# NVIDIA/DALI

A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.

Repository: https://github.com/NVIDIA/DALI
Canonical: https://ross.abutalabs.com/products/dali
Homepage: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/index.html
Language: C++
License: Apache-2.0
License Family: permissive
Topics: fast-data-pipeline, image-augmentation, data-augmentation, image-processing, data-processing, deep-learning, machine-learning, python, neural-network, gpu, gpu-tensorflow, audio-processing, pytorch, mxnet, paddle
Last push: 2026-08-26T17:33:31+00:00

## Health v2 (maintenance only)
Score: 92/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 78, longevity 100
- inputs: {"age_days": 3015, "days_push": 7, "days_rel": 65, "gap_med": 34, "n_releases_24m": 16}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5734, forks 674 (observed 2026-08-28T04:09:28.354408+00:00)

## What it is
NVIDIA DALI is a GPU-accelerated data loading and preprocessing library with optimized building blocks and an execution engine for deep learning training and inference. It provides portable pipelines for image, video, and audio processing that integrate with PyTorch, TensorFlow, JAX, and PaddlePaddle.

## Use cases
- accelerate data loading for deep learning training
- offload image preprocessing from CPU to GPU
- replace torchvision transforms with GPU-accelerated versions
- build fast image augmentation pipelines
- speed up audio and video preprocessing for ML
- fix CPU bottleneck in multi-GPU training
- portable preprocessing across PyTorch and TensorFlow

## When to choose
- training is bottlenecked on CPU data preprocessing
- you have NVIDIA GPUs and want to maximize GPU utilization
- you need consistent preprocessing across multiple DL frameworks
- you process large volumes of images, video, or audio

## When to avoid
- you have no NVIDIA GPU or run on Windows/macOS
- your data pipeline is small and not a bottleneck
- you need CPU-only or cloud-agnostic portability
- you rely on custom operators DALI does not provide

## Facets
- artifact type: library
- maturity: active
- function: image-processing, audio-processing, video-processing, machine-learning, gpu-computing, data-science, etl
- domain: deep-learning, machine-learning, computer-vision, image-processing, gpu-computing
- platform: python, cpp
- tags: data-loading, data-augmentation, gpu-acceleration, cuda, data-pipeline, pytorch, tensorflow, preprocessing, audio, data-engineering, linux, gpu

## Member repositories
- NVIDIA/DALI (main) score 92

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:28.354408+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:53:22.720619+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA/DALI (fetched 2026-08-28T04:09:28.354408+00:00, sha 23c8884d1e23)
  - homepage: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/index.html (fetched 2026-08-29T08:48:55.216718+00:00, sha 2f4a65dd72e9)
  - site_page: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/examples/getting_started/dynamic_mode.html (fetched 2026-08-29T08:48:55.234682+00:00, sha 93b1fd759543)
  - site_page: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/examples/getting_started/torchvision_api.html (fetched 2026-08-29T08:48:55.237531+00:00, sha 4a6d06af7669)
  - site_page: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/security.html (fetched 2026-08-29T08:48:55.239795+00:00, sha 9b62a133418f)
  - site_page: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/installation.html (fetched 2026-08-29T08:48:55.225879+00:00, sha c89ee9d493ee)
  - site_page: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/support_matrix.html (fetched 2026-08-29T08:48:55.227673+00:00, sha abc8ed5c3d8b)
  - site_page: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/examples/getting_started/index.html (fetched 2026-08-29T08:48:55.229568+00:00, sha 666847d5cb95)
  - site_page: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/examples/getting_started/pipeline_mode.html (fetched 2026-08-29T08:48:55.231722+00:00, sha f03c6b472027)
  - site_page: https://docs.nvidia.com/deeplearning/dali/user-guide/docs/pipeline_api_overview.html (fetched 2026-08-29T08:48:55.241363+00:00, sha b1261fe1365f)
- Data as of 2026-08-30T08:39:29.467469+00:00.