As expected, you don't really need all those steps. Here's a simplified version:
- Load puzzle, set low wiggle.
- SWSW on low.
- Refine density.
- SWSW on low.
- Refine density.
The score once again drops on step 5, and doesn't recover.
Restore credit best gives back the step 4 score, but it's no longer stable. The score drops to the step 5 score.
A manual save after step 4 is stable when restored.
It seems like the shaking and wiggling in step 4 is needed. Without the SWSW, a second refine density doesn't drop the score, and may actually improve it. Then shaking and wiggling continues to add points. Something about collapsing the wave I guess.
hi @LociOiling thank you for taking the time to put these steps together. i'll try it out now on both my windows and linux machines to see if I can replicate it this way. The bug with the previous devprev release had to do with quicksaves, and now based on your feedback there appears to alternatively now be an issue with certain types of autosaves.
to be clear, the additional steps after refine density are just trying to restore credit best or some other type of autosave?
After the first refine density, shaking and wiggling is necessary to create the problem. The actual problem doesn't appear yet.
Then after the second refine density, restore credit best seems to work, but then the score drops and doesn't recover. The same applies to restore very best.
I think recent best may also be affected, but I don't have exact steps to recreate.
Hi @LociOiling please let us know if you still find this issue in the latest DevPrev!
Still early going, but no problems so far. After some rebuilding, Refine Density produced a nice gain.
I'll do another Refine Density a little later. That's where problems showed up previously.
After wiggling out and fusing the first Refine Density, a second Refine Density dropped the score by over 250 points. Restore Credit Best gave back the previous score, which was still stable (not changed by shake + wiggle).
I'd say the issue with restore credit best has been resolved. Restore Very Best and Restore Recent Best also seem to be working as expected.
A third Refine Density on top of the second Refine Density gained a few points, but didn't get close to the high score. Restore credit best once again gave back a stable high score.
I'll continuing working on the puzzle in the usual way. So far manual saves seem to be working, so I'll look at saving, sharing, and then loading on a different client.
So far, I haven't found any problems with an unstable pose.
I'm trying a restart, where I plan to Refine Density near the end.
Initially, I rebuilt the protein for a while on low power (TvdL Enhanced DRW), then did Refine Density, then moved on to finishing steps on high power. That worked well, so we'll see how refining later compares.
Correction: Thanks to Loci's remark below regarding the version, I re-checked my version and it seems I was using version 33 Windows for this report. The solution saved apparently contains the correct score, but client 33 misread it. In version 34 this seems to be corrected. Apologies for the confusion and thanks to Loci for reminding me to check the version.
I partially agree with LociOiling. I am seeing the score drop even if no trim is used. restore credit best does not restore the indicated max score. A save that is done from one client with the shown max score does not reload into a different client with the indicated score. Solution 'inter34345' shared with scientists, which should have 34345 points (as indicated in the saved solution) loads with 34118 points.
Just for reference, I'm running release 34 on Windows, V34-20240506-85f0ab2ee2-win_x64-devprev. But I'm guessing the current test puzzle is only available on that release.
