Running from a Flash Drive
Is the instruction the same for the new installer?
1 Like
going by the stability target, for my hardest preset I got 91%, which is harsh, but these cards are incredibly difficult for me, so it makes sense. for some of my easier presets I got 72%, 76%, and 78%, which similarly makes sense to me.
Iāve tried a couple of decks and get 70% on all four alternatives.
-
Anki 25.06 beta 1
-
Anki 25.06 beta 6
In my opinion, the previous design was more elegant and visually appealing. The current implementation also feels a bit janky when hovering the mouse over the available options (see the gifs below).
Gifs showcasing the behavior
-
25.06b1
-
25.06b6
The same issue also affects other screens, such as the Change Note Type menu.
3 Likes
I opened a PR:
5 Likes
The Reset Window Sizes button has a weird appearanceā¦
I also checked the above mentioned PR (FIX dropdown buttons having borders by GithubAnon0000 Ā· Pull Request #4130 Ā· ankitects/anki Ā· GitHub) but the drop list remains with the same configuration, although I notice no difference at hovering.
Can you share the exact steps youāre taking? The PR works flawlessly for me:
Edit: I get it. I only tested with light mode but dark mode still has this issue. Iāll investigate.
1 Like
@chrislg I pushed a new commit which solves the issue you reported. Could you check for regressions in other buttons / if it works for you?
1 Like
Thanks. It works well. I only noticed the previous appearance in the drop menu of New/review order in the inactive selections.
1 Like
Thanks! It should be fixed with the latest commit.
1 Like
Hi all,
A new beta & updated launchers are now available:
Iām thinking it might be time for a soft release soon (published as stable on GitHub, but not updating the website for a week or two, which will give less wide-spread adoption, and distro maintainers some time to catch up). Is there anything people feel should block the release?
Seems reasonable. I think when it was originally implemented, we didnāt even have a view menu.
Works for me - make sure you start with anki-console.exe
Youāll need to copy the AnkiProgramFiles folder.
2 Likes
Thank you for the new beta.
Some feedback
-
Assorted minor issues with new launcher is closed, but Anki still appears as Python in the Task Manager.
-
uv cache is enabled by default, but I canāt see any benefit of doing so, especially if we anyway clear the cache when uninstalling Anki. During updates, the packages arenāt redownloaded if they were unchanged even if the cache was cleared.
Edit: Having packages cached helps when downgrading, but is downgrading so common that the cache should be enabled for all users by default?
2 Likes
TLDR: replace CMRR with a desired retention - workload graph
1 Like
Iām sorry for possibly beating a dead horse, but I think I might speak for some others: is there really no plans to give the health check more information to the user other than that the fit is optimized? Some sense of how good the fit is would be wonderful, even if it doesnāt have to do with exposing RMSE or Log Loss (something like a āhealth barā perhaps).
3 Likes
I would like this too. Apart from the fact that I would just like to see it and mentally track how the values drift over time, I also use it whenever a new version of FSRS is released with extra parameters. Usually the first optimization is worse than what I had and improves rapidly within the first week or so. So far Iāve always taken the hit, because it was possible to see how big the difference was. Now the only information we get is that itās āwell adaptedā without any quantification. Now it could go from < 4% to > 5% and you wouldnāt know.
1 Like
I agree and a problem is that the cool trick for seeing if splitting or merging presets results in better optimization for the cards that make them up can no longer be done as far as I know. I got the presets that I have now because I discovered that they were better than the more intuitive ones that I was using before, by using that evaulate trick, but if I want to try changing it again, I have no way to no if itās better or worse now.
of course, juggling presets for the sake of optimization of FSRS is ideally something you wouldnāt have to do (maybe by having knowledge of decks and presets itself) but currently if you want the best parameters, you do, and now you canāt
What would be the purpose of this? What would you do with this information?
What would I do personally you mean? Iām sure thereās a bunch of use cases for other people that I wouldnāt need, but for me, the whole reason I was able to see that I was misusing Hard was the fact that Hard Misuse drastically brought down my RMSE, and once I started using it in a reasonable way, I was able to see that the fit was better.
The whole point of taking away the Evaluate button was to make sure that FSRS is accessible to newer users. The idea of some āhealth barā or āstatus barā to how well FSRS fits will not take away from that purpose: in fact, it will probably enhance it (all users, as far as I am aware, like visual cues).
If you are not developer then the only thing which you should be concerned about is if you set retention meet the real one. If you set 90% then you should see 90% in stats. This is all.
Your goal is to memorize anything you want - not to understand each bit on the algorithm.
If you set 90& of your goal and in long term you do not see this then:
- correct your behaviour (ex.: stop misuse Hard)
- every week recalculate FSRS parameters
- Reschedule all cards (watch out for first backlog - could be devastating - just manual spread it for next month). Next weeks backlog will be symbolic in count.
- Enjoy your study - spend more time in reading textbooks rather than comparing some parameters
Why this works: when you optimize then you calculate best parameters for new cards. When you change you habits then new/young cards will start more and more influence on FSRS parameters. When you reschedule old cards you say: please adjust your intervals to my new habit and memory. This is of course long time process to repair large review history. Nevertheless it is better than analysing some numbers and spending hours to optimize something which should be done automatically.
I have been doing this for last few month:
This has been said many times in this thread. There is nothing wrong with my retention, and I donāt spend much time looking at optimizations. All you said doesnāt defeat the fact that a lot of users who have used FSRS had found nothing wrong with the Evaluate button before, and that a compromise between showing some sort of visual indicator for your optimizer, could make both sides happy.
1 Like