Python library for defining and evaluating multi-dimensional array expressions with GPU acceleration, historically used as the backend for early deep learning frameworks. Built and maintained by WIEWAVE for Azure Marketplace, on Ubuntu and Debian.
| Built & maintained by | WIEWAVE |
|---|---|
| Category | AI & Machine Learning |
| Operating systems | Ubuntu, Debian |
| Marketplaces | Azure Marketplace |
| Offer types | Public listing, with private offers on request |
| Marketplace | Status |
|---|---|
| Azure Marketplace | Published |
| AWS Marketplace | Available on request |
| Google Cloud Marketplace | Available on request |
Need it on another marketplace, or as a private offer for your organisation? cloud@wiewave.com
Each image follows its distribution's own provisioning model, package manager and security tooling — not one build relabelled several times.
LTS and interim releases, Minimal and Pro variants, built to Canonical's cloud-image conventions.
Stable and oldstable, with backports where a workload needs a newer runtime than the release ships.
The same four steps behind every offer we've published, including the hardening and CIS Benchmark checks every build goes through.
The distribution, licensing model and target marketplaces are agreed before anything is built.
Packer templates, Ansible provisioning and a pinned package set — then hardened, scanned and checked against the CIS Benchmark for its distribution.
Taken through each cloud's own certification pipeline before it goes live on the marketplace.
Rebuilt on the upstream security cadence and re-published, with old versions retired without breaking deployments.
If yours isn't here, ask our marketplace team directly.
GPU-ready training and inference images with drivers, CUDA and frameworks already matched to each other.
Compact deep learning framework with autograd and several accelerator backends, aimed at readable model code and experimentation, installed in a Python environment.
Lua-based scientific computing framework with tensor math and neural network packages, the predecessor of PyTorch, built with LuaJIT and CUDA bindings.
Model serving framework for PyTorch exposing inference and management REST endpoints, with batching, versioning and metrics for deployed model archives.
NVIDIA Triton serving TensorRT, ONNX and PyTorch models side by side, with dynamic batching and the metrics endpoint enabled.
Toolkit for training reinforcement-learning agents inside Unity environments, bundling the Python trainer package, PyTorch and the ml-agents communication library.
Vectice provides model documentation and governance for data science teams, capturing experiment lineage, dataset versions and validation evidence for review.
Tell us the distribution, the marketplace and the commercial model — we'll build, certify and publish it as a public listing or a private offer.
