FROM nvcr.io/nvidia/pytorch:22.01-py3

ENV HOME=/root
ENV LC_ALL C.UTF-8
ENV LANG C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive

# APP_PATH: assumed to be volume mounted with host
ENV APP_PATH=$HOME/np_app_text_retrieval_inference
ENV PYTHONPATH $APP_PATH

# EXTERNAL_PATH: assumed having system specific files, not volume mounted with host
ENV EXTERNAL_PATH=/workspace/external
RUN mkdir -p $EXTERNAL_PATH

WORKDIR $APP_PATH
COPY . $APP_PATH/

### ubuntu package
RUN apt-get update
RUN apt-get install git vim unzip wget ffmpeg libsm6 libxext6 -y

### conda
# note that retriever (based on pyserini) support cpu only for faiss
RUN conda install -c conda-forge faiss-cpu

### pip
# install editable mode
RUN pip install --upgrade pip && \
    pip install --no-binary :all: nmslib && \
    pip install -e .

# update java version >= 11
WORKDIR $EXTERNAL_PATH
RUN cp $APP_PATH/download_oracle_jdk.sh $EXTERNAL_PATH/
RUN sh download_oracle_jdk.sh
RUN ln -sf ${EXTERNAL_PATH}/jdk-20.0.2/bin/java /opt/conda/bin/java
ENV JAVA_HOME=${EXTERNAL_PATH}/jdk-20.0.2

# anserini: need for sparse retrieval/indexing
RUN conda install -c conda-forge maven
RUN mvn -v
RUN git clone https://github.com/castorini/anserini.git
RUN cd anserini && \
    mvn clean -Dmaven.test.skip=true package appassembler:assemble

# copy fatjar from anserini to app
# will be done at entrypoint
#COPY $EXTERNAL_PATH/anserini/target/anserini-*-fatjar.jar $APP_PATH/pyserini/resources/jars

## entrypoint: call entrypoint.sh instead
WORKDIR $APP_PATH
#CMD ["bash"]