Run Rapidminer algorithms (experiments) in MIP platform.
For the moment only classification algorithms...
It can run any RapidMiner experiments (based on classifier) as long as the .rmp template is present in the data/in/template volume folder
Use this image as the parent image to adapt a RapidMiner algorithm to the MIP platform:
Dockerfile
FROM hbpmip/java-base-build:3.6.0-jdk-11-0 as build-java-env
COPY pom.xml /project/pom.xml
COPY src/ /project/src
# Repeating the file copy works better. I dunno why.
RUN cp /usr/share/maven/ref/settings-docker.xml /root/.m2/settings.xml \
&& mvn package site
FROM hbpmip/java-rapidminer:0.9.4
MAINTAINER <your email>
ENV DOCKER_IMAGE=hbpmip/my-algo:1.0.0 \
JAVA_CLASSPATH=${JAVA_CLASSPATH}:/usr/share/jars/my-algo.jar \
JAVA_ARGS=/eu/humanbrainproject/mip/algorithms/rapidminer/myalgo/settings.properties \
MODEL=myalgo \
FUNCTION=java-rapidminer-myalgo
COPY --from=build-java-env /project/target/my-algo.jar /usr/share/jars/my-algo.jar
COPY --from=build-java-env /project/target/site/ /var/www/html/
COPY src/ /src/
RUN chown -R compute:compute /src/ \
&& chown -R root:www-data /var/www/html/- Parse the RapidMiner experiment template name from the algorithm field of the JSON input...
- Better validation input data types (classification, using real values features for now!)
- Add action section in PFA. Train the model using all the data as training set (separate RapidMiner process)