NNPACK

Visit Tool

NNPACK is an acceleration package for neural network computations, providing high-performance implementations of convnet layers for multi-core CPUs. It offers low-level performance primitives for deep learning frameworks.

Claim this tool

No Views Yet

At a glance

Pricing

Open Source

Free tier

Yes

API

Skill level

Technical

About

What is NNPACK?

NNPACK is an acceleration package specifically designed to optimize neural network computations on multi-core CPUs. It focuses on delivering high-performance implementations of convolutional neural network (convnet) layers. The tool is not intended for direct use by machine learning researchers but rather provides low-level performance primitives that are leveraged by leading deep learning frameworks such as PyTorch, Caffe2, MXNet, and Darknet. It supports various platforms including Linux, macOS, Android, and iOS, and offers multiple algorithms for convolutional layers, including Fourier transform, Winograd transform, and implicit matrix-matrix multiplication. Implemented in C99 and Python, NNPACK features multi-threaded SIMD-aware implementations and extensive unit test coverage.

Best used for

Ideal for developers who need to integrate high-performance convolutional neural network layers into deep learning frameworks, optimize neural network computations on multi-core CPUs, and accelerate both inference and training processes. Especially valuable for enhancing the performance of AI applications on various platforms including mobile devices.

Common actions

accelerate neural networks

optimize convnet layers

integrate deep learning primitives

enhance CPU performance

collaborationworkflowsdeepfakelow-code/no-codeautomated workflowopen-source"AI Agents"github copilotface swapping

Capabilities

Key features

High-performance convnet layers
Multi-threaded SIMD implementations
Multiple convolution algorithms
Inference/training optimized layers
C99 and Python implementation

Target Audience

developer

Integrations

pytorchcaffe2mxnettiny-dnncaffetorchdarknettvm

Pricing & Plans

Open Source

Free

FAQs

What kind of neural network layers does NNPACK accelerate?

NNPACK primarily accelerates convolutional layers, including inference-optimized forward propagation, training-optimized forward propagation, and backward gradient updates. It also supports fully-connected, max pooling, ReLU, and Softmax layers for both inference and training.

Which deep learning frameworks can leverage NNPACK?

NNPACK provides low-level primitives for leading deep learning frameworks. It is supported by PyTorch, Caffe2, MXNet, tiny-dnn, Caffe, Torch, and Darknet, among others. TVM also supports NNPACK for inference in convolutional layers.

What are the CPU requirements for NNPACK?

For x86-64 systems on Linux and macOS, NNPACK requires AVX2 and a 3-level cache hierarchy. For ARM-based systems on Linux, Android, and iOS, it requires NEON support. Specific Android ABIs have additional toolchain requirements.

Also listed in

This tool also appears in

AI Agents & Automation › AI Frameworks & Infra Research & Education › Scientific Computing

Explore

NNPACK

At a glance

What is NNPACK?

Key features

Target Audience

Integrations

This tool also appears in

Browse AI tools by category

What is NNPACK?

Key features

Target Audience

Integrations