JQ

why in Discriminator, you didn't use batchnormalization？

There are some problems when training the example

4

When I run this: >spark@master:~$ spark-submit --master yarn --deploy-mode cluster \ > --num-executors ${SPARK_WORKER_INSTANCES} \ > --files ${CAFFE_ON_SPARK}/data/lenet_memory_solver.prototxt,${CAFFE_ON_SPARK}/data/lenet_memory_train_test.prototxt \ > --conf spark.driver.extraLibraryPath="${LD_LIBRARY_PATH}" \ > --conf spark.executorEnv.LD_LIBRARY_PATH="${LD_LIBRARY_PATH}" \ > --class com.yahoo.ml.caffe.CaffeOnSpark...