## Docker GROBID image using deep learning models and/or CRF models
## See https://grobid.readthedocs.io/en/latest/Grobid-docker/
## usage example with version 0.8.0:
## docker build -t grobid/grobid:0.8.0 --build-arg GROBID_VERSION=0.8.0 --file Dockerfile.delft .
## no GPU:
## docker run -t --rm --init -p 8070:8070 -p 8071:8071 -v /home/lopez/grobid/grobid-home/config/grobid.properties:/opt/grobid/grobid-home/config/grobid.properties:ro grobid/grobid:0.8.0
## allocate all available GPUs (only Linux with proper nvidia driver installed on host machine):
## docker run --rm --gpus all --init -p 8070:8070 -p 8071:8071 -v /home/lopez/grobid/grobid-home/config/grobid.properties:/opt/grobid/grobid-home/config/grobid.properties:ro grobid/grobid:0.8.0
# -------------------
# build builder image
# -------------------
FROM openjdk:17-jdk-slim as builder
USER root
RUN apt-get update && \
apt-get -y --no-install-recommends install unzip
WORKDIR /opt/grobid-source
# gradle
COPY gradle/ ./gradle/
COPY gradlew ./
COPY gradle.properties ./
COPY build.gradle ./
COPY settings.gradle ./
# source
COPY grobid-home/ ./grobid-home/
COPY grobid-core/ ./grobid-core/
COPY grobid-service/ ./grobid-service/
COPY grobid-trainer/ ./grobid-trainer/
# cleaning unused native libraries before packaging
RUN rm -rf grobid-home/pdf2xml
RUN rm -rf grobid-home/pdfalto/lin-32
RUN rm -rf grobid-home/pdfalto/mac-64
RUN rm -rf grobid-home/pdfalto/mac_arm-64
RUN rm -rf grobid-home/pdfalto/win-*
RUN rm -rf grobid-home/lib/lin-32
RUN rm -rf grobid-home/lib/win-*
RUN rm -rf grobid-home/lib/mac-64
RUN ./gradlew clean assemble --no-daemon --info --stacktrace
WORKDIR /opt/grobid
RUN unzip -o /opt/grobid-source/grobid-service/build/distributions/grobid-service-*.zip && \
mv grobid-service* grobid-service
RUN unzip -o /opt/grobid-source/grobid-home/build/distributions/grobid-home-*.zip && \
chmod -R 755 /opt/grobid/grobid-home/pdfalto
RUN rm -rf grobid-source
# -------------------
# build runtime image
# -------------------
# use NVIDIA Container Toolkit to automatically recognize possible GPU drivers on the host machine
FROM tensorflow/tensorflow:2.7.0-gpu
# setting locale is likely useless but to be sure
ENV LANG C.UTF-8
# update NVIDIA Cuda key (following a key rotation in April 2022)
RUN apt-get install -y wget && \
apt-key del 7fa2af80 && \
rm /etc/apt/sources.list.d/cuda.list && \
rm /etc/apt/sources.list.d/nvidia-ml.list && \
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.0-1_all.deb && \
dpkg -i cuda-keyring_1.0-1_all.deb
# Add Tini
ENV TINI_VERSION v0.19.0
ADD https://github.com/krallin/tini/releases/download/${TINI_VERSION}/tini /tini
RUN chmod +x /tini
ENTRYPOINT ["/tini", "-s", "--"]
# install JRE, python and other dependencies
RUN apt-get update && \
apt-get -y --no-install-recommends install apt-utils build-essential gcc libxml2 libfontconfig unzip curl \
openjdk-17-jre-headless ca-certificates-java \
musl gfortran \
python3 python3-pip python3-setuptools python3-dev \
&& apt-get clean \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /opt/grobid
COPY --from=builder /opt/grobid .
RUN python3 -m pip install pip --upgrade
# install DeLFT via pypi
RUN pip3 install requests delft==0.3.3
# link the data directory to /data
# the current working directory will most likely be /opt/grobid
RUN mkdir -p /data \
&& ln -s /data /opt/grobid/data \
&& ln -s /data ./data
# disable python warnings (and fix logging)
ENV PYTHONWARNINGS="ignore"
WORKDIR /opt/grobid
ENV JAVA_OPTS=-Xmx4g
# install jep (and temporarily the matching JDK)
ENV JDK_URL=https://download.java.net/java/GA/jdk17.0.2/dfd4a8d0985749f896bed50d7138ee7f/8/GPL/openjdk-17.0.2_linux-x64_bin.tar.gz
RUN curl --fail --show-error --location -q ${JDK_URL} -o /tmp/openjdk.tar.gz && \
mkdir /tmp/jdk-17 && \
tar xvfz /tmp/openjdk.tar.gz --directory /tmp/jdk-17 --strip-components 1 --no-same-owner && \
/tmp/jdk-17/bin/javac -version && \
JAVA_HOME=/tmp/jdk-17 pip3 install jep==4.0.2 && \
rm -f /tmp/openjdk.tar.gz && \
rm -rf /tmp/jdk-17
ENV LD_LIBRARY_PATH=/usr/local/lib/python3.8/dist-packages/jep:grobid-home/lib/lin-64:grobid-home/lib/lin-64/jep:${LD_LIBRARY_PATH}
# remove libjep.so because we are providing our own version in the virtual env above
RUN rm /opt/grobid/grobid-home/lib/lin-64/jep/libjep.so
# preload embeddings, for GROBID all the RNN models use glove-840B (default for the script), ELMo is currently not loaded
# to be done: mechanism to download GROBID fine-tuned models based on SciBERT if selected (but not good enough for the moment)
COPY --from=builder /opt/grobid-source/grobid-home/scripts/preload_embeddings.py .
COPY --from=builder /opt/grobid-source/grobid-home/config/resources-registry.json .
RUN python3 preload_embeddings.py --registry ./resources-registry.json && \
ln -s /opt/grobid /opt/delft
RUN mkdir delft && \
cp ./resources-registry.json delft/
ENV GROBID_SERVICE_OPTS "--add-opens java.base/java.lang=ALL-UNNAMED"
CMD ["./grobid-service/bin/grobid-service"]
ARG GROBID_VERSION
LABEL \
authors="The contributors" \
org.label-schema.name="GROBID" \
org.label-schema.description="Image with GROBID service" \
org.label-schema.url="https://github.com/kermitt2/grobid" \
org.label-schema.version=${GROBID_VERSION}