Oh 2D didn’t work, augur assumes rectangular 2D array somehow or numpy didn’t allow irregular 2D array. I was getting address boundry error in the C code due to this and didn’t want to debug a SIGSEGV.

But didn’t their own LDA code, which you showed me, use an irregular array of arrays?

no the way they were making the array seemed like it could be used with irregular arrays but when I tried using it doesn’t work.

Also it was faster for me to just switch to 1D rather than spending time to figure out the seg fault, so I did.

Ok

So the way I’d put it, they picked their data to avoid a bug about irregular arrays

Even though their example model code says until N[d] and all

So are you timing augurv2 NB accuracy soon?

(I’m on my way to the office, in case you want to discuss in person)

btw, I’m back from PLDI so I’m around to talk at Luddy today

I’ll need to leave around 1pm

okay I will be there soon