Training: what it means.
Training uses data to develop a model capable of performing a task. Once trained, a model can be deployed for inference. NVIDIA describes this distinction in systems where training resources and deployed edge devices serve different roles. (NVIDIA: edge AI)
An accelerator suitable for one stage is not automatically sufficient for every stage. A deployment plan should separately account for model development, evaluation and execution in the intended application. (NVIDIA: edge AI)
Put the definition to work.
Separate the computer used to train a model from the hardware that runs it in a robot. They need not be the same device or even be in the same location. (NVIDIA: edge AI)
Questions to ask next.
Use these questions as a reading checklist, not as a device-selection recommendation. Exact manufacturer documentation takes priority over a general definition.
- Task and representative data
- Training and evaluation workflow
- Target inference hardware and software
For the broader explanation, compare ai processing roles. This page defines the term; the companion guide connects it to the rest of the system.
Common Training questions.
Must training occur on the robot that will use the model?
No. Training and inference can take place on different systems. Identify how the trained model reaches the target and how its behavior will be validated there. (NVIDIA: edge AI)