Valohai ML platform logo

Machine learning infrastructure without vendor lock-in

Building on open standards and open source

Open API

The Valohai platform is developed 100% API first, meaning everything in the platform can be accessed through an open API. You’re always in control. Integrate, orchestrate and make offline backups of your data and metadata through the API.

See the Valohai API

Open standards

Valohai is built on top of open standards and technologies, from Docker and Kubernetes to REST APIs standard frameworks that make it easy and safe to extend for anyone. Let your data science team choose the software environment they want to run their experiments on, select their runtime hardware and hit run.

Open source

Many parts of Valohai are open source so that you can integrate and in the best case add features to the tools. Regardless if you run against the Valohai cloud, your own cloud or on your own premises you’ll use the same tools.

Open for integrations

We believe in giving data scientists the tools to be productive but letting them use the frameworks and languages they are already comfortable with to make the learning curve as gentle as possible. Use any framework in any programming language of your choice.

Core components

valohai-cli

The most used way to execute runs on the Valohai platform. Fork, submit pull requests or report issues. It is also the prime reference on how to use the API Valohai is built on.

View on GitHub.com

valohai-yaml

Valohai YAML parses and validates the Valohai projects setup YAML file. Ever thought about building your own local clone of Valohai for smaller executions? Use the valohai-yaml project to ensure your project setup is interpreted in the same way.

View on GitHub.com

valohai-local-run

Run experiments designed for Valohai on your local Linux hardware. Can be used in air-gapped environments where execution is needed for instance for transfer learning on confidential content.

View on GitHub.com

Integration examples

tensorflow-example

Handwritten digit detection using Tensorflow and the MNIST dataset.

View on GitHub.com

keras-example

Keras examples with Theano or TensorFlow backend for Valohai platform.

View on GitHub.com

darknet-example

Darknet examples for Valohai platform.

View on GitHub.com

r-example

R examples using the official r-base Docker image.

View on GitHub.com

unity-example

Run Unity standalone binaries in combination with machine learning frameworks.

View on GitHub.com

pocketflow-example

Pocketflow example repository for Valohai platform.

View on GitHub.com

cntk-example

Microsoft Cognitive Toolkit example repository.

View on GitHub.com

airflow-valohai-plugin

Airflow plugin to launch executions remotely in Valohai.

View on GitHub.com

fasttext-example

Try fastText on your own dataset with Valohai.

Run on Valohai

huggingface-transformers-valohai

Huggingface Transformers provide state-of-the-art NLP for TensorFlow 2.0 and PyTorch.

Run on Valohai