PESTPP-IES : What are my options to decrease RAM usage ?

[BJEANNOT0]

Hello,

As my problem gets bigger and bigger, I am reaching a situation where my manager (which is on a computer with no agents running) gets out of memory issues. As a consequence, i wonder if there are smart options I did not think about that would help me reduce memory consumption.
-I think that options about binary format only save disk memory, not RAM, and are therefore not a solution to my problem.
-There is the option Ies_upgrades_in_memory that seems interesting. Based on what I read on the manual I understand I should set it to False. Am I correct ?
-Finally, the option ensemble_output_precision might be interesting, But I am afraid that reducing the default value would hinder the inversion process.

Do you have any comments on my problem ?
Thank you for your help.
Best regards,

[jtwhite79]

How many pars and obs do you have? The ies_upgrades_in_memory will help if you have lots of pars - it essentially stores the candidate parameter ensembles to disk during lambda testing rather than holding them in memory. The other way to keep the memory usage down is to use shorter par and obs names because these names get used in lots of different containers under the hood so keeping them short can help with memory, esp if you have lots of pars and obs...

[BJEANNOT0]

Thank you for your answer !
I have about 1k pars but about 320k obs for calibration and another 160kobs that are zero-weighed and only used for validation purposes.
I will try shortening the name of observations as you suggest.
Also, I understand that because i do not have that much pars, ies_upgrades_in_memory will not help that much.
Thanks again.
Cheers,

[jtwhite79]

with that many observations, ies will use a lot memory...

[BJEANNOT0]

Indeed ! I try to upscale the most possible but I might have to downscale a bit