The optimizer notebook seemed to fail on a SQL query about fsrs4anki

[L-M-Sherlock]

Like this add-on: Anki Simulator.

[L-M-Sherlock]

OK, I've added some documentation to the notebook just now.

[L-M-Sherlock]

What's your Anki version? I see the collection.anki21b in the files. I never saw it before.

[L-M-Sherlock]

I got it. anki21b is the filename extension of the collection for the latest version of Anki. I have supported it in the newest version now. You can try it at https://github.com/open-spaced-repetition/fsrs4anki/blob/main/fsrs4anki_optimizer.ipynb

[YorkZ]

Hi @L-M-Sherlock Thank you for the help. I'm using the Anki from git commit 5bf031f1e. I built this version a few days ago for the sake of trying out your amazing algorithm. I'll try your new notebook.

[L-M-Sherlock]

@YorkZ The problem still exists. Please export your collection file with Support older Anki versions and try it again.

[YorkZ]

Looks like it succeeded this time, thanks again. But I still have two questions:

1. Does the .colpkg file I exported include all my decks, or just one deck? If it includes all the decks, then how can I make it to include only one deck because I want to optimize each individual deck separately.
2. I still don't know how to use the results of running the notebook. My understanding is that the results would have to be copied back to my local computer because I run Anki on my local computer right? However, I wasn't able to find any document about how to incorporate the optimization results back into the local Anki system.

[L-M-Sherlock]

1. You can use .apkg file to optimize for each deck.
2. Copy the parameters in the notebook and replace the parameters in the custom scheduling of Anki deck options with them. The following output is an exmple.

print(f"const defaultDifficulty = {defaultDifficulty};")
print(f"const defaultStability = {defaultStability};")
print(f"const difficultyDecay = {difficultyDecay};")
print(f"const stabilityDecay = {stabilityDecay};")
print(f"const retrievabilityFactor = {retrievabilityFactor};")
print(f"const increaseFactor = {increaseFactor};")
print(f"const lapsesBase = {lapsesBase};")

[YorkZ]

Thank you so much. I would recommend that you document this step as well though. I'll try it out tomorrow.

[YorkZ]

Sorry I still don't fully understand it. So should I copy the print statements in the Result section from running the notebook, to the Custom scheduling box in Anki deck option? But which line should I insert these print statement to? I don't see any print statement in that javascript file.

[L-M-Sherlock]

The default parameters in the javascript file:

The output of the optimizer:

You can replace the default parameters with the output of the optimizer.

[YorkZ]

I see. I also guessed this, but I did compare the defaults generated by the notebook with the defaults in that javascript file, and found them to be exactly the same. The following was copied from my results of the notebook, and as you can see they have no difference with the ones in that javascript file.

const defaultDifficulty = 4.6179;
const defaultStability = 2.5636;
const difficultyDecay = -0.5913;
const stabilityDecay = -0.1382;
const retrievabilityFactor = 1.1951;
const increaseFactor = 3.201;
const lapsesBase = -0.0562;

[L-M-Sherlock]

Maybe your code hasn't run there. The default output is generated from my revlog.

[YorkZ]

I uploaded only the exported file for the deck I was interested in (the .apkg file), and then ran the notebook again. This time it did produce the different parameters. So everything is ready for me to try it out tomorrow. Thanks again.

[L-M-Sherlock]

More feedback is welcome.

[YorkZ]

Sure. But I suppose it would take some time to know the results right? Or is there a way to know the results quicker?

[L-M-Sherlock]

Yes, it needs time. Maybe one month or more. Thanks for your test. And I will develop a simulator for predicting the review in the future.

[YorkZ]

What did you mean by "predicting the review"?

[YorkZ]

Sounds really cool!