So here I am again, gearing up for my second GSoC evaluations. I guess this quote from Albert Einstein perfectly relays my experience:

“When you are courting a nice girl an hour seems like a second. When you sit on a red-hot cinder, a second seems like an hour. That’s relativity.”

I can say without any doubt that my experience with ArviZ has been akin to the former, as I still find it hard to believe that it has been almost two months since the Coding Period started. The learning curve for me during this time has been exponential, and I have a much better knowledge of different open source tools and am much more comfortable with python than before.

Progress Update

The time following my previous blog post was spent mainly on improving the remaining plots and learning more about Numba AOT and wheel. I also managed to speed up the _fast_kde_2d by a significant amount, which also resulted in the other plots gaining a substantial speedup. I had to write the benchmark tests for the new methods as well along with the required tests. Here is the result of the benchmark run:

Benchmark results on my machine Image

As you all can see, I was able to gain a speedup of almost 50% and 28% with _fast_kde and _fast_kde_2d respectively. My pull request for the same is ready and up for review.

Next Step

My next plan of action is to work on arviz.data and work on Numba AOT compilations.Plus need to keep fixing the bugs which pop up as a result of my previous pull requests and improve the benchmark tests. I will update you all soon !! Farewell.