GPU & Compute

cunn

Torch CUDA Neural Network Implementation for GPU-accelerated deep learning.

Status
UnmaintainedNo commits in 3275 days.
Last checked
Checked · Oct 7, 2026
License
Open source
Pricing
Free tier · from $0
Visit cunn Compare with alternatives

What it is

cunn is a library that provides GPU acceleration for neural networks using CUDA, enabling faster training and inference in Torch. It's essential for developers working with large datasets or complex models who require high performance.

cunn stands out as a specialized library within the Torch ecosystem, offering direct CUDA integration to accelerate neural network operations on GPUs, making it ideal for performance-critical applications.

When it fits — and when it doesn’t

A good fit for

  • Teams working with large datasets and complex neural network architectures who need GPU acceleration for faster training.

  • Projects requiring real-time inference capabilities on GPUs.

  • Developers already using Torch who want to leverage CUDA for performance improvements.

Look elsewhere if

  • Projects that do not require or cannot utilize GPU resources.

  • Teams looking for cloud-based solutions without the need for local setup and maintenance.

Known weaknesses

Each one links to the public issue, changelog or docs page it came from, where we have one.

  1. 01

    Limited language support due to primary reliance on Lua

    Lua is less popular compared to Python or JavaScript, limiting the developer community and available resources.

  2. 02

    Complex setup for CUDA environment

    Setting up CUDA can be challenging due to hardware and driver compatibility issues which may require significant effort to resolve.

  3. 03

    Small and potentially inactive community

    As an open-source project primarily in Lua, cunn might not have a large or active developer community for support and contributions.

  4. 04

    Performance limitations on non-NVIDIA hardware

    cunn relies heavily on CUDA which is optimized for NVIDIA GPUs. Performance on other GPU brands may be suboptimal or unsupported.

What it does well

  • GPU acceleration using CUDA for faster neural network training and inference.

  • Integration with Torch, a scientific computing framework.

  • Supports various deep learning models and architectures.

Pricing

Free tier
YesOpen source — free to use
Paid from
$0
Model
Flat rate
Enterprise plan
No

Pricing not independently confirmed — check the vendor’s page.

Performance benchmarks

How Fast Is It?

Set up cunn

Setup steps, common gotchas and what to check after 30 days.

Open the setup guide →