Author Topic: newest openquake implementation  (Read 1370 times)

Anne Hulsey

  • Jr. Member
  • **
  • Posts: 70
    • View Profile
newest openquake implementation
« on: September 21, 2021, 08:45:19 PM »
Hi Kuanshi,

Now that I have confirmed that openquake is producing the same results, I'm switching to using the standard gmms that the New Zealand NSHM is going to use (not the custom GNS ones I recently sent you). In particular, the Kuehn, Parker, and Abrahamson&Gulerce subduction models from 2020, all available here: https://github.com/gem/oq-engine/tree/master/openquake/hazardlib/gsim.

They were not automatically included in the openquake version that was packaged in the R2D build so I copied them over. However, openquake recently reorganized somethings and it looks like the R2D build cannot accommodate them. (For example, they require the base.add_alias function from the same folder I linked above.)

Would it be possible to include the openquake version that is on github, instead of what we are currently using? (I recall at one point I asked you to go back to a previous version - maybe that is what is included now.) Unfortunately, replacing the openquake folder wasn't sufficient because FetchOpenQuake calls openquake.commands.run.run, which is no longer included.

Thanks,
Anne
« Last Edit: September 21, 2021, 08:59:09 PM by Anne Hulsey »

kuanshi

  • Jr. Member
  • **
  • Posts: 54
    • View Profile
Re: newest openquake implementation
« Reply #1 on: September 22, 2021, 12:10:08 AM »
Hi Anne,

Yes, OpenQuake modified the main run command after 3.11.x and also reorganized "datastore" module (from baselib to commonlib) and a few other log methods since 3.12.x... So, I'm actually working on this to make the R2D compatible with released versions since 3.10.x.

In addition to this, I wonder if it's common that users may also like to have their own instance of OpenQuake (with unique local features) - if that is the case I can also add an option to use the local instance instead of the official release.

Many thanks,
Kuanshi

Anne Hulsey

  • Jr. Member
  • **
  • Posts: 70
    • View Profile
Re: newest openquake implementation
« Reply #2 on: September 22, 2021, 06:19:26 PM »
Hi Kuanshi,

Nice, thanks for having already been proactively thinking about/working on this! I don't know how most uses interact with openquake/local features. But I can definitely imagine it - in my case, that would mean having the custom gmms added in the local openquake version, without needing to remember to add them in the R2D version. Or having two local versions - one so I can replicate Nick's results on the old version and one to replicate the NSHM results on the new version.

One of the biggest benefits that I see is that it might allow me to switch back and forth between running openquake from the command line and from R2D, without needing to delete the db.sqlite3 file. Do you think that would be possible with what you are describing? For me, the biggest benefit of R2D will be when I can run it on DesignSafe for huge jobs, while there are helpful things about running openquake locally without R2D (such as being able to easily manipulate the .ini file, e.g., listing multiple poe values).

Thanks,
Anne

kuanshi

  • Jr. Member
  • **
  • Posts: 54
    • View Profile
Re: newest openquake implementation
« Reply #3 on: October 06, 2021, 07:13:17 PM »
Hi Anne,

Sorry about the late response! Yes, I added a version selection (3.10, 3.11, and 3.12) for OpenQuake in the latest developer built (https://drive.google.com/file/d/1zDj_xFEaXqLyuK4yB2zkq0LZUZACYXQz/view?usp=sharing) - when switching between versions the old db.sqlite3 file would be removed automatically by the backend before launching a new run.  Please feel free to test it and let us know if you have any questions or comments as usual.

I'm still working on to enable running openquake on design-safe to support your need and would get you back for testing and feedback once it is completed.

Regards,
Kuanshi