Initial parallel computation wrokflow for ML model optimization

NLeSC / team-atlas

1 stars 0 forks source link

Initial parallel computation wrokflow for ML model optimization #44

Open rogerkuou opened 4 years ago

rogerkuou commented 4 years ago

As the PO of Vegetation project, I would like to have an initial parallel computation workflow for the existing ML model optimization work flow. We can start this by executing the Cartesius built-in parallel network training.

Acceptable criterion:

[ ] Get access to Cartesius which Manuel use
[ ] Reproduce the a small fraction of current CNN optimization workflow on Cartesius
[ ] Reproduce the a small fraction of current CNN optimization workflow on Cartesius, with parallel tranning.

rogerkuou commented 4 years ago

TU Delft use personal accesses to Cartesius therefore it is not possible to get quick access via them. I will first move this card to blocks. Instead I am going to do a local parallelization test with TensorFlow Multi-worker training solution