Documentation
Access all resources and documentation or contact us for additional assistance.
Deploy
Initial deployment
Brainome produces code (currently Python3, but other formats will be available in the future) that can be deployed anywhere, in any workflow or pipeline. Brainome predictors do not require any GPU support. However, Python3 predictors do require the installation of Numpy in addition to the standard Python3 libraries. Inserting the output file produced by Brainome into an existing production environment is by far the most common way that our users deploy their predictors. This approach usually fits seamlessly into existing build-QA-deploy-maintain operations cycles without major changes. Indeed, this approach is so simple that, from an operations perspective, the predictor can usually be treated exactly like any other system component.
Ongoing deployment and reactions
If you have a stream of incoming data on which you are running a predictor and want to know if changes in operational conditions are changing your predictor’s performance, you need to annotate the new data and use Brainome to analyze it. Pay special attention to the Memory Equivalent Capacity (MEC) measurement. This should not be drastically different for newly added data compared to the MEC of your training data. If it is, it’s highly recommended to retrain as your data complexity has changed, indicating a change in the underlying behavior.
Keep Going!
After completing these How-to Guides, you are good to go! Like everything else, mastery of measurements and model building can be achieved through practice. So we really suggest you start with an initial project and have fun. We maintain a Brainome FAQ where common questions are answered. You can also reach us via email at support@brainome.ai.
