@startuml [*] -> VALIDATED VALIDATED : The validated package is sequestered into the catalog. This protects the process from a VALIDATED : non-validated package from being supplied to the cataloging step that is different from the VALIDATED : one provided on the validation step.
VALIDATED -down-> TRAINING_REQUIRED : ML detected VALIDATED -> AVAILABLE : No ML TRAINING_REQUIRED -> AVAILABLE : Training Complete AVAILABLE -> TRAINING_REQUIRED : Retraining Required
TRAINING_REQUIRED : Training Iteration is tracked. It is initialized with a zero at validation. TRAINING_REQUIRED : During training multiple training iterations may be applied. TRAINING_REQUIRED : Therefore the count returned from training may increment by TRAINING_REQUIRED : more than one. How we manage the availablity for different TRAINING_REQUIRED : trained instances is FFS. For now only the latest trained TRAINING_REQUIRED : version is "AVAILABLE".
AVAILABLE : This is an application in which the Run-Time can create a configuration for. AVAILABLE : However, we may need to consider training iteration count as after a AVAILABLE : configuration is created, additional training might become required AVAILABLE : a determination needs to occur as FFS if we invalidate existing configuration AVAILABLE : which raises a question on "RUNNING" instances. Or do we allow the AVAILABLE : deployment of any iteration that reaches the AVAILABLE state.
@enduml |