Runners

Module: polygraphy.backend.trt

class TrtRunner(engine, name=None)

Bases: polygraphy.backend.base.runner.BaseRunner

Runs inference using TensorRT.

Note that runners are not designed for production deployment and should generally be used only for prototyping, testing, and debugging.

Parameters

• engine (Callable() -> Union[trt.ICudaEngine, trt.IExecutionContext]) –

  A callable that can supply either a TensorRT engine or execution context. If an engine is provided, the runner will create a context automatically. This callable is invoked whenever the runner is activated.

  Alternatively, the engine or context may be supplied directly instead of through a callable, in which case the runner will not take ownership of it, and therefore will not destroy it.

• name (str) – The human-readable name prefix to use for this runner. A runner count and timestamp will be appended to this prefix.

set_profile(index)

Sets the active optimization profile for this runner. The runner must already be active (see __enter__() or activate()).

This only applies if your engine was built with multiple optimization profiles.

In TensorRT 8.0 and newer, the profile will be set asynchronously using this runner’s CUDA stream (runner.stream).

By default, the runner uses the first profile (profile 0).

Parameters

index (int) – The index of the optimization profile to use.

infer(feed_dict, check_inputs=None)

Runs inference using the provided feed_dict.

Parameters

• feed_dict (OrderedDict[str, numpy.ndarray]) – A mapping of input tensor names to corresponding input NumPy arrays.

• check_inputs (bool) – Whether to check that the provided feed_dict includes the expected inputs with the expected data types and shapes.

Returns

A mapping of output tensor names to their corresponding NumPy arrays.

IMPORTANT: Runners may reuse these output buffers. Thus, if you need to save outputs from multiple inferences, you should make a copy with copy.deepcopy(outputs).

Return type

OrderedDict[str, numpy.ndarray]

__enter__()

Activate the runner for inference. This may involve allocating GPU buffers, for example.

__exit__(exc_type, exc_value, traceback)

Deactivate the runner.

If the POLYGRAPHY_INTERNAL_CORRECTNESS_CHECKS environment variable is set to 1, this will also check that the runner was reset to its state prior to activation.

activate()

Activate the runner for inference. This may involve allocating GPU buffers, for example.

Generally, you should use a context manager instead of manually activating and deactivating. For example:

with RunnerType(...) as runner:
    runner.infer(...)

deactivate()

Deactivate the runner.

If the POLYGRAPHY_INTERNAL_CORRECTNESS_CHECKS environment variable is set to 1, this will also check that the runner was reset to its state prior to activation.

Generally, you should use a context manager instead of manually activating and deactivating. For example:

with RunnerType(...) as runner:
    runner.infer(...)

get_input_metadata()

Returns information about the inputs of the model. Shapes here may include dynamic dimensions, represented by None. Must be called only after activate() and before deactivate().

Returns

Input names, shapes, and data types.

Return type

TensorMetadata

last_inference_time()

Returns the total inference time required during the last call to infer().

Returns

The time in seconds, or None if runtime was not measured by the runner.

Return type

float