In the HPO application I see the following explanation:

'Maximum iterations per experiment after which it will be stopped. Iterations are based on the experiments' own reporting (for example, if experiments report every epoch, then iterations=epochs)'