polygraphy.backend.trt.loader.EngineFromNetwork Class Reference

Inheritance diagram for polygraphy.backend.trt.loader.EngineFromNetwork:

Collaboration diagram for polygraphy.backend.trt.loader.EngineFromNetwork:

Public Member Functions
def	__init__ (self, network, config=None)
def	__call__ (self)

Private Attributes
	_network
	_config

Constructor & Destructor Documentation

__init__()

def polygraphy.backend.trt.loader.EngineFromNetwork.__init__	(		self,
			network,
			config = None
	)

Functor that uses a TensorRT ``INetworkDefinition`` to build an engine.

Args:
    network (Callable() -> trt.Builder, trt.INetworkDefinition):
    A callable capable of returning a TensorRT Builder and INetworkDefinition. The returned builder
    and network are owned by EngineFromNetwork and should not be freed manually. The callable may
    have at most 3 return values if another object needs to be kept alive for the duration of the network,
    e.g., in the case of a parser. EngineFromNetwork will take ownership of the third return value, and,
    like the network, it should not be freed by the callable. The first and second return values must
    always be the builder and network respectively.
    If instead of a loader, the network, builder, and optional parser arguments are provided directly,
    then EngineFromNetwork will *not* deallocate them.


    config (Callable(trt.Builder, trt.INetworkDefinition) -> trt.IBuilderConfig):
    A callable that returns a TensorRT builder configuration. If not supplied,
    a `CreateConfig` instance with default parameters is used.

Member Function Documentation

__call__()

def polygraphy.backend.trt.loader.EngineFromNetwork.__call__

(

self

)

Builds a TensorRT engine.

Returns:
    trt.ICudaEngine: The engine that was created.

Here is the call graph for this function:

Member Data Documentation

_network

polygraphy.backend.trt.loader.EngineFromNetwork._network

private

_config

polygraphy.backend.trt.loader.EngineFromNetwork._config

private

---

The documentation for this class was generated from the following file:

• backend/trt/loader.py