The whole afternoon disappeared into tracking down a bug Ken sent me…
I can still write stuff today. But I am not sure who is writing what right now, and I don’t want to waste time on things others are already working on.
I’m still working on the compilation section
More specifically: @ccshan are you working on the Histogram section?
I did fix that bug, but now I don’t know if the result is ‘good’, so that we can use it as an example! I need someone who understands what counts as good output to check.
I’m not actively writing anything, just preparing to write the histogram section. So please feel free to write a histogram section, or at least import the existing text.
Thanks for tracking down the bug, but I believe without trying that your fix is a non-starter because it breaks all our Gibbs sampling benchmarks. I just came back from teaching and will write more soon.
I’ve just pushed about a page on compilation
We’re at about 9.5 pages, and plenty more remains to be written
@rjnw I think we need to focus on getting the benchmarks running, regardless of the optimization state at the moment
and then returning to optimization after we have plottable data
Also I created a submission on the PLDI site: https://pldi18.hotcrp.com/paper/54/edit
I am going to do that for the first three before going to sleep
along with the most of the optimizations running. Then it is only naive bayes and lda
Also I need help with the input data for that
If someone can pull out the script for getNews and store the input data for lda and naivebayes. I will be very greatfull
I can look into that