-
Notifications
You must be signed in to change notification settings - Fork 175
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
SortingAnalyzer
issues
#3031
Comments
Just a comment, there is documentation for the extension dependency graph but it might not be easily discoverable: |
@h-mayorquin I see, thanks, should have read that |
I wonder about this. I too like this idea to some extent especially since a chunk of users really only want to use default for everything and don't care about the fine-tuning... But I think the idea for the analyzer was that it would force the user to make these choices for two reasons: 1) tuneability so that it is easy to optimize and reproduce and 2) education so the user can see how their choices affect quality. Could I ask what your specific use-case is? You just want to use defaults?
This will be solved with #2785. :) Just not merged yet.
this is interesting. What is the actual example you're thinking about?
this shouldn't happen. Could you double-check that this is happening within the context of the dependency graph. |
Hi Kyu,
Normally compute, launch the computation only on the targetted list or dict. This is strange.
Having several instances of an extension with differents parameters would be extremely complicated and also error prone - I think -
I think I prefer no.
I am agree with you but Alessio convinced me that most of end users would prefer to load everything at once. |
Thanks @zm711 and @samuelgarcia
Feel free to close this issue once #2785 is merged. |
Assumption here is that if you run it again it is because you have changed something and want to re-run it. But I agree I have made this mistake many times accidentally and it is frustrating. Maybe we could add a check such that if the params are the exact same then we don't re-run?
I think that is just for default... I you look at the code if you force it by saying metrics_names = [nn...] then even though the default PC will be skipped the fact it is listed as a desired metric should include this. Based on my reading.But I also agree this is a little annoying that we remove something at default. Not sure the thought process, ie, was it for testing? |
We removed |
@alejoe91 @samuelgarcia
I just started trying the new sorting analyzer and wanted to share some issues that I have encountered:
sorting_analyzer.compute
, you recompute all the extensions. I think that it would be better if it first checks if the extension exists with the same parameters and not re-do it if possible. Maybe there could be arecompute=True
parameter that you can pass.principal_components
extension could be computed withn_components=5
, but a PC metric that depends on it may ask forn_components=10
. In that case it seems like the user should get at least a warning, or a error message saying that the parameters are not compatible.si.load_sorting_analyzer
should haveload_extensions=False
by default (rather than load all the extensions by default, which could take a long time)load_sorting_analyzer
and tried to compute the extensions, I got an error saying that it requires a recording and the sorting analyzer doesn't have one. I can load the recording independently, but I don't know how it can be associated with the loaded sorting analyzer again. Would I 'recreate' the sorting analyzer again withcreate_sorting_analyzer
again in that case?The text was updated successfully, but these errors were encountered: