- 24 Feb, 2022 5 commits
-
-
Clément Haëck authored
-
Clément Haëck authored
Place option type in conf file as a string. Add support for default fixes. Add comments in example conf file.
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
set of years could be unsorted
-
- 18 Feb, 2022 2 commits
-
-
Clément Haëck authored
-
Clément Haëck authored
-
- 17 Feb, 2022 7 commits
-
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
fix mix-up in masks add some print statements specify default proba mask threshold
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
- 16 Feb, 2022 5 commits
-
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
so they are in output string order
-
- 04 Feb, 2022 10 commits
-
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
For some reason, having a mask (frt and bkg) with missing values does not play nicely with the rest of the script. Would get wrong values. Simple logical operation works better. (confirmed I get exact same result with the previous method, just faster)
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
All zones and variables in a single file Less files for cluter, but most importantly all computations are done in a single execution of the compute_hists script. A lot less overhead (especially on ciclad where it can take time), and a better usage of dask delayed computation gives a speed up of about 6 times faster (!). This adds complexity to lib.data.hists functions which now have to deal with different variables histograms in the same dataset. I tried to refactor all loops on variable in a decorator. It is overly complex and quite experimental but is able to deliver wanted results with flexibility.
-
Clément Haëck authored
-
- 03 Feb, 2022 1 commit
-
-
Clément Haëck authored
-
- 02 Feb, 2022 4 commits
-
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
Clément Haëck authored
-
- 01 Feb, 2022 1 commit
-
-
Clément Haëck authored
-
- 28 Jan, 2022 1 commit
-
-
Clément Haëck authored
no need to include its computation in the loop
-
- 26 Jan, 2022 2 commits
-
-
Clément Haëck authored
-
Clément Haëck authored
-
- 25 Jan, 2022 2 commits
-
-
Clément Haëck authored
rather than put mode to 1 (which fix the low part of the distribution at 1), we put the 95% quantile at HI=9.5 (which fix the tail of the HI values distribution)
-
Clément Haëck authored
Compare to underlying gaussian distribution using trapezoidal rule essentially
-