NaN vs. 1.e20 issue on ciclad with CDMS

eguil / Density_bining

Density bining code

2 stars 5 forks source link

Closed eguil closed 5 years ago

eguil commented 6 years ago

mme_ave_msk.py is creating files with NaNf instead of 1.e20 for valmask.
Using the exact same files and python routines on Eric's mac does not produce this problem
Try reinstalling CMDS via metrics package on ciclad.

durack1 commented 6 years ago

@eguil let me know about this because it could be a cdms bug and we'll need to create an issue in that repo - note that the latest version of CDMS is 2.12 (see https://anaconda.org/conda-forge/cdms2/files?channel=main)

I have had issues in the past with regard to NaN numpy values

eguil commented 6 years ago

Installed on Eric's mac cdms2, cdutil etc. via conda install <pckg> -c conda-forge for local testing vs. ciclad