=""; \
fi && \
chmod +x $TOOLKIT_PKG $KERNELS_PKG $NNAL_PKG && \
./$TOOLKIT_PKG --quiet --install --install-path=$ASCEND_BASE --install-for-all $CHIPOPTION && \
./$KERNELS_PKG --quiet --install --install-path=$ASCEND_BASE --install-for-all && \
. /usr/local/Ascend/ascend-toolkit/set_env.sh && \
./$NNAL_PKG --quiet --install --install-path=$ASCEND_BASE && \
rm -f $TOOLKIT_PKG $KERNELS_PKG $NNAL_PKG
ENV GLOG_v=2 \
LD_LIBRARY_PATH=$TOOLKIT_PATH/lib64:$LD_LIBRARY_PATH \
TBE_IMPL_PATH=$TOOLKIT_PATH/opp/op_impl/built-in/ai_core/tbe \
PATH=$TOOLKIT_PATH/ccec_compiler/bin:$PATH \
ASCEND_OPP_PATH=$TOOLKIT_PATH/opp \
ASCEND_AICPU_PATH=$TOOLKIT_PATH
ENV PYTHONPATH=$TBE_IMPL_PATH:$PYTHONPATH
SHELL ["/bin/bash", "-c"]
RUN echo "source /usr/local/Ascend/ascend-toolkit/set_env.sh" >> ~/.bashrc && \
echo "source /usr/local/Ascend/nnal/atb/set_env.sh --cxx_abi=0" >> ~/.bashrc && \
. ~/.bashrc
# dlinfer
# timm is required for internvl2 model
WORKDIR /opt/
RUN --mount=type=cache,target=/root/.cache/pip \
pip3 install torch==2.3.1+cpu torchvision==0.18.1+cpu --index-url=https://download.pytorch.org/whl/cpu && \
pip3 install torch-npu==2.3.1 && \
pip3 install transformers timm && \
git clone https://github.com/DeepLink-org/dlinfer.git && \
cd dlinfer && DEVICE=ascend python setup.py develop
# lmdeploy
FROM build_temp as copy_temp
RUN rm -rf /tmp/*.run
FROM cann_builder as final_builder
COPY --from=copy_temp /tmp /opt/lmdeploy
WORKDIR /opt/lmdeploy
RUN --mount=type=cache,target=/root/.cache/pip \
sed -i '/triton/d' requirements/runtime.txt && \
pip3 install -v --no-build-isolation -e . |
|
@CyCle1024 output error: The above exception was the direct cause of the following exception: Traceback (most recent call last): The above exception was the direct cause of the following exception: Traceback (most recent call last): |
Checklist
Describe the bug
when i run :
DOCKER_BUILDKIT=1 docker build -t lmdeploy-aarch64-ascend:latest
-f docker/Dockerfile_aarch64_ascend .
ERROR: Cannot install torch-npu==2.3.1, torch==2.3.1 and torchvision==0.18.1 because these package versions have conflicting dependencies.
341.8
341.8 The conflict is caused by:
341.8 The user requested torch==2.3.1
341.8 torchvision 0.18.1 depends on torch==2.3.1
341.8 torch-npu 2.3.1 depends on torch==2.3.1+cpu
341.8
341.8 To fix this you could try to:
341.8 1. loosen the range of package versions you've specified
341.8 2. remove package versions to allow pip to attempt to solve the dependency conflict
341.8
341.8 ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts
Dockerfile_aarch64_ascend:110
109 | # timm is required for internvl2 model
110 | >>> RUN --mount=type=cache,target=/root/.cache/pip
111 | >>> pip3 install torch==2.3.1 torchvision==0.18.1 torch-npu==2.3.1 &&
112 | >>> pip3 install transformers timm &&
113 | >>> pip3 install dlinfer-ascend
114 |
ERROR: failed to solve: process "/bin/bash -c pip3 install torch==2.3.1 torchvision==0.18.1 torch-npu==2.3.1 && pip3 install transformers timm && pip3 install dlinfer-ascend" did not complete successfully: exit code: 1
Reproduction
DOCKER_BUILDKIT=1 docker build -t lmdeploy-aarch64-ascend:latest
-f docker/Dockerfile_aarch64_ascend .
Environment
Error traceback
The text was updated successfully, but these errors were encountered: