Our processors benefit from “SIMD” instructions. These instructions can operate on several values at once, thus greatly accelerating some algorithms. Earlier, I reported that you can multiply the speed of common (fast) random number generators such as PCG and xorshift128+ by a factor of three or four by vectorizing... Read more

As you encounter various data elements you should come across categorical data. Some individuals simply discard this data in their analysis or do not bring it into their models. That is certainly an option, however many times the categorical data represents information that we would typically want to bring in to... Read more

PyTorch is a flexible open source framework for Deep Learning experimentation. In this post, you will learn how to train PyTorch jobs on Amazon SageMaker. I’ll show you how to: build a custom Docker container for CPU and GPU training, pass parameters to a PyTorch script, save the trained model. As usual, you’ll find my code... Read more

Roaring bitmaps are a popular data structure to represents sets of integers. Given such sets, you can quickly compute unions, intersections, and so forth. It is a convenient tool when doing data processing. I used to joke that Roaring bitmaps had been implemented in every language (Java, C, Rust, Go,... Read more

This work is supported by Anaconda Inc To increase transparency I’m trying to blog more often about the current work going on around Dask and related projects. Nothing here is ready for production. This blogpost is written in haste, so refined polish should not be expected. Current efforts for June... Read more

I learned a new thing this weekend: we apparently can apply a patch onto a branch/fork using git apply . There’s a few things to unpack here. First off, what’s a patchfile? The long story cut short is that a patchfile is nothing more than a plain text file that contains all information... Read more

This work is supported by Anaconda Inc. History For the first year of Dask’s life it focused exclusively on single node parallelism. We felt then that efficiently supporting 100+GB datasets on personal laptops or 1TB datasets on large workstations was a sweet spot for productivity, especially when avoiding the pain... Read more