A machine assessment stage is a sequence of ML capabilities that processes your data. To create a new machine assessment stage for your process, first either create a new process as per the Creating Processes page, or navigate to an existing process by clicking on Assessments -> Processes and selecting your process from the list.
Once you are on your process page:
- Click on the Stage Definition tab
- Click the 'New Machine Assessment Stage' button
- Give your machine assessment stage a name
- From the Previous Stage dropdown select a previous stage
- Click the 'Save' button
- Now contact Silverpond to continue setting up the ML capabilities of your machine assessment stage
Choosing the best training run for each capability:
- A 'capability' in highlighter specifies inputs and outputs. A 'training run' is a set of trained weights that can be used to process the inputs and generate the outputs.
- There are often multiple training runs per capability, because each time we try different training configurations or add new data, we will generate new trained weights.
- Depending on your use case, you may want to choose the one with the highest performance, or you may want to choose one based on a specific dataset, or the one that was most recently trained.
- The progress of research on each capability is tracked under Agents -> Evaluations. Go there, find your capability's evaluation, and find the best training run in that evaluation. Take note of the capability id and the training run id.
- You can use this information when contacting Silverpond to help with setting up the ML capabilities of your machine assessment stage.