Hello,
Trying to use fluid.umap~ which causes a crash whenever I call fittransform or fit. I have a build which doesnt use AVX instructions as I have an ARM chip in my machine. The error is linked below along with the patch.
error.txt.zip (35.2 KB)
umapping.zip (1.3 MB)
And for context I am loading only mono files… The data looks sane before it goes in!
Application Specific Information:
terminating with uncaught exception of type
std::runtime_error: TridiagEigen: eigen decomposition failed
abort() called
I’ll bounce this towards @groma who will know better than I what could cause the decomposition to fail. Doesn’t seem crash-worthy though, sorry about that.
its so negative about the whole thing too. you didnt fail sweetie you tried your best!
The important thing is the friends we made along the way
cmake is not my friend.
james, maybe you should install that hacked linux that boots on apple arm on your machine and then build for that. i hear you actually have to build the os with cmake. then maybe you will be friends?
im pretty sure what you just said is hate speech in some circles
Any news on what might be causing this ? 
is it because I’m on the dreaded ARM?
It’s been logged with the relevant authorities.
The root cause is part of the algorithm not converging with the particular data you’re feeding it. The crash comes because the library we’re using for that bit turns out to throw an exception under these circumstances, and this doesn’t get caught and handled anywhere (hence, termination). That will get fixed, but whatever is causing this data to bork things will likely still remain. Does the reference python implementation handle this data set ok?
Ah!! That would make sense. As I just tested some other data and it worked in Max. AFAIK the python code I’ve been running handles the data fine.
I’ll tap you up for the failing data by e-mail if that’s ok: possibly useful to have a reproduceble case where Python succeeds we don’t.