@@ -33,12 +33,11 @@ ENV KMP_AFFINITY='granularity=fine,verbose,compact,1,0' \
3333 KMP_BLOCKTIME=1 \
3434 KMP_SETTINGS=1
3535
36- ARG TF_VERSION
37-
3836WORKDIR /
3937COPY requirements.txt .
4038
41- RUN python -m pip install --no-cache-dir -r requirements.txt
39+ RUN python -m pip install --no-cache-dir -r requirements.txt && \
40+ rm -rf requirements.txt
4241
4342ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/master/third-party-programs/dockerlayer/THIRD-PARTY-PROGRAMS.txt /licenses/
4443ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/master/third-party-programs/dockerlayer/third-party-program-of-intel-extension-for-tensorflow.txt /licenses/
@@ -53,12 +52,13 @@ ENV KMP_AFFINITY='granularity=fine,verbose,compact,1,0' \
5352ENV PATH /usr/bin:/root/conda/envs/idp/bin:/root/conda/condabin:~/conda/bin/:${PATH}
5453
5554ENV TF_ENABLE_ONEDNN_OPTS=1
56- ARG TF_VERSION
5755
5856WORKDIR /
5957COPY requirements.txt .
6058
61- RUN python -m pip install --no-cache-dir -r requirements.txt
59+ RUN conda run -n idp python -m pip install --no-cache-dir -r requirements.txt && \
60+ rm -rf requirements.txt && \
61+ conda clean -y --all
6262
6363ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/master/third-party-programs/dockerlayer/THIRD-PARTY-PROGRAMS.txt /licenses/
6464ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/master/third-party-programs/dockerlayer/third-party-program-of-intel-extension-for-tensorflow.txt /licenses/
@@ -77,37 +77,43 @@ EXPOSE 8888
7777
7878CMD ["bash" , "-c" , "source /etc/bash.bashrc && jupyter notebook --notebook-dir=/jupyter --port 8888 --ip 0.0.0.0 --no-browser --allow-root --ServerApp.token= --ServerApp.password= --ServerApp.allow_origin=* --ServerApp.base_url=$NB_PREFIX" ]
7979
80- FROM tf-base-${PACKAGE_OPTION} AS openmpi
80+ FROM tf-base-${PACKAGE_OPTION} AS multinode
8181
8282RUN apt-get update -y && apt-get install -y --no-install-recommends --fix-missing \
83+ build-essential \
84+ cmake \
85+ g++ \
86+ gcc \
87+ git \
88+ libgl1-mesa-glx \
89+ libglib2.0-0 \
8390 libopenmpi-dev \
91+ numactl \
8492 openmpi-bin \
85- openmpi-common
93+ openmpi-common \
94+ python3-dev \
95+ unzip \
96+ virtualenv
8697
87- WORKDIR /
88- COPY ompi-requirements.txt .
98+ ENV SIGOPT_PROJECT=.
8999
90- RUN python -m pip install --no-cache-dir -r ompi-requirements.txt
100+ WORKDIR /
101+ COPY multinode/requirements.txt requirements.txt
91102
92- FROM openmpi AS horovod
103+ RUN python -m pip install --no-cache-dir -r requirements.txt && \
104+ rm -rf requirements.txt
93105
94- ENV LD_LIBRARY_PATH /lib64/: /usr/lib64/ :/usr/local/lib64
106+ ENV LD_LIBRARY_PATH= "/lib/x86_64-linux-gnu:${LD_LIBRARY_PATH}: /usr/local/lib/python${PYTHON_VERSION}/dist-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib :/usr/local/lib/python${PYTHON_VERSION}/dist-packages/oneccl_bindings_for_pytorch/lib"
95107
96108RUN apt-get install -y --no-install-recommends --fix-missing \
97- unzip \
98109 openssh-client \
99110 openssh-server && \
100- rm /etc/ssh/ssh_host_*_key \
101- /etc/ssh/ssh_host_*_key.pub
102-
103- ENV OMPI_ALLOW_RUN_AS_ROOT=1
104- ENV OMPI_ALLOW_RUN_AS_ROOT_CONFIRM=1
105-
106- ENV OMPI_MCA_tl_tcp_if_exclude="lo,docker0"
111+ rm /etc/ssh/ssh_host_*_key \
112+ /etc/ssh/ssh_host_*_key.pub && \
113+ apt-get clean && \
114+ rm -rf /var/lib/apt/lists/*
107115
108- # Install OpenSSH for MPI to communicate between containers
109- RUN mkdir -p /var/run/sshd && \
110- echo 'LoginGraceTime 0' >> /etc/ssh/sshd_config
116+ RUN mkdir -p /var/run/sshd
111117
112118# Install Horovod
113119ARG HOROVOD_WITH_TENSORFLOW=1
@@ -116,43 +122,32 @@ ARG HOROVOD_WITHOUT_PYTORCH=1
116122ARG HOROVOD_WITHOUT_GLOO=1
117123ARG HOROVOD_WITH_MPI=1
118124
119- RUN apt-get install -y --no-install-recommends --fix-missing \
120- build-essential \
121- cmake \
122- g++ \
123- gcc \
124- git \
125- libgl1-mesa-glx \
126- libglib2.0-0 \
127- python3-dev && \
128- apt-get clean && \
129- rm -rf /var/lib/apt/lists/*
130-
131- WORKDIR /
132- COPY hvd-requirements.txt .
133-
134- RUN python -m pip install --no-cache-dir -r hvd-requirements.txt
135-
136- ENV SIGOPT_PROJECT=.
137-
138- RUN wget --progress=dot:giga --no-check-certificate https://github.com/intel/neural-compressor/raw/master/docker/third-party-programs-tensorflow.txt -O /licenses/inc-third-party-programs-tensorflow.txt && \
139- wget --progress=dot:giga --no-check-certificate https://raw.githubusercontent.com/intel/neural-compressor/master/LICENSE -O /licenses/INC_LICENSE
125+ ENV LD_LIBRARY_PATH /lib64/:/usr/lib64/:/usr/local/lib64
140126
141- FROM horovod AS multinode- pip
127+ RUN python -m pip install --no-cache-dir horovod==0.28.1
142128
143- WORKDIR /
144- COPY multinode-requirements.txt .
129+ ARG PYTHON_VERSION
145130
146- RUN python -m pip install --no-cache-dir -r multinode-requirements.txt
131+ COPY multinode/generate_ssh_keys.sh /generate_ssh_keys.sh
147132
148- FROM horovod AS multinode-idp
133+ # modify generate_ssh_keys to be a helper script
134+ # print how to use helper script on bash startup
135+ # Avoids loop for further execution of the startup file
136+ ARG PACKAGE_OPTION=pip
137+ ARG PYPATH="/usr/local/lib/python${PYTHON_VERSION}/dist-packages"
138+ RUN if [ "${PACKAGE_OPTION}" = "idp" ]; then PYPATH="/opt/conda/envs/idp/lib/python${PYTHON_VERSION}/site-packages" ; fi && \
139+ echo "source ${PYPATH}/oneccl_bindings_for_pytorch/env/setvars.sh" >> ~/.startup && \
140+ cat '/generate_ssh_keys.sh' >> ~/.startup && \
141+ rm -rf /generate_ssh_keys.sh
149142
150- WORKDIR /
151- COPY multinode-requirements.txt .
143+ COPY multinode/dockerd-entrypoint.sh /usr/local/bin/dockerd-entrypoint.sh
144+ COPY multinode/sshd_config /etc/ssh/sshd_config
145+ COPY multinode/ssh_config /etc/ssh/ssh_config
152146
153- RUN python -m pip install --no-cache-dir -r multinode-requirements.txt
147+ RUN wget --progress=dot:giga --no-check-certificate https://github.com/intel/neural-compressor/raw/master/docker/third-party-programs-tensorflow.txt -O /licenses/inc-third-party-programs-tensorflow.txt && \
148+ wget --progress=dot:giga --no-check-certificate https://raw.githubusercontent.com/intel/neural-compressor/master/LICENSE -O /licenses/INC_LICENSE
154149
155- FROM ${PYTHON_BASE} AS itex-xpu-base-pip
150+ FROM ${PYTHON_BASE} AS itex-xpu-base
156151
157152RUN apt-get update && \
158153 apt-get install -y --no-install-recommends --fix-missing \
@@ -219,54 +214,7 @@ ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/maste
219214
220215ENV LD_LIBRARY_PATH=/opt/intel/oneapi/redist/lib:$LD_LIBRARY_PATH
221216
222- FROM ${PYTHON_BASE} AS itex-xpu-base-idp
223-
224- RUN apt-get update && \
225- apt-get install -y --no-install-recommends --fix-missing \
226- apt-utils \
227- build-essential \
228- clinfo \
229- git \
230- gnupg2 \
231- gpg-agent \
232- rsync \
233- unzip \
234- wget && \
235- apt-get clean && \
236- rm -rf /var/lib/apt/lists/*
237-
238- ARG ICD_VER
239- ARG LEVEL_ZERO_GPU_VER
240- ARG LEVEL_ZERO_VER
241- ARG LEVEL_ZERO_DEV_VER
242-
243- RUN no_proxy="" NO_PROXY="" wget -qO - https://repositories.intel.com/gpu/intel-graphics.key | \
244- gpg --dearmor --output /usr/share/keyrings/intel-graphics.gpg
245- RUN echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu jammy/lts/2350 unified" | \
246- tee /etc/apt/sources.list.d/intel-gpu-jammy.list
247-
248- RUN no_proxy="" NO_PROXY="" apt-get update && \
249- apt-get install -y --no-install-recommends --fix-missing \
250- intel-opencl-icd=${ICD_VER} \
251- intel-level-zero-gpu=${LEVEL_ZERO_GPU_VER} \
252- level-zero=${LEVEL_ZERO_VER} \
253- level-zero-dev=${LEVEL_ZERO_DEV_VER} && \
254- apt-get clean && \
255- rm -rf /var/lib/apt/lists/*
256-
257- ARG ITEX_VER="2.15.0.1"
258-
259- RUN conda install -n idp -y intel-extension-for-tensorflow=${ITEX_VER}=*xpu* \
260- -c https://software.repos.intel.com/python/conda
261-
262- ENV LD_LIBRARY_PATH=/opt/conda/envs/idp/lib:$LD_LIBRARY_PATH
263-
264- ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/master/third-party-programs/dockerlayer/THIRD-PARTY-PROGRAMS.txt /licenses/
265- ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/master/third-party-programs/dockerlayer/third-party-program-of-intel-extension-for-tensorflow.txt /licenses/
266- ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/master/third-party-programs/dockerlayer/third-party-programs-of-intel-tensorflow.txt /licenses/
267- ADD https://raw.githubusercontent.com/intel/intel-extension-for-tensorflow/master/third-party-programs/dockerlayer/third-party-programs-of-intel-optimization-for-horovod.txt /licenses/
268-
269- FROM itex-xpu-base-${PACKAGE_OPTION} AS itex-xpu-jupyter
217+ FROM itex-xpu-base AS itex-xpu-jupyter
270218
271219WORKDIR /jupyter
272220COPY jupyter-requirements.txt .
![dependabot[bot]](https://avatars.githubusercontent.com/in/29110?v=4&size=40){kind=avatar}
0 commit comments