Anki 23.10 Beta 1-4

Anki Beta Testing
82

Guys, any news about the difference in weights between the versions? I did some tests here and noticed how it spaced out in my case. In the new version, it was very conservative, throwing more than 300 cards for today (I had already finished my studies and decided to compare). The point is that I am much more comfortable with these values, because I always thought that the original was spacing out too much.

image
image1235×515 92.8 KB

83

Notes are not being merged even when the “Merge notetypes” option was enable. Below are the screenshots. I hope I am using the feature right, please let me know if this is not the intended way to use it.

image
image837×389 19.1 KB

Screenshot 2023-09-28 at 20.55.29
Screenshot 2023-09-28 at 20.55.29834×273 16.2 KB

84

A fatal error occurred, and Anki must close. Please report this message on the forums.
Anki 23.10 (d2f392ff) Python 3.9.15 Qt 6.5.2 PyQt 6.5.2
Platform: Windows-10-10.0.19045
Flags: frz=True ao=True sv=?
Add-ons, last update check: 2023-09-29 03:48:42

Caught exception:
Traceback (most recent call last):
File “aqt.taskman”, line 138, in _on_closures_pending
File “aqt.taskman”, line 82, in
File “aqt.taskman”, line 102, in wrapped_done
File “aqt.operations”, line 122, in wrapped_done
File “concurrent.futures._base”, line 439, in result
File “concurrent.futures._base”, line 391, in __get_result
File “concurrent.futures.thread”, line 58, in run
File “aqt.operations”, line 107, in wrapped_op
File “aqt.operations.note”, line 22, in
File “decorator”, line 232, in fun
File “anki.hooks”, line 89, in decorator_wrapper
File “anki.hooks”, line 84, in repl
File “anki.collection”, line 597, in add_note
File “anki._backend_generated”, line 1201, in add_note
File “anki._backend”, line 145, in _run_command
pyo3_runtime.PanicException: assertion failed: (left == right)
left: 1695952847185,
right: 0

Traceback (most recent call last):
File “aqt.browser.table.model”, line 103, in _fetch_row_from_backend
File “anki.collection”, line 931, in browser_row_for_id
File “anki._backend_generated”, line 1631, in browser_row_for_id
File “anki._backend”, line 145, in _run_command
pyo3_runtime.PanicException: called Result::unwrap() on an Err value: PoisonError { … }

85

There are some differences in revlog filtering and training in the Python and Rust optimizers that make their outputs not completely identical.

Would you be able to create an issue on GitHub - ankitects/anki: Anki's shared backend and web components, and the Qt frontend and attach the .apkg file after compressing it into a .zip, so we can try to reproduce this?

Can you reproduce it without add-ons?

1 Like
86

Hey @dae, thanks for your response.
I have created an issue on Github.

1 Like
87

The Easy button is not being affected by the integrated FSR for new cards.

I did a test with the same weight:

0.3524, 0.3869, 1.9765, 22.3079, 4.7508, 1.0115, 0.8749, 0.0000, 1.5617, 0.1000, 1.0117, 2.1119, 0.1170, 0.2724, 1.2289, 0.0626, 2.6100

In the FSRS (javascript) the Easy button shows 20-24 days
However, in the integrated FSRS, the Easy button shows 6-8 days

88

I encountered the problem only once until now, and I did have many add-ons active when it happened.
I suspect it took place when I attempted to add a new note, and two ‘add new note’ operations were triggered simultaneously or nearly simultaneously, likely due to lag and a double click.
I will keep an eye open for further occurrences.

89

@dae and @L.M.Sherlock

I investigated my problem more deeply and discovered that the weights are being saved by DECK GROUP. Shouldn’t it be universal (similar to “Custom scheduling”)?

Even the deck groups with weights that remain empty exhibit a different behavior;

90

Tried the new beta version on Windows, both Qt5 and Qt6. Qt6 runs, but it gives a strange “mouse hitching” behavior periodically (like my mouse is running at maybe 5fps for a few seconds every now and then) (and mainly on the collection page), and with Qt5 doesn’t even run Anki, no crash message or anything, just mouse cursor loading for a moment and then nothing, and no text in %appdata%\Roaming\Anki2\anki.log. All addons disabled as well. Running as administrator doesn’t change anything either. Restarting my computer also did not help with the hitching.

It seems to only occur when I “proc” some graphical changes, like first clicking on the window after being alt-tabbed, or hovering over some of my decks (and the “drop shadow” highlight appears). Funny enough, when I attempted to record it with OBS, my mouse moved smoothly in the recording, but not on the screen.

Edit: Was fixed when I changed the “Video driver” in settings to be OpenGL rather than Direct3D. Maybe Direct3D shouldn’t be the default.

1 Like
91

@AnKingMed

As a medical student, it was always recommended to follow the rules of the famous anking video. I wanted a recommendation for what a new metric would look like. Any suggestion?

image
image739×393 15.5 KB

1 Like
92

Seconded!

1 Like
93

In fact, I was wondering if this could be an innate error of Anki and FSRS due to presenting any value greater than 1d

94

I have not had time to research or test the FSRS algorithm. From what I have read peripherally, it may be a better algorithm, but not necessarily. I’m currently still using the SM-2 (anki’s default) but plan to test FSRS out on one of my decks in the future and see what I think.

1 Like
95

Actually it’s better in almost all cases. Even it’s comparable to SM17.

https://www.reddit.com/r/Anki/comments/15mab6e/fsrs_explained_part_2_accuracy/

5 Likes
96

I have the exact same issues. Regarding the first issue, changing the driver to OpenGL does help, thank you.

1 Like
97

I’ve seen this. It’s fascinating. I don’t know how to measure it on an actual day to day level though. Yes it’s better at predicting memory likely, but you also have to factor in simplicity and being able to understand the algorithm so that you can manually adjust things if needed, take days off, reschedule, reset, etc. I do anticipate that this is better, I just haven’t had time to test it out thoroughly. Hoping to soon

3 Likes
98

If you mean tweaking the parameters themselves, then you don’t have to tweak them. It is the optimizer’s job. A lot of people seem to be concerned about not understanding the meaning of parameters or how to tweak said parameters, but that’s the thing - you shouldn’t be doing that at all. Leave that to the optimizer. The only really important setting in FSRS that you should understand is desired retention.

The helper add-on of FSRS offers this functionality, but it’s unknown when (or even if) this functionality will be added to the Anki version of FSRS as well.

3 Likes
99

I agree for someone doing SRS in a general sense, not messing with the parameters is good. However what about if you’re a med student and have tests every 2 weeks and want to review a bit more closely initially or schedule additional reviews prior to a test? It complicates things

2 Likes
100

In that case, you have 2 options

  1. Temporarily increase your desired retention and reschedule all cards (currently not supported in the beta)
  2. Use the Advance option to make some cards appear sooner (currently not supported in the beta)
2 Likes
101

Neither of those are great options when you’re doing 600+ reviews a day. However I do think there are solutions to this. I just need to explore it more.

Beyond that, I don’t think FSRS is the end goal. AI analysis for individual patterns, analyzing the content of cards and then adjusting reviews based on similar content you may or may not have seen recently, etc are all additional variables that could be factored into the equation.

I’m just working 70 hour weeks right now so I haven’t had time… gotta prioritize my health and my family

2 Likes