#6767th of 126 in ML frameworks and libraries
dali
GPU-accelerated data loading and preprocessing library for deep learning pipelines
- Stars
- 5.8k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Oct 2026
- Language
- C++
5,634 on 10 Apr 20265,790 on 30 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
DALI moves image, video and audio decoding and augmentation (crop, resize, normalize) from the CPU to the GPU, using its own execution engine with prefetching and parallel execution. Pipelines are written in a Python functional API, in pipeline mode or the newer dynamic mode, and plug into PyTorch, TensorFlow, PaddlePaddle and JAX. It also has a Triton Inference Server backend.
Who it is for: ML engineers on NVIDIA GPUs whose training is bound by CPU data loading
Strengths
- Decodes JPEG, JPEG 2000, H.264, VP9, HEVC and audio formats on the GPU
- One pipeline definition retargets to PyTorch, TensorFlow, PaddlePaddle or JAX
- Supports GPUDirect Storage and multi-GPU scaling
- Custom operators and a Triton Inference Server backend
Weaknesses
- Requires an NVIDIA driver and CUDA Toolkit; README documents no AMD or Apple GPU path
- Pipeline mode and dynamic mode are two separate APIs to learn
- Installation is tied to a CUDA major version via separate pip packages
- Roadmap link in the README refers to a 2024 plan
What it needs
- GPU required
- Docker
- Needs NVIDIA driver, CUDA Toolkit
- Models: PyTorch, TensorFlow, PaddlePaddle, JAX
Quick answers
Is dali open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Can you run dali with Docker?
Yes. Its repository has a Docker setup: a Dockerfile, a Compose file or a published image.
Does dali need a GPU?
Yes, it needs a GPU.
Also in ML frameworks and libraries
See all 126| Rank | Project | Score |
|---|---|---|
| 1 | pytorchPython tensor library with GPU support and tape-based autograd for deep learning | 95 out of 100 |
| 2 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 93 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts | 93 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 91 out of 100 |