-
Notifications
You must be signed in to change notification settings - Fork 377
/
Dockerfile
196 lines (160 loc) · 9.24 KB
/
Dockerfile
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
ARG BUILD_TYPE
ARG CUDA_VERSION
ARG UBUNTU_VERSION
########################################################################
FROM nvidia/cuda:${CUDA_VERSION}-cudnn8-runtime-ubuntu${UBUNTU_VERSION} as thinbuild
ARG PYTHON_VERSION=3.11
# build-essential: installs gcc which is needed to install some deps like rasterio
# libGL1: needed to avoid following error when using cv2
# ImportError: libGL.so.1: cannot open shared object file: No such file or directory
# See https://stackoverflow.com/questions/55313610/importerror-libgl-so-1-cannot-open-shared-object-file-no-such-file-or-directo
RUN --mount=type=cache,target=/var/cache/apt apt update && \
apt install -y wget=1.21.2-2ubuntu1 build-essential=12.9ubuntu3 libgl1=1.4.0-1 curl=7.81.0-1ubuntu1.13 git=1:2.34.1-1ubuntu1.10 tree=2.0.2-1 gdal-bin=3.4.1+dfsg-1build4 libgdal-dev=3.4.1+dfsg-1build4 python${PYTHON_VERSION} python3-pip && \
curl -fsSL https://deb.nodesource.com/node_16.x | bash - && \
apt install -y nodejs=16.20.2-deb-1nodesource1 && \
update-alternatives --install /usr/bin/python3 python3 /usr/bin/python${PYTHON_VERSION} 1 && \
update-alternatives --install /usr/bin/python python /usr/bin/python${PYTHON_VERSION} 1 && \
apt autoremove
########################################################################
FROM nvidia/cuda:${CUDA_VERSION}-cudnn8-runtime-ubuntu${UBUNTU_VERSION} as fullbuild
ARG TARGETPLATFORM
ARG PYTHON_VERSION=3.11
# wget: needed below to install conda
# build-essential: installs gcc which is needed to install some deps like rasterio
# libGL1: needed to avoid following error when using cv2
# ImportError: libGL.so.1: cannot open shared object file: No such file or directory
# See https://stackoverflow.com/questions/55313610/importerror-libgl-so-1-cannot-open-shared-object-file-no-such-file-or-directo
RUN apt-get update && \
apt-get install -y wget=1.* build-essential libgl1 curl git tree && \
apt-get autoremove && apt-get autoclean && apt-get clean
RUN case ${TARGETPLATFORM} in \
"linux/arm64") LINUX_ARCH=aarch64 ;; \
*) LINUX_ARCH=x86_64 ;; \
esac && echo ${LINUX_ARCH} > /root/linux_arch
# needed for jupyter lab extensions
RUN curl -fsSL https://deb.nodesource.com/node_16.x | bash - && \
apt-get install -y nodejs
# Install Python and conda/mamba (mamba installs conda as well)
RUN wget -q -O ~/micromamba.sh https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-Linux-$(cat /root/linux_arch).sh && \
chmod +x ~/micromamba.sh && \
bash ~/micromamba.sh -b -p /opt/conda && \
rm ~/micromamba.sh
ENV PATH /opt/conda/bin:$PATH
ENV LD_LIBRARY_PATH /opt/conda/lib/:$LD_LIBRARY_PATH
RUN mamba init
RUN mamba install -y python=${PYTHON_VERSION}
RUN python -m pip install --upgrade pip
# env variable required by uv
ENV CONDA_PREFIX=/opt/conda
RUN pip install uv
# We need to install GDAL first to install Rasterio on non-AMD64 architectures.
# The Rasterio wheels contain GDAL in them, but they are only built for AMD64 now.
RUN mamba update mamba -y && mamba install -y -c conda-forge gdal=3.6.3
ENV GDAL_DATA=/opt/conda/lib/python${PYTHON_VERSION}/site-packages/rasterio/gdal_data/
# Needed for GDAL 3.0
ENV PROJ_LIB /opt/conda/share/proj/
# This is to prevent the following error when starting the container.
# bash: /opt/conda/lib/libtinfo.so.6: no version information available (required by bash)
# See https://askubuntu.com/questions/1354890/what-am-i-doing-wrong-in-conda
RUN rm /opt/conda/lib/libtinfo.so.6 && \
ln -s /lib/$(cat /root/linux_arch)-linux-gnu/libtinfo.so.6 /opt/conda/lib/libtinfo.so.6
# This gets rid of the following error when importing cv2 on arm64.
# We cannot use the ENV directive since it cannot be used conditionally.
# See https://github.com/opencv/opencv/issues/14884
# ImportError: /lib/aarch64-linux-gnu/libGLdispatch.so.0: cannot allocate memory in static TLS block
RUN if [ "${TARGETARCH}" = "arm64" ]; \
then echo "export LD_PRELOAD=/lib/$(cat /root/linux_arch)-linux-gnu/libGLdispatch.so.0:$LD_PRELOAD" >> /root/.bashrc; fi
########################################################################
FROM ${BUILD_TYPE:-fullbuild} AS final_stage
ARG TARGETARCH
ENV LC_ALL C.UTF-8
ENV LANG C.UTF-8
WORKDIR /opt/src/
#------------------------------------------------------------------------
# Ideally we'd just pip install each package, but if we do that, then
# a lot of the image will have to be re-built each time we make a
# change to the code. So, we split the install into installing all the
# requirements in bunches (filtering out any prefixed with
# rastervision_*), and then copy over the source code. The
# dependencies are installed in bunches rather than package-by-package
# or on a per-RV component basis to reduce the build time, the number
# of layers, and the overall image size, and to reduce churn
# (installing and uninstalling of Python packages during the build).
#
# The bunches are heuristic and are meant to keep the heaviest and/or
# least-frequently-changing dependencies before the more variable
# ones. At time of writing, the amount of image size attributable to
# PyTorch (and the amount of image size overall) is heavily dominated
# by PyTorch, so it is first.
# Install requirements.
# -E "^\s*$|^#|rastervision_*" means exclude blank lines, comment lines,
# and rastervision plugins.
COPY ./rastervision_pytorch_learner/requirements.txt /opt/src/pytorch-requirements.txt
RUN --mount=type=cache,target=/root/.cache/pip cat pytorch-requirements.txt | sort | uniq > all-requirements.txt && \
uv pip install $(grep -ivE "^\s*$|^#|rastervision_*" all-requirements.txt) && \
rm all-requirements.txt
COPY ./rastervision_aws_batch/requirements.txt /opt/src/batch-requirements.txt
COPY ./rastervision_aws_s3/requirements.txt /opt/src/s3-requirements.txt
COPY ./rastervision_core/requirements.txt /opt/src/core-requirements.txt
# Pip wheels for triangle are missing for ARM64 architectures and building
# from source fails, so we skip it.
RUN if [ "${TARGETARCH}" = "arm64" ]; \
then sed -i '/^triangle.*$/d' /opt/src/core-requirements.txt; fi
COPY ./rastervision_gdal_vsi/requirements.txt /opt/src/gdal-requirements.txt
COPY ./rastervision_pipeline/requirements.txt /opt/src/pipeline-requirements.txt
COPY ./rastervision_aws_sagemaker/requirements.txt /opt/src/sagemaker-requirements.txt
COPY ./requirements-dev.txt /opt/src/requirements-dev.txt
RUN --mount=type=cache,target=/root/.cache/pip cat \
/opt/src/batch-requirements.txt \
/opt/src/s3-requirements.txt \
/opt/src/core-requirements.txt \
/opt/src/gdal-requirements.txt \
/opt/src/pipeline-requirements.txt \
/opt/src/sagemaker-requirements.txt \
/opt/src/requirements-dev.txt \
| sort | uniq > all-requirements.txt && \
uv pip install $(grep -ivE "^\s*$|^#|rastervision_*" all-requirements.txt) && \
rm all-requirements.txt
#########################
# Docs
#########################
# Install docs/requirements.txt
COPY ./docs/requirements.txt /opt/src/docs/pandoc-requirements.txt
# Install pandoc, needed for rendering notebooks
# Get latest release link from here: https://github.com/jgm/pandoc/releases
RUN --mount=type=cache,target=/root/.cache/pip uv pip install -r docs/pandoc-requirements.txt && \
wget https://github.com/jgm/pandoc/releases/download/3.1.12.2/pandoc-3.1.12.2-1-${TARGETARCH}.deb && \
dpkg -i pandoc-3.1.12.2-1-${TARGETARCH}.deb && rm pandoc-3.1.12.2-1-${TARGETARCH}.deb
#------------------------------------------------------------------------
ENV PYTHONPATH=/opt/src:$PYTHONPATH
ENV PYTHONPATH=/opt/src/rastervision_aws_batch/:$PYTHONPATH
ENV PYTHONPATH=/opt/src/rastervision_aws_s3/:$PYTHONPATH
ENV PYTHONPATH=/opt/src/rastervision_core/:$PYTHONPATH
ENV PYTHONPATH=/opt/src/rastervision_gdal_vsi/:$PYTHONPATH
ENV PYTHONPATH=/opt/src/rastervision_pipeline/:$PYTHONPATH
ENV PYTHONPATH=/opt/src/rastervision_aws_sagemaker/:$PYTHONPATH
ENV PYTHONPATH=/opt/src/rastervision_pytorch_backend/:$PYTHONPATH
ENV PYTHONPATH=/opt/src/rastervision_pytorch_learner/:$PYTHONPATH
COPY scripts /opt/src/scripts/
COPY scripts/rastervision /usr/local/bin/rastervision
COPY tests /opt/src/tests/
COPY integration_tests /opt/src/integration_tests/
COPY .flake8 /opt/src/.flake8
COPY .coveragerc /opt/src/.coveragerc
COPY ./rastervision_aws_batch/ /opt/src/rastervision_aws_batch/
COPY ./rastervision_aws_s3/ /opt/src/rastervision_aws_s3/
COPY ./rastervision_core/ /opt/src/rastervision_core/
COPY ./rastervision_gdal_vsi/ /opt/src/rastervision_gdal_vsi/
COPY ./rastervision_pipeline/ /opt/src/rastervision_pipeline/
COPY ./rastervision_aws_sagemaker/ /opt/src/rastervision_aws_sagemaker/
COPY ./rastervision_pytorch_backend/ /opt/src/rastervision_pytorch_backend/
COPY ./rastervision_pytorch_learner/ /opt/src/rastervision_pytorch_learner/
# needed for this image to be used by the AWS SageMaker PyTorch Estimator
RUN uv pip install sagemaker_pytorch_training==2.8.1
ENV SAGEMAKER_TRAINING_MODULE=sagemaker_pytorch_container.training:main
# Install a onnxruntime-gpu version compatible with CUDA 12. Specifying
# --extra-index-url in requirements.txt seems to cause problems with the
# RTD build.
RUN if [ "${TARGETARCH}" != "arm64" ]; then \
uv pip install --upgrade onnxruntime-gpu==1.17 --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/; fi
CMD ["bash"]