Does the RMSE/log-loss being exposed in debug console not count? That was the compromise we already made (but the required changes haven’t yet been yet iirc).
1 Like
That would be fine. I was just trying to maybe push for an idea that would accommodate for the interests of both new users and long-time users. But if there’s no plan for that, I guess it would be nice to have some chance of looking behind the scenes.
FYI, the current health check is more complicated than simple “how high RMSE is” like evaluate. It depends also on your average retention and number of reviews, so those changing can change what health check shows.
My conclusion was “can’t fix without a different architecture”, but I’d love to be proved wrong. I’ve logged it in a separate issue: Anki appears as Python in the task manager on Windows · Issue #4144 · ankitects/anki · GitHub
It makes for a nicer out-of-the-box experience when switching between versions, and given how many users depend on add-ons, it’s not so uncommon for them to downgrade when they find an add-on is not working. I can appreciate the argument for making it default to off, and that may happen in the future, but I think defaulting to on might make more sense at the moment, especially given how skeptical people are to change, and how they may play around with this at first.
Generally blockers are regressions, not extra features 
I think this and Jarrett’s last review time work are probably not going to be ready on time, and are best deferred to the next update - I don’t anticipate it following far behind the current one, and this new system has made making releases easier than it was previously.
Logged on Expose old evaluate API to the debug console · Issue #4143 · ankitects/anki · GitHub so it doesn’t get missed.
3 Likes
I understand, but right now CMRR is practically useless, regardless of which option is used. A graph would be way, and I mean WAY more useful. Please consider postponing the release until CMRR is replaced with a graph.
Oh, and please make sure to say “Go ahead” in that thread if you don’t object to this change.
I see ongoing discussions there, and presume this is not going to be ready for testing within a few days. It’s been 4 months since a non-patch update, and we can’t keep deferring it indefinitely - there’ll always be something else that comes up. And as I mentioned earlier, it’ll be a soft release in any case.
I have no objections to a graph instead of a specific value, but given you’re still talking about it, a ‘go ahead’ seems a tad premature.
but given you’re still talking about it
I’m talking about it because the same user that proposed this idea now wants to change it 
EDIT: ok, he agreed with me
Ok, so what’s the plan then? Remove CMRR and add the graph in the next release? That seems like a bad idea IMO. I can guarantee you users will keep asking, “Why not just add the graph now, why do we have to wait for the next release?”
@A_Blokee can you make a DR - workload graph in a few days?
Well currently the answer is “Because it hasn’t been made yet.”. But when that’s no longer the case, it’ll be “It’s just been made. You can give it a try by enabling betas in the new launcher, and grabbing the latest release.” And as I mentioned earlier, this is not a release we’ll be actively advertising to users quickly. Your post-release summaries on Reddit are informative and well-loved, but you might want to think about delaying the next one a little if you feel this is a critical feature.
1 Like
but you might want to think about delaying the next one a little if you feel this is a critical feature.
Nah, I’ll just say “CMRR has been removed and a graph will be added in the next release” in my reddit post
Well currently the answer is “Because it hasn’t been made yet.”
Guess I’ll say that ¯\(ツ)/¯
1 Like
I downloaded the installer just now and I’m running Version 25.06b7 (ad34b76f) on Windows 11, and I do see anki.exe in the Task Manager.
However, when searching for “mpv” in the Task Manager search bar, it shows Python instead, but then you can click-to-expand to show mpv as a subfolder of Python.
1 Like
I’ve been checking the forums everyday now for the past few weeks reading the beta threads. For you dae as an admin to Anki development, it sounds like you have it rough managing these constant requests that sometimes are against what you would prefer. Soft or not, I would very much like a release so I can use FSRS-6. That’s why I’ve been checking the forums everyday now.
I do not care about CMRR in the next release. I am not one of the users that is waiting for a CMRR graph in the next release. So Expertimum’s guarantee holds less weight now. I’m not trying to attack the graph, I’m indifferent about it. That’s all, I thought you might like to hear it from a side that is not as involved with the whole development/community and is just strictly a user of Anki - for what its worth, I agree with your current plan. In case I was in the majority of users, I didn’t want to remain non-vocal about it. Thank you.
7 Likes
Opened a pr:
3 Likes
I don’t know if this is new or a longstanding issue, but in the latest beta (Version 25.06b7 (ad34b76f)) on Windows 11, keyboard shortcuts using the Ctrl key only work with the left Ctrl key, not the right Ctrl key.
Edit: Okay, mystery solved. This only happens when the French (Canada) Canadian Multilingual Standard keyboard layout is in effect. Online search shows that this layout remaps the right Ctrl key to become AltGr, so typing Right-Ctrl + e gives œ, and similarly for various other special characters. Sorry for the false alarm.
To make sure the right Ctrl key isn’t broken on the keyboard, you can try holding it down while turning the mouse scroll wheel, which increases or decreases the font size in most Anki windows.
I downloaded the installer just now and I’m running Version 25.06b7 (ad34b76f) on Windows 11, and I do see anki.exe in the Task Manager.
However, this anki.exe seems to be a dummy placeholder, because it uses only 1MB of memory and 0% of CPU…
yep. I guess so too. It also causes issues with some window trackers like the Logitech G Software Hub. I use it to reprogram some of my mouse keybinds for more convenient reviewing and it usually switches back to the desktop profile for applications other than anki, but it unfortunately no longer recognized anki as a standalone software, likely because pythonw.exe is now doing the heavy lifting(?). This is clearly a niche use case, and you can solve it by manually adding the pythonw.exe to the list of paths it should recognize as “anki”, but good luck explaining that to end-users lol
3 Likes
That’s definitely not been the case through 25.02. I tried several in Browse that worked fine with right-Ctrl (and right-Alt and right-Shiftt) – like suspend, flag, card info, remove tags.
The mystery is solved. I was temporarily using a keyboard layout that I don’t often use, and it takes over the right-Ctrl key and remaps it to become an AltGr key. Not an Anki issue.
3 Likes
Today I noticed that the latest iOS beta (250500.1) and the latest desktop beta (25.06b7) have different parameters after optimizing. What the desktop thinks is better, the iOS app says is worse (the iOS app still shows the RMSE). Is the iOS version missing a fix to FSRS 6 and is it an important one?
Edit: or is the way it’s evaluated different? Does it optimize for something else?
Edit 2: I guess it’s related to this change: Feat/evaluate FSRS with time series split by L-M-Sherlock · Pull Request #3962 · ankitects/anki · GitHub. I didn’t realize that it’s not just presented differently, but also evaluated differently. If my understanding of this change is correct, that is. Now it also makes more sense how it’s presented in the user interface.