Bruno Goutorbe, Chief Data Scientist at Cdiscount
Julien Jouganous, Data Scientist at Cdiscount

Sometimes, computational performances become a crucial point in production algorithms.
That’s especially true in data science where proof of concept pieces of code can have a poor scalability in terms of computation time as well as resource (CPU, memory) use.

There are multiple ways to overcome this type of issues, but before pulling out the heavy artillery (parallel computing, cython…), one may want to make sure his code has been written efficiently. Optimizing code in python is not always trivial and can be a bit counterintuitive. Hopefully, several profiling tools are here to make python devs and data scientists lives easier!

This presentation (in french) gives a quick overview of the profiling tools we use in the data science team.

link to the slides

The script used to illustrate the slides is provided here.

See you soon!

Write A Comment