Running Keras-Tensorflow Deep Learning (TFoS) on Hadoop Distributions (CDH)

I successfully managed to test and setup TensorflowOnSpark on our Cloudera environment. Following these links:
TFoS Yarn Setup
Resolved Issues

I now want to design and push all our deep learning jobs onto our Cloudera infrastructure. I did not find any Spark deep learning nodes. But I did find Spark and Deep Learning Network nodes, Is there a way to combine these two types and make a example for TensorflowOnSpark ?

For example: Is there a way to run “Train_MNIST_Classifier” example, available in the KNIME example server to run on Hadoop.

Any help appreciated. Thanks !

Hi mohammedayub,

There is no way I could think of to combine those nodes, unfortunately.
Nonetheless, we’re definitely interested in supporting deep learning in a distributed setup in the future, I can’t make any promises, though.


Thanks Marcel. Looking forward to the updates from your team. We can definitely pilot test some of that work on our CDH clusters.