Symmetry Mode Refinements

antons-l

Hello to all Topas users!
I am extremely interested in using symmetry mode refinements for my research work. I have already performed all tutorials on this topic and found out the structure for my material using modified structure of the parent compound. However the number of symmetry modes is too high (ca.400), so I tried to select the most important modes by following the tutorial #40 at https://topas.webspace.durham.ac.uk/topas_user_menu/ .
I repeated all the steps for WO3 in the tutorial and it worked well (though I used Anaconda3 for python script). Then I tried to do the same with my own data and end up with the error that repeats continuously in the anaconda terminal:

"WARNING: Individual did not converge - Retrying. If this continues then consider restarting kernel and looking at this individual refinement manually."

If I refine the data individually (using .inp file) the refinement converges, though only is I add randomize_on_error (which for some reason wasn't used in .inp0 file in WO3 tutorial). Additionaly, if I refine the data with all symmetry mode parameters switched on it converges as well.
I would be grateful if someone could help me to understand how to proceed on and to fix this issue.

P.S. For my refinement I used only PDX data (synchrotron data of good quality) and I deleted TOF part of code from the tutorial template.

Anton

johnsoevans

Anton,

I'm travelling and haven't looked at this tutorial for a while! You shouldn't need the randomize_on_errors as all the update values you need should be handled by the drand macro. I'd remove that line.

I think the error you describe comes if the fitness factory python script doesn't find a file called "inpfile.txt" which would have been created when topas ran. Are you sure that your command to call tc.exe is correct? Another possibility is that you have tc.exe running already (or perhaps a hung version - check task manager) so it won't run for licence reasons. If you type "C:\topasa_v6\tc.exe inpfile.inp" at a command prompt does it work?

John

antons-l

Thank you for the quick reply, John!

Strangely, I can't find any files "inpfile.txt" or "inpfile.inp" in my topas folder even if I look for hidden files or run random topas refinement or python script with WO3 example.
The path to tc.exe file in this folder for my computer is C:\TOPAS6\tc.exe. I typed - "C:\TOPAS6\tc.exe" inpfile.inp - in command terminal but all I get is "Cannot open file inpfile.inp". (the path C:\topasa_v6\tc.exe can't be found).

I think the command to call tc.exe is correct because I use exactly the same string to call tc.exe in WO3 example and all works well. It isn't hunging topas, I checked task manager, rebooted laptop, cleared cash etc. and WO3 example works even if I run another topas window simultaneously.

How do you think, could it be a problem with too-far-from-solution initial values of irreps? Maybe I need to start with some initial values that give more or less good fit?
Anton

antons-l

Finally, I found a solution. The solution to my problem was a choice of correct fitness function. I don't know why but for my compound only "onemax" fitness function works, any other returns that annoying message (Individual didn't converge). Hope it will help someone with the same problem.
Now it would be nice to know what is the difference between all that fitness functions :)

Best wishes,
Anton S-L

johnsoevans

Anton: glad you found an answer!

alexb

Dear both,

I have recently been attempting to apply the GA tutorial to a research problem too; like Anton above I also found that I have to use the 'onemax' fitness function, but with that change made I run into a second problem and am hoping you might have some insight.

My low-symmetry model has 54 modes (ie less than the 96 of WO3), and I have replaced the WO3 modes in the get_irrep_block_lists.py file, but when I run gacontrol.py the 'best modepool' includes modes numbered 55-96 (ie it can't be using the modes of my model). If I amend wo3modelsize = 96 #number of alelles per individual in gacontrol.py to a value of 54 then the modes in best modepool are numbered =<54 but the fitness/Rwp are sustantially worse than the fit of my parent model, so I have a gut feeling that gacontrol is still not using the 'right' modes. So my question is how to get gacontrol to do so. Are there other sections in the code that need updating to replace the tutorial data with my model? (I'm rather stumped in terms of understanding what the parameters under # mag specific information and # Apply problem-specific information are doing so it's entirely possible I've just missed something obvious!).

Thanks in advance,

Alex