Loading...

​​​🕘​ Meeting notes: Machine Learning WG

​​Jan 6, 2019

​​Attendees

​​​@Joe H​ 

​​​@David G​ 

​​

​​Quick update:

​​

• ​​Joe:

• ​​Proposals

• ​​Still working on Met disaggregation ML work

• ​​David:

• ​​Getting ready for AMS

• ​​Proposals

• ​​Jeff:

• ​​Working on continental scale streamflow prediction, finishing up data prep

• ​​Zhonghua

• ​​https://github.com/pangeo-data/ml-workflow-examples/blob/master/AWS-S3-Dask-XGBoost.ipynb

• ​​working on climate temperature emulator

• ​​using xgboost

​​

​​Agenda:

​​

• ​​NVIDIA Jlab extension for GPUs

• ​​
  

​​

​​Action Items:

​​Dec 2, 2019

​​Attendees

​​​@Noah B​ 

​​​@Ryan A​ 

​​​@Joe H​ 

​​​@Tom​

​​​@Sebastian A​ 

​​

​​Quick update:

• ​​Noah: AGU next week, presentations on ML for climate modeling on Thursday

• ​​Ryan: 

• ​​Changes to data catalogs, turning off free egress (requester pays)

• ​​Still working on ML environment in the cloud

• ​​Tom: gcsfs now supports requester pays and fallback project auth. 

• ​​Sebastian: new data scientist at UW, a bit of ML background, going to work on GPUs on the cloud

• ​​Joe

• ​​ML Cloud Environment

• ​​https://github.com/pangeo-data/pangeo-stacks/pull/83

• ​​https://github.com/pangeo-data/pangeo-cloud-federation/issues/425

• ​​Casper JupyterHub - https://jupyterhub.ucar.edu/dav-dev, GPUs on NCAR super computer in Jupyter

• ​​ML for met disaggregation: Looking into best practices with packaging trained ML models in a python api.

• ​​Tom pointed us to ONNX which looks promising.

• ​​https://github.com/onnx/keras-onnx