Anki 23.12 Beta

[Expertium]

Bug: when “Limits start from top” is enabled, a card from the parent deck can appear when reviewing cards from a subdeck.
Suppose my decks look like this: ParentDeck::SubDeck. I am reviewing cards from the SubDeck. When “Limits start from​ top” is enabled, I see a card that belong​s to ParentDeck, not to SubDeck.​

[Freezable0596]

Bug on ARM MacOS 14.2.1 with RC1 as mentioned by @medanki in previous versions. Every time I try to use the bar at the top to get the browse screen Anki Crashes. This doesn’t happen if I press B from the Home Screen but if I click on a deck or synchronise it does too. If I’ve clicked once on any button on the anki window B will cause it crash.

This happens after completing all the steps in the troubleshooting guide.

[sprvlcn]

I’ve been using FSRS (without the Helper add-on, at default 0.90 retention) since 23.10 came out. Have optimized a few times since then.

Today I gave “Reschedule cards on change” a try for the first time. Despite the warning that a large number of cards could become due, I was surprised that no cards at all became due (I turned on Reschedule, clicked Optimize, clicked Save). I have tens of thousands of cards, and all decks use the same preset.

This isn’t a tech support request, I’m just posting here in case 23.12rc1 somehow accidentally broke “Reschedule cards on change” functionality. Highly unlikely, but maybe someone can do a quick check?

The yellow text says: “Use this option sparingly, as it will add a review entry to each of your cards, and increase the size of your collection.” But looking at Card Info for various non-new cards does not show any new review entries in the timestamped log listing at the bottom of the Info window.

[Expertium]

Have you tried changing desired retention and then turning on “Reschedule cards on change” and clicking “Save”?

[sprvlcn]

No, I kept the same retention.

Also, I recall that I had finished all my due cards (and none overdue), and then did an optimization without Reschedule enabled. Then a short time later I thought about giving Reschedule a try, and so did another optimization, but this time with Reschedule enabled. The parameters surely didn’t change at all (but I didn’t actually check this), because no new reviews had been done since the previous optimization a short time earlier. Maybe that was the issue?

[Expertium]

“Reschedule cards on change” only does things when you, well, change something. By “something” I mean either parameters or desired retention.

1. Turn “Reschedule cards on change” on
2. Change desired retention or parameters
3. Click “Save”

[sprvlcn]

OK, that’s probably it then. Thanks. I’ll give Reschedule another try when I do another optimization a few weeks from now.

[sprvlcn]

Speaking of rescheduling, Searching - Anki Manual should document resched: somewhere near where introduced: is documented.

I think Rescheduled is a fairly recent addition to the Today sidebar submenu? Which version added it? I couldn’t find that in Release notes or Changes.

[snowtimeglass]

In AnkiMobile 23.10.1 build 4, setting 0 second (the default value) to the Auto Advance options doesn’t seem to disable the auto advance feature, in contrast to the behavior on Anki Desktop and the explanation on the help screen; that is, just after starting the auto advance in AnkiMobile, many cards get shown and answered in an instant and then the congrat screen gets shown immediately.

[dae]

@Expertium

Another suggestion: to ensure that users cannot end up with worse performance after optimization, do the following:

It feels a bit complicated. What if log loss increases and RMSE decreases, or vice versa? And is it safe to always assume that the generated parameters are worse when those metrics increase? Would it be possible that certain changes to the parameters could lead to a situation where the average card was scheduled better, but RMSE increased?

@Aleksej

In a filtered back, pressing Enter in the Search field does not press the Rebuild button.

I can’t reproduce this. Which platform/Qt version?

@Aleksej

I have “Reschedule cards on change” disabled, and enabling it is not retained

If users set it and forget to turn it off, they’ll end up adding review log entries to their cards every time they optimize their parameters.

@Expertium

Deck size, Days to simulate and Minutes study/day should be remembered on a per-preset basis. It’s much more likely that the user will want to re-compute desired retention with the same values of these parameters than with the default values.

Maybe in the future, when that feature is no longer experimental.

@Gustaf-C

Minor issue (this might be old), the scroll bar in the editor slightly overlaps the sticky/html buttons, would be nice with some more space.

This requires some refactoring of our editing screens to fix, so is a low priority at the moment.

@Expertium

Bug: when “Limits start from top” is enabled, a card from the parent deck can appear when reviewing cards from a subdeck.

Thanks for catching this. I’ve logged it on Cards from parent decks are being included · Issue #2906 · ankitects/anki · GitHub

@Freezable0596

Bug on ARM MacOS 14.2.1 with RC1 as mentioned by @medanki in previous versions.

If resetting window sizes doesn’t help and neither did the other troubleshooting steps, the only other thing I can think of to try is the qt5 version.

@sprvlcn

I think Rescheduled is a fairly recent addition to the Today sidebar submenu?

Don’t know about the sidebar, but the feature itself has been around for around 3 years I believe.

@snowtimeglass

In AnkiMobile 23.10.1 build 4, setting 0 second (the default value) to the Auto Advance options doesn’t seem to disable the auto advance feature

Thanks for catching this!

[Expertium]

@dae

It feels a bit complicated. What if log loss increases and RMSE decreases, or vice versa? And is it safe to always assume that the generated parameters are worse when those metrics increase? Would it be possible that certain changes to the parameters could lead to a situation where the average card was scheduled better, but RMSE increased?

@L.M.Sherlock what’s your opinion?

[L.M.Sherlock]

If the Log Loss and RMSE(bins) both increase in the same dataset, the new parameters are worse than the old one.

[dae]

That didn’t really answer my questions I fear other intended side effects of this too - for example, we assume that users should recalculate their params when updating from FSRS 4 to 4.5, but what if when they tried to do so, Anki said “your old params were better”? That feels like it’s going to create confusion and a support burden.

[L.M.Sherlock]

I think the best method is to hide everything technical.

[Freezable0596]

I tried using the both the qt6 and qt5 intel versions as there is no qt5 version for arm. In the qt5 version the after clicking around a bit, the buttons in the top bar work and don’t cause the application to crash. In the intel qt6 version after pressing sync, none of the buttons in the bar seem to work. They don’t cause the application to crash, but nothing happens. The keyboard shortcuts aka b or cmd + , don’t work either.

[Expertium]

I agree, it shouldn’t notify the users.
But the question is: what if RMSE is better and logloss is worse? Or vice versa? Which parameters should be used?

[vaibhav]

What if the user still wants to know the new optimized parameters (for e.g. to see how much worse they are)?
Also, the aim should be to figure out why the new parameters are worse than the older ones and to correct the issue. The method proposed by @Expertium is just a stopgap measure.

[Aleksej]

@dae

@Aleksej

In a filtered back, pressing Enter in the Search field does not press the Rebuild button.

I can’t reproduce this. Which platform/Qt version?

Version ⁨23.12 (fc492a41)⁩
Python 3.9.15 Qt 6.6.1 PyQt 6.6.1
Debian, Xfce

[dae]

23.12 was released as stable today. Season’s Greetings everyone!

Releases · ankitects/anki

Changes since rc1:

• Exclude suspended cards from FSRS optimizing
• Select cancel button by default when exiting card layout screen
• Fix new cards from parent decks being gathered (#2907) (thanks to Rumo)
• Improve confirmation message in Add screen (#2903) (thanks to Abdo)
• Update to FSRS-4.5 (#2901) (thanks to Jarrett)

[sprvlcn]

Is it just me, or does https://apps.ankiweb.net/ still have 23.10.1? Ctrl+F5 doesn’t update the webpage.