2019-05-15 Brown Dog Meeting notes

absent

Luigi is having issues running transformation catalog - needs to talk to Yan

Galaxy Swarm - was able to run as independent program and simple extractor

Working on it now as a docker

We won't maintain image set

Tensor flow on Nebula cluster - install ver 1.5 to work

Naming: Extractor Galaxies - https://opensource.ncsa.illinois.edu/bitbucket/projects/CATS/repos/extractors-galaxies/browse\https://bengalewsky.github.io/openstack/docker/2018/12/10/BuildingTensorFlowForOpenstack1.html

Looking at Yan's PR - looking at flask

Don't spend too much time without asking for help

trying to get a simple python script to call scala code and call simple extractor

getting errors - possibly related to docker - no java in the docker

Rob explained a process to go forward

docker compose is old in extractor template

https://opensource.ncsa.illinois.edu/bitbucket/projects/CATS/repos/pyclowder2/raw/docker-compose.yml
Greg, you can use this docker compose file. We need to update the one in the extractor template project. (Thanks, will try it. - GJ)

Finished code for punchcard extractor, but got blocked for a while on testing setup. Should have it tested out this week and ready for push and inclusion. (Had forgotten about some Linux-specific docker networking I have had to do in the past.)
I'll also still need to add a callback for contributing file-level extractions of card text to a dataset level aggregate.

Making training data set into tensor flow format - did 100 training images

got access to deep learning machine at NCSA

avg time is 10 sec per step

label based on their labeling in the shape file - train for number to see if it is faster