Skip to content

Commit 23b4dd5

Browse files
hephaexhephaex
committed
fixed conflict
2 parents 711a656 + 73c8fb9 commit 23b4dd5

File tree

4 files changed

+296
-292
lines changed

4 files changed

+296
-292
lines changed

‎vendor/ngc-pytorch/Dockerfile.22.03-py3

Lines changed: 232 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,232 @@
1+
FROM nvcr.io/nvidia/pytorch:22.03-py3
2+
# NVIDIA NGC PyTorch with Python 3.8 (CONDA)
3+
4+
ENV DEBIAN_FRONTEND=noninteractive \
5+
MPLBACKEND=Svg \
6+
PIP_IGNORE_INSTALLED=0 \
7+
PYTHONUNBUFFERED=1 \
8+
LD_LIBRARY_PATH="/usr/local/cuda/compat/lib:/usr/local/cuda/extras/CUPTI/lib64:/usr/local/cuda/lib:/usr/local/cuda/lib64:/usr/local/nvidia/lib64:/usr/local/cuda-11.6/include:/usr/include/x86_64-linux-gnu:$LD_LIBRARY_PATH" \
9+
PATH="/usr/local/nvm/versions/node/v16.6.1/bin:/opt/conda/lib/python3.8/site-packages/torch_tensorrt/bin:/opt/conda/bin:/usr/local/mpi/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/local/ucx/bin:/opt/tensorrt/bin:/usr/local/src/lightgbm/LightGBM:/usr/local/bin/mecab" \
10+
mecab_dicdir=/usr/local/lib/mecab/dic/mecab-ko-dic \
11+
CPLUS_INCLUDE_PATH=/usr/include/gdal \
12+
C_INCLUDE_PATH=/usr/include/gdal \
13+
# CPATH=/usr/local/cuda-10.1/targets/x86_64-linux/include:$CPATH \
14+
LANG=C.UTF-8
15+
16+
RUN apt-get update && \
17+
apt-get install -y --no-install-recommends \
18+
automake \
19+
bzip2 \
20+
cabextract \
21+
ffmpeg \
22+
fonts-nanum \
23+
fonts-nanum-coding \
24+
fonts-nanum-extra \
25+
gfortran \
26+
htop \
27+
libasound2-dev \
28+
libatlas-base-dev \
29+
libavresample-dev \
30+
libdc1394-22-dev \
31+
libeigen3-dev \
32+
libfaac-dev \
33+
libgdal-dev \
34+
libgflags-dev \
35+
libgoogle-glog-dev \
36+
libgphoto2-dev \
37+
libgstreamer-plugins-base1.0-dev \
38+
libgstreamer1.0-dev \
39+
libgtk-3-dev \
40+
libhdf5-dev \
41+
libjemalloc-dev \
42+
libmp3lame-dev \
43+
libncurses-dev \
44+
libopenblas-dev \
45+
libopencore-amrnb-dev \
46+
libopencore-amrwb-dev \
47+
libprotobuf-dev \
48+
libtheora-dev \
49+
libvorbis-dev \
50+
libx264-dev \
51+
libxext6 \
52+
libxrender-dev \
53+
libxvidcore-dev \
54+
libsm6 \
55+
libtbb-dev \
56+
mercurial \
57+
ncurses-term \
58+
openjdk-8-jdk \
59+
pbzip2 \
60+
pv \
61+
pdsh \
62+
protobuf-compiler \
63+
v4l-utils \
64+
x264
65+
66+
# Install CUDA + cuDNN
67+
RUN ln -s /usr/lib/x86_64-linux-gnu/libcudnn.so.8.3.2 /usr/local/cuda/lib64/libcudnn.so && \
68+
rm -rf /var/lib/apt/lists/* && \
69+
ldconfig
70+
71+
# Install cmake
72+
RUN wget https://github.com/Kitware/CMake/releases/download/v3.22.0/cmake-3.22.0-Linux-x86_64.sh \
73+
-q -O /tmp/cmake-install.sh && \
74+
chmod u+x /tmp/cmake-install.sh && \
75+
mkdir /usr/bin/cmake && \
76+
/tmp/cmake-install.sh --skip-license --prefix=/usr/bin/cmake && \
77+
rm /tmp/cmake-install.sh
78+
79+
# nvtop install
80+
WORKDIR /tmp
81+
RUN git clone https://github.com/Syllo/nvtop.git && \
82+
mkdir -p nvtop/build && \
83+
cd /tmp/nvtop/build && \
84+
cmake .. && \
85+
cmake .. -DNVML_RETRIEVE_HEADER_ONLINE=True && \
86+
make -j$(nproc) && \
87+
make install
88+
89+
RUN update-alternatives --install /opt/conda/bin/python python /opt/conda/bin/python3 2
90+
91+
# install NLP packages *mecab-ko & khai*
92+
WORKDIR /tmp
93+
RUN curl -LO https://bitbucket.org/eunjeon/mecab-ko/downloads/mecab-0.996-ko-0.9.2.tar.gz && \
94+
tar zxfv mecab-0.996-ko-0.9.2.tar.gz && \
95+
cd mecab-0.996-ko-0.9.2 && \
96+
./configure && \
97+
make -j$(nproc) && \
98+
make check && \
99+
make install
100+
101+
RUN echo "Install mecab-ko-dic" && \
102+
cd /tmp && \
103+
ldconfig && \
104+
curl -LO https://bitbucket.org/eunjeon/mecab-ko-dic/downloads/mecab-ko-dic-2.1.1-20180720.tar.gz && \
105+
tar -zxvf mecab-ko-dic-2.1.1-20180720.tar.gz && \
106+
cd mecab-ko-dic-2.1.1-20180720 && \
107+
./autogen.sh && \
108+
./configure && \
109+
make -j$(nproc) && \
110+
sh -c 'echo "dicdir=/usr/local/lib/mecab/dic/mecab-ko-dic" > /usr/local/etc/mecabrc' && \
111+
make install && \
112+
cd /tmp && \
113+
git clone https://bitbucket.org/eunjeon/mecab-python-0.996.git && \
114+
python3 -m pip install /tmp/mecab-python-0.996
115+
116+
RUN curl -sL https://deb.nodesource.com/setup_16.x | bash - && \
117+
apt-get update -y && \
118+
apt-get install -y nodejs
119+
120+
WORKDIR /tmp
121+
RUN git clone -q --branch=v0.3.20 https://github.com/xianyi/OpenBLAS.git && \
122+
cd OpenBLAS && \
123+
make DYNAMIC_ARCH=1 NO_AFFINITY=1 NUM_THREADS=48 FC=gfortran && \
124+
make install && \
125+
cd /tmp && \
126+
git clone --recursive https://github.com/bodono/scs-python.git && \
127+
cd /tmp/scs-python && \
128+
python setup.py install --scs --gpu
129+
130+
RUN /opt/conda/bin/conda install -cᅧ conda-forge opencv ffmpeg spacy
131+
132+
WORKDIR /tmp
133+
COPY ./requirements.22.03.txt /tmp/requirements.txt
134+
RUN /opt/conda/bin/python3 -m pip install --no-cache-dir -r requirements.txt && \
135+
rm -f /tmp/*.whl /tmp/requirements.txt
136+
137+
# install git-lfs
138+
WORKDIR /tmp
139+
RUN curl -sLO https://github.com/git-lfs/git-lfs/releases/download/v3.0.2/git-lfs-linux-amd64-v3.0.2.tar.gz && \
140+
tar -zxf git-lfs-linux-amd64-v3.0.2.tar.gz && \
141+
bash install.sh && \
142+
rm -rf /tmp/*
143+
144+
WORKDIR /tmp
145+
RUN git clone https://github.com/aristocratos/bashtop.git && \
146+
cd bashtop && \
147+
make install
148+
149+
RUN curl -fL https://github.com/cdr/code-server/releases/download/v4.0.2/code-server-4.0.2-linux-amd64.tar.gz \
150+
| tar -C /usr/local/lib -xz && \
151+
mv /usr/local/lib/code-server-4.0.2-linux-amd64 /usr/local/lib/code-server-4.0.2 && \
152+
ln -s /usr/local/lib/code-server-4.0.2/bin/code-server /usr/local/bin/code-server
153+
154+
# Install Open MPI
155+
RUN mkdir /tmp/openmpi && \
156+
cd /tmp/openmpi && \
157+
wget https://download.open-mpi.org/release/open-mpi/v4.1/openmpi-4.1.1.tar.gz && \
158+
tar zxf openmpi-4.1.1.tar.gz && \
159+
cd openmpi-4.1.1 && \
160+
./configure --enable-orterun-prefix-by-default && \
161+
make -j $(nproc) all && \
162+
make install && \
163+
ldconfig && \
164+
rm -rf /tmp/openmpi* && \
165+
# Create a wrapper for OpenMPI to allow running as root by default
166+
mv /usr/local/bin/mpirun /usr/local/bin/mpirun.real && \
167+
echo '#!/bin/bash' > /usr/local/bin/mpirun && \
168+
echo 'mpirun.real --allow-run-as-root "$@"' >> /usr/local/bin/mpirun && \
169+
chmod a+x /usr/local/bin/mpirun && \
170+
171+
# Configure OpenMPI to run good defaults:
172+
echo "btl_tcp_if_exclude = lo,docker0" >> /usr/local/etc/openmpi-mca-params.conf
173+
174+
# Install Horovod, temporarily using CUDA stubs
175+
RUN ldconfig /usr/local/cuda/targets/x86_64-linux/lib/stubs && \
176+
# cp /usr/local/cuda-11.6/bin/nvcc /bin/nvcc && \
177+
# ldconfig /usr/local/cuda/targets/x86_64-linux/lib/stubs && \
178+
HOROVOD_CUDA_HOME=$CONDA_PREFIX HOROVOD_GPU_ALLREDUCE=NCCL HOROVOD_GPU_BROADCAST=NCCL HOROVOD_NCCL_LINK=SHARED \
179+
HOROVOD_WITHOUT_TENSORFLOW=1 HOROVOD_WITH_PYTORCH=1 HOROVOD_WITHOUT_MXNET=1 \
180+
pip install --no-cache-dir horovod==0.24.2 && \
181+
ldconfig
182+
183+
RUN python3 -m pip install --no-cache-dir \
184+
mpi4py==3.1.2 \
185+
nni==2.5 \
186+
mlflow==1.21.0 \
187+
scikit-nni==0.2.1
188+
189+
RUN rm /usr/local/bin/node
190+
RUN jupyter nbextensions_configurator enable && \
191+
jupyter contrib nbextension install && \
192+
jupyter nbextension enable --py --sys-prefix widgetsnbextension && \
193+
jupyter serverextension enable --py jupyterlab --sys-prefix && \
194+
jupyter labextension install --no-build @jupyter-widgets/jupyterlab-manager && \
195+
jupyter labextension install --no-build @krassowski/jupyterlab-lsp && \
196+
jupyter serverextension enable --py jupyter_lsp && \
197+
jupyter labextension install --no-build @jupyterlab/toc && \
198+
jupyter nbextension enable execute_time/ExecuteTime && \
199+
jupyter nbextension enable toc2/main && \
200+
jupyter lab build
201+
202+
RUN apt autoclean && \
203+
sed -i 's/source \/usr\/local\/nvm\/nvm.sh//' /etc/bash.bashrc && \
204+
ln -sf /usr/share/terminfo/x/xterm-color /usr/share/terminfo/x/xterm-256color && \
205+
rm -rf /var/lib/apt/lists/* && \
206+
rm -rf /root/.cache && \
207+
rm -rf /tmp/*
208+
209+
RUN /opt/conda/bin/python3 -m ipykernel install \
210+
--prefix=/opt/conda/ \
211+
--display-name "PyTorch 1.11 (NGC 22.03/Python 3.8 Conda) on Backend.AI" && \
212+
cat /opt/conda/share/jupyter/kernels/python3/kernel.json
213+
214+
# Backend.AI specifics
215+
COPY ./service-defs /etc/backend.ai/service-defs
216+
COPY ./runner-scripts/bootstrap.sh runner-scripts/setup_multinode.py /opt/container/
217+
218+
LABEL ai.backend.kernelspec="1" \
219+
ai.backend.envs.corecount="OPENBLAS_NUM_THREADS,OMP_NUM_THREADS,NPROC" \
220+
ai.backend.features="batch query uid-match user-input" \
221+
ai.backend.base-distro="ubuntu16.04" \
222+
ai.backend.accelerators="cuda" \
223+
ai.backend.resource.min.cpu="1" \
224+
ai.backend.resource.min.mem="1g" \
225+
ai.backend.resource.min.cuda.device=1 \
226+
ai.backend.resource.min.cuda.shares=0 \
227+
ai.backend.base-distro="ubuntu16.04" \
228+
ai.backend.runtime-type="python" \
229+
ai.backend.runtime-path="/opt/conda/bin/python3" \
230+
ai.backend.service-ports="ipython:pty:3000,jupyter:http:8091,jupyterlab:http:8090,vscode:http:8180,tensorboard:http:6006,mlflow-ui:preopen:5000,nniboard:preopen:8080"
231+
232+
WORKDIR /home/work

0 commit comments

Comments
 (0)