NVIDIA-Merlin/NVTabular

Health v2 · maintenance only

Adoption not part of the score

1150 stars · 149 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

NVTabular is a GPU-accelerated feature engineering and preprocessing library for tabular data, built to handle terabyte-scale datasets for training deep learning recommender systems. It is part of NVIDIA Merlin and uses RAPIDS Dask-cuDF to process datasets that exceed GPU and CPU memory.

Use cases

• preprocess terabyte-scale tabular datasets for deep learning
• feature engineering for recommender systems on GPU
• accelerate data loading for GPU model training
• build end-to-end recommender system pipelines with NVIDIA Merlin
• apply training-time preprocessing to inference with Triton

When to choose

• you train deep learning recommenders on very large tabular datasets
• you need GPU-accelerated ETL that out-of-memory scales via Dask-cuDF
• you want preprocessing steps reusable at inference time within the Merlin ecosystem

When to avoid

• you work with small datasets where pandas or scikit-learn suffice
• you have no NVIDIA GPU available
• you need general-purpose CPU-only feature engineering

Facets

library · maturity maintenance

etl machine-learning data-science machine-learning deep-learning gpu-computing python recommender-systems feature-engineering tabular-data rapids nvidia-merlin cudf preprocessing data-engineering gpu linux docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NVIDIA-Merlin/NVTabular")