huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-11 04:14:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

dockerfile change to ipex cpu/xpu

Browse Source 
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
This commit is contained in:
Wang, Yi A 2025-02-11 19:51:05 +00:00
parent 4b8cda684b
commit a31641c1b6
1 changed files with 54 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

83
Dockerfile_intel
View File

@ -97,24 +97,34 @@ ENV HF_HOME=/data \
WORKDIR /usr/src WORKDIR /usr/src
RUN pip install https://intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn/ipex_dev/xpu/torch-2.5.0a0%2Bgite84e33f-cp311-cp311-linux_x86_64.whl --no-cache-dir
RUN pip install https://intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn/ipex_dev/xpu/torchaudio-2.5.0a0%2B56bc006-cp311-cp311-linux_x86_64.whl --no-cache-dir
RUN pip install https://intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn/ipex_dev/xpu/torchvision-0.20.0a0%2B8e8a208-cp311-cp311-linux_x86_64.whl --no-cache-dir
RUN pip install https://intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn/ipex_dev/xpu/oneccl_bind_pt-2.5.0%2Bxpu-cp311-cp311-linux_x86_64.whl --no-cache-dir
RUN pip install triton-xpu==3.0.0b2 --no-cache-dir
# Install server # Install server
COPY proto proto COPY proto proto
COPY server server COPY server server
COPY server/Makefile server/Makefile COPY server/Makefile server/Makefile
ENV UV_SYSTEM_PYTHON=1 ENV UV_SYSTEM_PYTHON=1
RUN cd server && \
make gen-server && \
pip install -U pip uv && \
uv pip install -e ".[accelerate, compressed-tensors, peft, outlines]" --no-cache-dir
ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/intel/oneapi/pti/0.9/lib:/opt/conda/lib RUN cd server && \
pip install -U pip uv && uv venv --seed && \
uv sync --frozen --extra gen --extra accelerate --extra compressed-tensors --extra quantize --extra peft --extra outlines --no-install-project && \
. ./.venv/bin/activate && \
make gen-server-raw
RUN cd server && \
uv sync --frozen --extra gen --extra accelerate --extra compressed-tensors --extra quantize --extra peft --extra outlines && \
. ./.venv/bin/activate && \
pwd && \
text-generation-server --help
ENV PATH=/usr/src/server/.venv/bin:$PATH
RUN pip install https://intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn/ipex_dev/xpu/torch-2.5.0a0%2Bgite84e33f-cp311-cp311-linux_x86_64.whl --no-cache-dir
RUN pip install https://intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn/ipex_dev/xpu/torchaudio-2.5.0a0%2B56bc006-cp311-cp311-linux_x86_64.whl --no-cache-dir
RUN pip install https://intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn/ipex_dev/xpu/torchvision-0.20.0a0%2B8e8a208-cp311-cp311-linux_x86_64.whl --no-cache-dir
RUN pip install https://intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn/ipex_dev/xpu/oneccl_bind_pt-2.5.0%2Bxpu-cp311-cp311-linux_x86_64.whl --no-cache-dir
RUN pip uninstall -y triton && pip install ruamel.yaml triton-xpu==3.0.0b2 --no-cache-dir
ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/intel/oneapi/pti/0.9/lib:/opt/conda/lib:/usr/src/server/.venv/lib
ENV CCL_ZE_IPC_EXCHANGE=sockets ENV CCL_ZE_IPC_EXCHANGE=sockets
#ENV TORCH_LLM_ALLREDUCE=1 #ENV TORCH_LLM_ALLREDUCE=1
#ENV CCL_TOPO_FABRIC_VERTEX_CONNECTION_CHECK=0 #ENV CCL_TOPO_FABRIC_VERTEX_CONNECTION_CHECK=0
@ -162,7 +172,7 @@ ARG MAMBA_VERSION=23.1.0-1
ARG PYTHON_VERSION='3.11.10' ARG PYTHON_VERSION='3.11.10'
# Automatically set by buildx # Automatically set by buildx
ARG TARGETPLATFORM ARG TARGETPLATFORM
ENV PATH /opt/conda/bin:$PATH ENV PATH=/opt/conda/bin:$PATH
# TGI seem to require libssl.so.1.1 instead of libssl.so.3 so we can't use ubuntu 22.04. Ubuntu 20.04 has python==3.8, and TGI requires python>=3.9, hence the need for miniconda. # TGI seem to require libssl.so.1.1 instead of libssl.so.3 so we can't use ubuntu 22.04. Ubuntu 20.04 has python==3.8, and TGI requires python>=3.9, hence the need for miniconda.
# Install mamba # Install mamba
@ -186,11 +196,6 @@ RUN case ${TARGETPLATFORM} in \
RUN conda install -c conda-forge gperftools mkl RUN conda install -c conda-forge gperftools mkl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torch-2.5.0.dev20240815%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torchvision-0.20.0.dev20240815%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torchaudio-2.4.0.dev20240815%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install triton==3.1.0 py-libnuma
WORKDIR /usr/src WORKDIR /usr/src
@ -198,16 +203,8 @@ RUN git clone https://github.com/intel/intel-extension-for-pytorch && cd intel-e
RUN git clone https://github.com/intel/torch-ccl.git && cd torch-ccl && git checkout v2.4.0+cpu+rc0 RUN git clone https://github.com/intel/torch-ccl.git && cd torch-ccl && git checkout v2.4.0+cpu+rc0
RUN sed -i 's/VERSION_MINOR 6/VERSION_MINOR 5/' intel-extension-for-pytorch/version.txt RUN sed -i 's/VERSION_MINOR 6/VERSION_MINOR 5/' intel-extension-for-pytorch/version.txt
RUN cd intel-extension-for-pytorch && git submodule sync && git submodule update --init --recursive && python setup.py install
RUN cd torch-ccl && git submodule sync && git submodule update --init --recursive && pip install .
ENV LD_PRELOAD=/opt/conda/lib/libtcmalloc.so ENV LD_PRELOAD=/opt/conda/lib/libtcmalloc.so
ENV CCL_ROOT=/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch
ENV I_MPI_ROOT=/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch
ENV FI_PROVIDER_PATH=/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib/prov:/usr/lib64/libfabric
ENV LD_LIBRARY_PATH=/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib:/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/lib
ENV LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/opt/conda/lib/"
# Install server # Install server
COPY proto proto COPY proto proto
@ -215,9 +212,34 @@ COPY server server
COPY server/Makefile server/Makefile COPY server/Makefile server/Makefile
ENV UV_SYSTEM_PYTHON=1 ENV UV_SYSTEM_PYTHON=1
RUN cd server && \ RUN cd server && \
make gen-server && \ pip install -U pip uv && uv venv --seed && \
pip install -U pip uv && \ uv sync --frozen --extra gen --extra accelerate --extra compressed-tensors --extra quantize --extra peft --extra outlines --no-install-project && \
uv pip install -e ".[accelerate, compressed-tensors, peft, outlines]" --no-cache-dir . ./.venv/bin/activate && \
make gen-server-raw
RUN cd server && \
uv sync --frozen --extra gen --extra accelerate --extra compressed-tensors --extra quantize --extra peft --extra outlines && \
. ./.venv/bin/activate && \
pwd && \
text-generation-server --help
ENV PATH=/usr/src/server/.venv/bin:$PATH
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torch-2.5.0.dev20240815%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torchvision-0.20.0.dev20240815%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torchaudio-2.4.0.dev20240815%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install triton==3.1.0 py-libnuma
RUN cd intel-extension-for-pytorch && git submodule sync && git submodule update --init --recursive && python3 setup.py install
RUN cd torch-ccl && git submodule sync && git submodule update --init --recursive && pip install .
ENV CCL_ROOT=/usr/src/server/.venv/lib/python3.11/site-packages/oneccl_bindings_for_pytorch
ENV I_MPI_ROOT=/usr/src/server/.venv/lib/python3.11/site-packages/oneccl_bindings_for_pytorch
ENV FI_PROVIDER_PATH=/usr/src/server/.venv/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib/prov:/usr/lib64/libfabric
ENV LD_LIBRARY_PATH=/usr/src/server/.venv/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib:/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/lib
ENV LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/opt/conda/lib:/usr/src/server/.venv/lib"
# Install benchmarker # Install benchmarker
COPY --from=builder /usr/src/target/release-opt/text-generation-benchmark /usr/local/bin/text-generation-benchmark COPY --from=builder /usr/src/target/release-opt/text-generation-benchmark /usr/local/bin/text-generation-benchmark
@ -231,5 +253,8 @@ ENV ATTENTION=flashdecoding-ipex
ENV PREFIX_CACHING=1 ENV PREFIX_CACHING=1
ENV PREFILL_CHUNKING=1 ENV PREFILL_CHUNKING=1
ENV CUDA_GRAPHS=0 ENV CUDA_GRAPHS=0
ENTRYPOINT ["text-generation-launcher"] COPY ./tgi-entrypoint.sh /tgi-entrypoint.sh
RUN chmod +x /tgi-entrypoint.sh
ENTRYPOINT ["/tgi-entrypoint.sh"]
CMD ["--json-output"] CMD ["--json-output"]