DL4J Java - Working on a GAN - getting silent error in MultiLayerNetwork.class

Question

I started with this example: https://github.com/wmeddie/dl4j-gans/tree/master/src/main/java/io/skymind/example

I updated it to instead of using MNIST 28X28 grayscale it uses 100X100 RGB 3 channel images.

Flow:

1. Get data set iterator for real images
2. create the 3 models: Gen, Dis, Gan
3. iterate over training data
4. Create INDArray of real and fake and merge them
5. I verified that the real and fake both have the shape: [130,3,100,100]
6. Call fit() on the discriminator

inside the fit implementation for the discriminator I get a silent error/application crash:

Class:MultiLayerNetwork.class Function: fitHelper(INDArray features, INDArray labels, INDArray featuresMask, INDArray labelsMask) Line 2357: solver.optimize(workspaceMgr);

Tess4j: 5.8.0 deeplearning4j-core: 1.0.0-M2.1 dl4j-spark_2.12: 1.0.0-M2.1 nd4j-native-platform: 1.0.0-M2.1 deeplearning4j-nn: 1.0.0-M2.1

Thanks.

I didn't notice CPU/GPU/SSD/RAM near any limits.