Cray Machine Learning

From Research Computing @ TAMUQ
Jump to navigation Jump to search


Setup Conda Environment

We will set up a Virtual Conda environment which will be used later for running different BigData/ML applications. The list of packages to be installed in this Virtual environment is provided in alldemo.yml file. This file lists all the packages which may be required for demo purposes of subsequent applications. You can modify alldemo.yml file based upon your requirements.


muarif092@raad2-login2:~> module load analytics

muarif092@raad2-login2:~> cp /lustre/share/urika_training/uxc1.2/Anaconda/alldemo.yml .

muarif092@raad2-login2:~> salloc -N 1 --account=ut --reservation=ut --partition=l_long --qos=lcustom3 start_analytics -d

muarif092@nid00248:~$ conda env create -f alldemo.yml

muarif092@nid00248:~$ exit

Conda environment has been created and can be used in later examples as required.

Keras

Interactive

module load analytics
salloc -N 1 -p debug
srun --pty bash -i
start_analytics


Batch

Keras Dist. Training

PyTorch

PyTorch Dist. Training

TensorFlow Dist. Training

==