how to tell how long it will take?

crest-lab / crest

CREST - A program for the automated exploration of low-energy molecular chemical space.

https://crest-lab.github.io/crest-docs/

GNU Lesser General Public License v3.0

182 stars 42 forks source link

how to tell how long it will take? #225

Closed jouteurestaque closed 3 months ago

jouteurestaque commented 9 months ago

Hello I have for the first time perform a CREST on a very big flexible molecule (lipid derivative), it is running for more that 8 dyas on a 36 CPU clusters. is There any chances I can tell how low it is supposed to last? Seems I am on a loop cause it is finding lower conformers all the iteration time...

and anyhow can it be restart somehow?? I don't think I ever found that? can it be parallelized with several nodes? can't retrieve that either//

cheers

crestsolong.zip

pprcht commented 9 months ago

No, there is no way to tell a priori how long the calculation will take. With the system size and flexibility you are looking at it is not unexpected to have quite long runtimes. If you are using a queuing system like slurm, keep in mind that CREST (at least in the current main version) produces a lot of I/O that can slow down the calculation if you are not running it on the node scratch.

There is no dedicated restart function so far (I'm working on it) but you can use the intermediate ensembles of course. Likewise, multi-node parallelization is not available, the code is not (yet) optimized for MPI.

jouteurestaque commented 9 months ago

Thank you Philipp yeah it is on the scratch indeed. ok I will pray it finishes before the 20th..thak you so much.. by te way I never say how good that CREST/CENSO thing is at finding the good conformers for using 1H NMR prediction,,but it is really great, I am gonna have to train the NMR guy at the univ cause he is asking me all the time, do you think your soft can solve that problematic...and so far it never failed...

Ever wonder how to predict your 1H NMR with something powerful and free. Use the Combo from Grimme/Neese groups,the CREST/CENSO/ORCA/ANMR @grimme @faccts_orca .
a recent example how everything is down to conformers and optimization of the ensemble. pic.twitter.com/NvizyvF62a
— Galano Jean-Marie (@JeanGalano) June 30, 2023

github-actions[bot] commented 3 months ago

This issue had no activity for 6 months. It will be closed in 1 week unless there is some new activity.