polygraphy.backend.trt.util Namespace Reference

Functions
def	load_plugins (plugins)
def	check_onnx_parser_errors (parser)
def	get_layer_class_mapping ()
def	get_input_metadata (network)
def	get_output_metadata (network)
def	str_from_network (network, mode="full")
def	_get_network_outputs (network)
def	check_outputs_not_found (not_found, available_outputs)
def	mark_outputs (network, outputs)
def	mark_layerwise (network)
def	unmark_outputs (network, outputs)
def	str_from_config (config)
def	check_profile (profile)
def	build_default_profile (builder, network, default_shape_value=None)
def	build_profile (builder, network, profile)
def	get_input_metadata_from_profile (profile, network)
def	add_binding_to_metadata (engine, binding, metadata)
def	get_input_metadata_from_engine (engine, start_binding, end_binding)
def	get_output_metadata_from_engine (engine, start_binding, end_binding)
def	str_from_engine (engine)
def	get_bindings_per_profile (engine)
def	get_active_profile_bindings (context)

Variables
	TRT_LOGGER = trt.Logger()

Function Documentation

load_plugins()

def polygraphy.backend.trt.util.load_plugins

(

plugins

)

check_onnx_parser_errors()

def polygraphy.backend.trt.util.check_onnx_parser_errors

(

parser

)

get_layer_class_mapping()

def polygraphy.backend.trt.util.get_layer_class_mapping

(

)

Here is the caller graph for this function:

get_input_metadata()

def polygraphy.backend.trt.util.get_input_metadata

(

network

)

Here is the caller graph for this function:

get_output_metadata()

def polygraphy.backend.trt.util.get_output_metadata

(

network

)

Here is the caller graph for this function:

str_from_network()

def polygraphy.backend.trt.util.str_from_network	(		network,
			mode = "full"
	)

Converts a TensorRT network to a human-readable representation

Args:
    network (trt.INetworkDefinition): The network.
    mode (str): Controls what is displayed for each layer. Choices: ["none", "basic", "attrs", "full"]

Returns:
    str

Here is the call graph for this function:

_get_network_outputs()

def polygraphy.backend.trt.util._get_network_outputs (   network )

private

Here is the caller graph for this function:

check_outputs_not_found()

def polygraphy.backend.trt.util.check_outputs_not_found	(		not_found,
			available_outputs
	)

Here is the caller graph for this function:

mark_outputs()

def polygraphy.backend.trt.util.mark_outputs	(		network,
			outputs
	)

Mark the specified outputs as network outputs.

Args:
    network (trt.INetworkDefinition): The network in which to mark outputs.
    outputs (Sequence[str]): The names of tensors to mark as outputs.

Here is the call graph for this function:

Here is the caller graph for this function:

mark_layerwise()

def polygraphy.backend.trt.util.mark_layerwise

(

network

)

Here is the call graph for this function:

unmark_outputs()

def polygraphy.backend.trt.util.unmark_outputs	(		network,
			outputs
	)

Here is the call graph for this function:

str_from_config()

def polygraphy.backend.trt.util.str_from_config

(

config

)

check_profile()

def polygraphy.backend.trt.util.check_profile

(

profile

)

Here is the caller graph for this function:

build_default_profile()

def polygraphy.backend.trt.util.build_default_profile	(		builder,
			network,
			default_shape_value = None
	)

Here is the call graph for this function:

build_profile()

def polygraphy.backend.trt.util.build_profile	(		builder,
			network,
			profile
	)

Here is the call graph for this function:

get_input_metadata_from_profile()

def polygraphy.backend.trt.util.get_input_metadata_from_profile	(		profile,
			network
	)

Returns metadata about the inputs based on a profile

Args:
    profile (trt.IOptimizationProfile): The profile from which to retrieve input metada.
    network (trt.INetworkDefinition): The network

Returns:
    TensorMetadata: A mapping of input names to their types and shapes.

add_binding_to_metadata()

def polygraphy.backend.trt.util.add_binding_to_metadata	(		engine,
			binding,
			metadata
	)

Here is the caller graph for this function:

get_input_metadata_from_engine()

def polygraphy.backend.trt.util.get_input_metadata_from_engine	(		engine,
			start_binding,
			end_binding
	)

Here is the call graph for this function:

Here is the caller graph for this function:

get_output_metadata_from_engine()

def polygraphy.backend.trt.util.get_output_metadata_from_engine	(		engine,
			start_binding,
			end_binding
	)

Here is the call graph for this function:

Here is the caller graph for this function:

str_from_engine()

def polygraphy.backend.trt.util.str_from_engine

(

engine

)

Here is the call graph for this function:

get_bindings_per_profile()

def polygraphy.backend.trt.util.get_bindings_per_profile

(

engine

)

Here is the caller graph for this function:

get_active_profile_bindings()

def polygraphy.backend.trt.util.get_active_profile_bindings

(

context

)

Gets the start and end binding indices for the active optimization profile.

Args:
    engine (trt.ICudaEngine): The engine in question.
    context (trt.IExecutionContext): The context where the profile is currently set.

Returns:
    Tuple[int, int]: The start and end bindings indices, in that order

Here is the call graph for this function:

Variable Documentation

TRT_LOGGER

polygraphy.backend.trt.util.TRT_LOGGER = trt.Logger()