Urika 1.2

Machine Learning

Keras

Information on how to use Keras with the Urika-XC 1.2 test release

To run using start_analytics (interactive)

ssh gateway2
module use /project/s1001/urika_training/uxc_1.2/modulefiles
module load analytics/1.01.0000.201810291633_0135
salloc -N 1 -p debug
srun --pty bash -i
start_analytics

To run the MNIST example from the Keras examples directory:
Here we can use the sameple py36_tf_cpu conda environment with Keras 2.2.4
already installed.

source activate py36_tf_cpu
python mnist_cnn.py

More Keras examples corresponding to Keras version 2.2.4 can be found here
https://github.com/keras-team/keras/tree/2.2.4/examples

This can also be easy run using the run_training script

ssh gateway2
module use /project/s1001/urika_training/uxc_1.2/modulefiles
module load analytics/1.01.0000.201810291633_0135
salloc -N 1 -p debug

run_training -n 1 "python mnist_cnn.py"

By default this will use the optimized Tensorflow corresponding to the node
type (cpu or gpu).