mratsim/Arraymancer

What is it?

Key features

• 01N-dimensional tensors up to 6 dimensions with numpy-style slicing and explicit broadcasting via +.
• 02Reshaping and joining operations such as concat, reshape, split, permute and transpose.
• 03Read and write CSV, NumPy (.npy) and HDF5 files.
• 04Linear algebra routines including least squares, eigendecomposition and covariance matrices.
• 05Autograd-based neural network DSL with Linear, Conv2D, pooling and GRU layers.
• 06CPU, CUDA and OpenCL tensor backends, with CUDA and OpenCL offering fewer operations.

Use cases

• Numerical computing projects in Nim that need fast multidimensional arrays.
• Prototyping small neural networks in Nim with autograd and a concise model definition syntax.
• Deploying models on resource-constrained devices where a large Python stack is impractical.
• Classical machine learning tasks like PCA, k-means and least squares fitting in Nim code.

Quick start

1. 1

   Install Nim with choosenim

2. 2

   Install Arraymancer with nimble

   nimble install arraymancer

Things to know

• !The project is marked experimental, and its deep learning interface is considered unstable and still changing.
• !Arraymancer requires a BLAS and a LAPACK library to be installed on the system.
• !The CUDA and OpenCL backends are less feature-complete than the CPU tensors.

Topics

FAQ

What is Arraymancer?

A fast, ergonomic N-dimensional tensor library in Nim with CPU, CUDA and OpenCL backends, aimed at numerical computing and deep learning. Arraymancer is an ndarray (tensor) library written in Nim, inspired by NumPy and PyTorch. It provides N-dimensional arrays with slicing, broadcasting, reshaping and linear algebra primitives such as matrix multiplication, least squares, eigendecomposition, PCA and k-means.

Is Arraymancer free to use? What license does it have?

Arraymancer is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.

Is Arraymancer still maintained?

Actively maintained. The last commit to Arraymancer was on Oct 6, 2026.

What can I use Arraymancer for?

Numerical computing projects in Nim that need fast multidimensional arrays. Prototyping small neural networks in Nim with autograd and a concise model definition syntax. Deploying models on resource-constrained devices where a large Python stack is impractical. Classical machine learning tasks like PCA, k-means and least squares fitting in Nim code.

How do I install Arraymancer?

Install Nim with choosenim Install Arraymancer with nimble: nimble install arraymancer

Where can I get Arraymancer?

The source code and installation instructions are on GitHub: https://github.com/mratsim/Arraymancer