Advanced9h 16m4 modules · 21 leçons0 inscrits

RAPIDS in Production — cuDF, cuML and the Full GPU Pipeline

The RAPIDS course for practitioners: you already know what a GPU is, and now you want to run real pipelines on one. cuDF and cudf.pandas, the Polars GPU engine, Numba-compiled UDFs, spilling and unified memory when VRAM saturates, Dask-cuDF across multiple GPUs, cuML with its scikit-learn API parity, GPU XGBoost and accelerated SHAP explainability, cuxfilter for exploring tens of millions of points, and finally production deployment: Amdahl's law, real cost, and GDPR. Every lesson measures the gain honestly, on concrete French use cases. 2026 edition of the RAPIDS stack.

Par Équipe Curriculum Nexus

69€

Accès complet · Certificat inclus

S'inscrire maintenant

Programme du cours

4 modules

01
Module
RAPIDS & cuDF Foundations: the DataFrame on GPU
Understand why a GPU accelerates data science, how RAPIDS and cuDF replicate the pandas API, how to load and type your data efficiently, and how to measure the real gains from accelerated operations.
5 leçons·2h 10

Module

Data Manipulation at Scale

Move from basic operations to the transformations that structure a real pipeline: complex aggregations, joins, rolling windows, vectorized string processing, controlled memory overflow, and multi-GPU distribution with Dask-cuDF.

6 leçons·2h 37

RAPIDS in Production — cuDF, cuML and the Full GPU Pipeline

Programme du cours

RAPIDS & cuDF Foundations: the DataFrame on GPU

Data Manipulation at Scale

Machine Learning with cuML

From GPU Data Science to Production