Add QDEIM option to some examples #229
Conversation
|
@dylan-copeland is this ready for review? |
I am planning to add a hyperreduction function that will reduce the boilerplate code for the various types of hyperreduction methods. |
…of hyperreduction sampling functions.
dylan-copeland
requested review from
chldkdtn,
dreamer2368 and
siuwuncheung
| // | ||
| // Online phase with full sampling: | ||
| // ./nonlinear_elasticity_global_rom -online -dt 0.01 -tf 5.0 -s 14 -vs 100 -hyp -rvdim 40 -rxdim 10 -hdim 71 -nsr 1170 -sc 1.00 | ||
| // ./nonlinear_elasticity_global_rom -online -dt 0.01 -tf 5.0 -s 14 -vs 100 -hyp -hrtype gnat -rvdim 40 -rxdim 10 -hdim 71 -nsr 1170 -sc 1.0 |
There was a problem hiding this comment.
I added this to expose the api
There was a problem hiding this comment.
Looks good in general. Tried the nonlinear elastic examples on my local machine and the results are as before. Varying the hyperreduction options works. I added some lines to the nlelast examples so that the new api gets shown.
Not a big fan of this many formatting changes in a PR. It makes it harder to read and review.
|
Sorry for the unnecessary commits. I thought that I committed on another branch. |
|
This PR needs 1 more review @chldkdtn @siuwuncheung @dreamer2368 |
| args.AddOption(&use_sopt, "-sopt", "--sopt", "-no-sopt", "--no-sopt", | ||
| "Use S-OPT sampling instead of DEIM for the hyperreduction."); | ||
| args.AddOption(&samplingType, "-hrtype", "--hrsamplingtype", | ||
| "Sampling type for hyperreduction."); |
There was a problem hiding this comment.
I recommend adding different hyper-reduction sampling types in the regression test for this example, similar to the mixed nonlinear diffusion.
| // mpirun -n 1 ./mixed_nonlinear_diffusion -online -rrdim 8 -rwdim 8 -nldim 20 -nsdim 20 -sopt | ||
| // mpirun -n 1 ./mixed_nonlinear_diffusion -online -rrdim 8 -rwdim 8 -nldim 20 -nsdim 20 -hrtype deim | ||
| // mpirun -n 1 ./mixed_nonlinear_diffusion -online -rrdim 8 -rwdim 8 -nldim 20 -nsdim 20 -hrtype qdeim | ||
| // mpirun -n 1 ./mixed_nonlinear_diffusion -online -rrdim 8 -rwdim 8 -nldim 20 -nsdim 20 -hrtype sopt | ||
| // mpirun -n 1 ./mixed_nonlinear_diffusion -online -rrdim 8 -rwdim 8 -nldim 20 -nsdim 20 -ns 1 -eqp | ||
| // | ||
| // Relative l2 error of ROM solution at final timestep using DEIM sampling: 1.096776797994166e-08 |
There was a problem hiding this comment.
I believe this result corresponds to -hrtype deim? Perhaps we should report the example results for Q-DEIM and S-OPT here and in the libROM webpage.
There was a problem hiding this comment.
My first thought was not to add it, because the QDEIM result is almost the same as DEIM. For completeness, I went ahead and added it. I agree that the webpage should be updated as well, to let users know about the QDEIM option.
I actually updated all the run times. Reporting run times is troublesome, because they vary even on the same machine. When a new result is added, all runs have to be updated for a fair comparison.
| // Output message: | ||
| // Elapsed time for time integration loop 1.80759 | ||
| // Relative error of ROM position (x) at t_final: 5 is 0.000231698 | ||
| // Relative error of ROM velocity (v) at t_final: 5 is 0.466941 | ||
| // | ||
| // Online phase with strong hyper-reduction: | ||
| // ./nonlinear_elasticity_global_rom -online -dt 0.01 -tf 5.0 -s 14 -vs 100 -hyp -rvdim 40 -rxdim 10 -hdim 71 -nsr 100 -sc 1.00 | ||
| // ./nonlinear_elasticity_global_rom -online -dt 0.01 -tf 5.0 -s 14 -vs 100 -hyp -hrtype gnat -rvdim 40 -rxdim 10 -hdim 71 -nsr 100 -sc 1.0 |
There was a problem hiding this comment.
Perhaps we should report the example commands and results for Q-DEIM and S-OPT here and in the libROM webpage.
No description provided.