Set up local inference
Install the required support and record execution consent.
Saving a Model and installing an inference engine are separate steps. Knotest checks execution readiness when a model or test needs local inference.
Follow the setup action
Use the readiness action shown in the Model or test flow. Setup downloads the required support to the service host. Open Settings → Model inference to inspect installed versions and available updates.
Available engine profiles include ONNX and Torch/YOLO CPU support. NVIDIA profiles require a supported Windows GPU. Use the packages offered for the host and task rather than assuming that a package for another platform applies.
Record license consent
Installed support still requires recorded license consent before execution. Review and accept the applicable license through the app’s consent step. Installation and consent remain separate readiness checks.
Match the model configuration
An installed engine does not determine how the Model’s outputs should be read. Confirm task, preprocessing, and output interpretation before starting a Run.
For a configured remote prediction service, use a verified API profile. That connection has its own request mapping and data destination.