Working With the Sandbox
The sandbox is a custom space on Neu.ro that was set up by our team to provide an out-of the box experience to test the platform's capabilities.
For example, our OSS dogs sandbox was created to showcase a model that's able to classify two specific dog breeds from images.
If you have followed the Getting Started guide and already have a deployed Pachyderm cluster that's ready to create and run pipelines, you can receive access to the sandbox and start working with it according to its
readme
file without the need to set up integrations with various third-party tools.The Dogs sandbox comes integrated with a few required tools right away:
- Pachyderm - A data layer that provides functionality for facilitating a full machine learning lifecycle. We use it for creating a pipeline that triggers model re-training on dataset updates that affect image labels.
- object detection, and semantic segmentation functionality that we need in this sandbox.
- MLFlow - A ML lifecycle management platform with main focus on experimentation, reproducibility, deployment, and a central model registry.
- SHAP - A tool that employs a game theory approach to the task of explaining outputs of machine learning models.
The general setup of the sandbox is performed on a per-client basis and includes several Kubernetes-native tools. This implies initial setup of the cluster from our side, so if you're interested in getting your own sandbox, please contact us at
[email protected]
.