Skip to content

Latest commit

 

History

History
29 lines (20 loc) · 796 Bytes

README.md

File metadata and controls

29 lines (20 loc) · 796 Bytes
Raw

flink-mnist

Run MNIST inference in Apache Flink

Usage

  1. Package the project.
mvn clean package # you could get MNISTInference.jar in target/ folder
  1. Download the native libraries of JCuda and JCublas to lib/ folder of your Flink distribution
./add-jcuda-dependency.sh <your_flink_dist>

  1. Prepare image data. You could download the training data or test data. Then, decompress it by gzip xxx.gz.

  2. Configure to enable GPU plugin, please refer to user document.

  3. Start the Flink cluster and run the MNIST inference job.

cd $FLINK_HOME
bin/start-cluster.sh
bin/flink run path/to/MNISTInference.jar --image-file path/to/imagefile