modin VS cudf

The interesting thing about Polars is that it does not try to be a drop-in replacement to pandas, like Dask, cuDF, or Modin, and instead has its own expressive API. Despite being a young project, it quickly got popular thanks to its easy installation process and its “lightning fast” performance.

Yeah, tried Polars a couple of times: the API seems worse than Pandas to me too. eg the decision only to support autoincrementing integer indexes seems like it would make debugging "hmmm, that answer is wrong, what exactly did I select?" bugs much more annoying. Polars docs write "blazingly fast" all over them but I doubt that is a compelling point for people using single-node dataframe libraries. It isn't for me.

Modin (https://github.com/modin-project/modin) seems more promising at this point, particularly since a migration path for standing Pandas code is highly desirable.

If anyone wants a faster version of pandas it’s not hard to find, modin for example uses multiple cores to speed it up, so if you have 4 cores it’s about 4 times faster than pandas, and has the same API as pandas.

Modin should fit. It implements Pandas APIs with e.g. Ray as backend. https://github.com/modin-project/modin

If you really want speed you should try modin.pandas which makes pandas multi-threaded.

I am guessing XGB is fairly optimised as it is. If you would want to use the sklearn libraries with pandas, look into Modin

Source

The interesting thing about Polars is that it does not try to be a drop-in replacement to pandas, like Dask, cuDF, or Modin, and instead has its own expressive API. Despite being a young project, it quickly got popular thanks to its easy installation process and its “lightning fast” performance.

Perhaps the largest for package size is the NVIDIA developed rapids toolkit https://rapids.ai/ . Even still adding things like pandas and some geospatial tools, you rapidly end up with an image well over a gigabyte, despite following cutting edge best practice with docker and python.

Yes sure, that is how OpenMP does; but on the other side: you seem to already do some basic type inference, and building an AST, no? Then you know as well the size and type of your vectors, and can execute actions in parallel if there is enough data to be worth parallelizing. Is there anyone who don't want their code to execute faster if it is possible? Those that do work in big data domain do use threads and vectorized instructions without user having to type in any directive; just import different library. Example, numpy or numpy with cuda backend, or similar GPU accelerated libraries like cudf.

Not the answer about Ray: you could use rapids.ai. I'm using it for for dataframe manipulation on GPU

To put Data Analytics on GPU Steroids, Try RAPIDS cudf https://rapids.ai/

You can scope out https://rapids.ai/. Nvidia's AI toolkits. They have some handy notebooks to poke at to get you started.

try https://rapids.ai/

As a companion to the other comments, I'd like to mention that the RAPIDS library cuML provides GPU-accelerated versions of quite a few of the algorithms mentioned in this thread (HDBSCAN, UMAP, SVM, PCA, {Exact, Approximate} Nearest Neighbors, DBSCAN, KMeans, etc.).