Dynamic Batcher

model_navigator.api.triton.DynamicBatcher dataclass

Dynamic batching configuration.

Read more in Triton Inference server model configuration

Parameters:

• max_queue_delay_microseconds (int) –

  The maximum time, in microseconds, a request will be delayed in the scheduling queue to wait for additional requests for batching.

• preferred_batch_size (Optional[list]) –

  Preferred batch sizes for dynamic batching.

• preserve_ordering –

  Should the dynamic batcher preserve the ordering of responses to match the order of requests received by the scheduler.

• priority_levels (int) –

  The number of priority levels to be enabled for the model.

• default_priority_level (int) –

  The priority level used for requests that don't specify their priority.

• default_queue_policy (Optional[QueuePolicy]) –

  The default queue policy used for requests.

• priority_queue_policy (Optional[Dict[int, QueuePolicy]]) –

  Specify the queue policy for the priority level.

__post_init__

__post_init__()

Validate the configuration for early error handling.

Source code in model_navigator/triton/specialized_configs/common.py

def __post_init__(self):
    """Validate the configuration for early error handling."""
    if self.default_priority_level > self.priority_levels:
        raise ModelNavigatorWrongParameterError(
            "The `default_priority_level` must be between 1 and " f"{self.priority_levels}."
        )

    if self.priority_queue_policy:
        if not self.priority_levels:
            raise ModelNavigatorWrongParameterError(
                "Provide the `priority_levels` if you want to define `priority_queue_policy` "
                "for Dynamic Batching."
            )

        for priority in self.priority_queue_policy.keys():
            if priority < 0 or priority > self.priority_levels:
                raise ModelNavigatorWrongParameterError(
                    f"Invalid `priority`={priority} provided. The value must be between "
                    f"1 and {self.priority_levels}."
                )

model_navigator.api.triton.QueuePolicy dataclass

Model queue policy configuration.

Used for default_queue_policy and priority_queue_policy fields in DynamicBatcher configuration.

Read more in Triton Inference server model configuration

Parameters:

• timeout_action (TimeoutAction) –

  The action applied to timed-out request.

• default_timeout_microseconds (int) –

  The default timeout for every request, in microseconds.

• allow_timeout_override (bool) –

  Whether individual request can override the default timeout value.

• max_queue_size (int) –

  The maximum queue size for holding requests.

model_navigator.api.triton.TimeoutAction

Bases: enum.Enum

Timeout action definition for timeout_action QueuePolicy field.

Read more in Triton Inference server model configuration

Parameters:

• REJECT –

  "REJECT"

• DELAY –

  "DELAY"