Enum anira_warmup_mode#

enum anira_warmup_mode#

The warm-up pass of a Hard contract, at anira_handler_prepare: ordinary inferences of every plan, in table order, through the handler’s session on the inference threads, each timed on the thread that runs it (the declared state’s feed and capture, the processor’s before_inference and after_inference, the engine call; never pre_process or post_process, never the wait for a thread or an instance). The warm-up pass also measures the inference time; a measured budget is the maximum of its last four inferences; under an explicit budget the timings are only reported. The session is sized again after the pass, which starts the stream over: nothing of the pass reaches the stream or a declared state. Apart from the pass, every executor of a built-in engine runs one inference over its own scratch when it is made (at load, or at an exclusive handler’s prepare).

Values:

enumerator ANIRA_WARMUP_NONE#

No pass; legal only with an explicit budget.

enumerator ANIRA_WARMUP_FIXED#

n inferences per plan at prepare (the v2 warm_up, which ran per instance at load). With fewer than four the figure is the maximum of all n, the cold start included; UNTIL_STABLE or n >= 5 leaves it out. FIXED 0 is NONE.

enumerator ANIRA_WARMUP_UNTIL_STABLE#

Until the last four inference times all lie within 20 % of their median, or within one step of the clock that took them when that is more, at most 32 inferences (default).

enumerator ANIRA_WARMUP_MODE_FORCE32#